math.big constants were no longer initialized

Fixes #3243

.github/workflows/ci.yml

@@ -3,6 +3,7 @@ on: [push, pull_request, workflow_dispatch]
 
 jobs:
   build_linux:
+    name: Ubuntu Build, Check, and Test
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v1
@@ -46,6 +47,9 @@ jobs:
       - name: Odin check examples/all for Linux i386
         run: ./odin check examples/all -vet -strict-style -target:linux_i386
         timeout-minutes: 10
+      - name: Odin check examples/all for Linux arm64
+        run: ./odin check examples/all -vet -strict-style -target:linux_arm64
+        timeout-minutes: 10
       - name: Odin check examples/all for FreeBSD amd64
         run: ./odin check examples/all -vet -strict-style -target:freebsd_amd64
         timeout-minutes: 10
@@ -53,6 +57,7 @@ jobs:
         run: ./odin check examples/all -vet -strict-style -target:openbsd_amd64
         timeout-minutes: 10
   build_macOS:
+    name: MacOS Build, Check, and Test
     runs-on: macos-latest
     steps:
       - uses: actions/checkout@v1
@@ -92,13 +97,49 @@ jobs:
           cd tests/internal
           make
         timeout-minutes: 10
-      - name: Odin check examples/all for Darwin arm64
-        run: ./odin check examples/all -vet -strict-style -target:darwin_arm64
+  build_macOS_arm:
+    name: MacOS ARM Build, Check, and Test
+    runs-on: macos-14 # This is an arm/m1 runner.
+    steps:
+      - uses: actions/checkout@v1
+      - name: Download LLVM, botan and setup PATH
+        run: |
+          brew install llvm@13 botan
+          echo "/opt/homebrew/opt/llvm@13/bin" >> $GITHUB_PATH
+          TMP_PATH=$(xcrun --show-sdk-path)/user/include
+          echo "CPATH=$TMP_PATH" >> $GITHUB_ENV
+      - name: build odin
+        run: ./build_odin.sh release
+      - name: Odin version
+        run: ./odin version
+        timeout-minutes: 1
+      - name: Odin report
+        run: ./odin report
+        timeout-minutes: 1
+      - name: Odin check
+        run: ./odin check examples/demo -vet
         timeout-minutes: 10
-      - name: Odin check examples/all for Linux arm64
-        run: ./odin check examples/all -vet -strict-style -target:linux_arm64
+      - name: Odin run
+        run: ./odin run examples/demo
+        timeout-minutes: 10
+      - name: Odin run -debug
+        run: ./odin run examples/demo -debug
+        timeout-minutes: 10
+      - name: Odin check examples/all
+        run: ./odin check examples/all -strict-style
+        timeout-minutes: 10
+      - name: Core library tests
+        run: |
+          cd tests/core
+          make
+        timeout-minutes: 10
+      - name: Odin internals tests
+        run: |
+          cd tests/internal
+          make
         timeout-minutes: 10
   build_windows:
+    name: Windows Build, Check, and Test
     runs-on: windows-2022
     steps:
       - uses: actions/checkout@v1

.github/workflows/nightly.yml

@@ -7,6 +7,7 @@ on:
 
 jobs:
   build_windows:
+    name: Windows Build
     if: github.repository == 'odin-lang/Odin'
     runs-on: windows-2022
     steps:
@@ -29,6 +30,7 @@ jobs:
           cp LICENSE dist
           cp LLVM-C.dll dist
           cp -r shared dist
+          cp -r base dist
           cp -r core dist
           cp -r vendor dist
           cp -r bin dist
@@ -39,6 +41,7 @@ jobs:
           name: windows_artifacts
           path: dist
   build_ubuntu:
+    name: Ubuntu Build
     if: github.repository == 'odin-lang/Odin'
     runs-on: ubuntu-latest
     steps:
@@ -56,46 +59,92 @@ jobs:
           cp LICENSE dist
           cp libLLVM* dist
           cp -r shared dist
+          cp -r base dist
           cp -r core dist
           cp -r vendor dist
           cp -r examples dist
+          # Zipping so executable permissions are retained, see https://github.com/actions/upload-artifact/issues/38
+          zip -r dist.zip dist
       - name: Upload artifact
         uses: actions/upload-artifact@v1
         with:
           name: ubuntu_artifacts
-          path: dist
+          path: dist.zip
   build_macos:
+    name: MacOS Build
     if: github.repository == 'odin-lang/Odin'
-    runs-on: macOS-latest
+    runs-on: macos-latest
     steps:
       - uses: actions/checkout@v1
       - name: Download LLVM and setup PATH
         run: |
-          brew install llvm@13
+          brew install llvm@13 dylibbundler
           echo "/usr/local/opt/llvm@13/bin" >> $GITHUB_PATH
           TMP_PATH=$(xcrun --show-sdk-path)/user/include
           echo "CPATH=$TMP_PATH" >> $GITHUB_ENV
       - name: build odin
-        run: make nightly
-      - name: Odin run
-        run: ./odin run examples/demo
-      - name: Copy artifacts
+        # These -L makes the linker prioritize system libraries over LLVM libraries, this is mainly to
+        # not link with libunwind bundled with LLVM but link with libunwind on the system.
+        run: CXXFLAGS="-L/usr/lib/system -L/usr/lib" make nightly
+      - name: Bundle
         run: |
           mkdir dist
           cp odin dist
           cp LICENSE dist
           cp -r shared dist
+          cp -r base dist
           cp -r core dist
           cp -r vendor dist
           cp -r examples dist
+          dylibbundler -b -x dist/odin -d dist/libs -od -p @executable_path/libs
+          # Zipping so executable permissions are retained, see https://github.com/actions/upload-artifact/issues/38
+          zip -r dist.zip dist
+      - name: Odin run
+        run: ./dist/odin run examples/demo
       - name: Upload artifact
         uses: actions/upload-artifact@v1
         with:
           name: macos_artifacts
-          path: dist
+          path: dist.zip
+  build_macos_arm:
+    name: MacOS ARM Build
+    if: github.repository == 'odin-lang/Odin'
+    runs-on: macos-14 # ARM machine
+    steps:
+      - uses: actions/checkout@v1
+      - name: Download LLVM and setup PATH
+        run: |
+          brew install llvm@13 dylibbundler
+          echo "/opt/homebrew/opt/llvm@13/bin" >> $GITHUB_PATH
+          TMP_PATH=$(xcrun --show-sdk-path)/user/include
+          echo "CPATH=$TMP_PATH" >> $GITHUB_ENV
+      - name: build odin
+        # These -L makes the linker prioritize system libraries over LLVM libraries, this is mainly to
+        # not link with libunwind bundled with LLVM but link with libunwind on the system.
+        run: CXXFLAGS="-L/usr/lib/system -L/usr/lib" make nightly
+      - name: Bundle
+        run: |
+          mkdir dist
+          cp odin dist
+          cp LICENSE dist
+          cp -r shared dist
+          cp -r base dist
+          cp -r core dist
+          cp -r vendor dist
+          cp -r examples dist
+          dylibbundler -b -x dist/odin -d dist/libs -od -p @executable_path/libs
+          # Zipping so executable permissions are retained, see https://github.com/actions/upload-artifact/issues/38
+          zip -r dist.zip dist
+      - name: Odin run
+        run: ./dist/odin run examples/demo
+      - name: Upload artifact
+        uses: actions/upload-artifact@v1
+        with:
+          name: macos_arm_artifacts
+          path: dist.zip
   upload_b2:
     runs-on: [ubuntu-latest]
-    needs: [build_windows, build_macos, build_ubuntu]
+    needs: [build_windows, build_macos, build_macos_arm, build_ubuntu]
     steps:
       - uses: actions/checkout@v1
       - uses: actions/setup-python@v2
@@ -126,6 +175,11 @@ jobs:
         with:
           name: macos_artifacts
 
+      - name: Download macOS arm artifacts
+        uses: actions/download-artifact@v1
+        with:
+          name: macos_arm_artifacts
+
       - name: Create archives and upload
         shell: bash
         env:
@@ -140,8 +194,9 @@ jobs:
           echo Uploading artifcates to B2
           chmod +x ./ci/upload_create_nightly.sh
           ./ci/upload_create_nightly.sh "$BUCKET" windows-amd64 windows_artifacts/
-          ./ci/upload_create_nightly.sh "$BUCKET" ubuntu-amd64 ubuntu_artifacts/
-          ./ci/upload_create_nightly.sh "$BUCKET" macos-amd64 macos_artifacts/
+          ./ci/upload_create_nightly.sh "$BUCKET" ubuntu-amd64 ubuntu_artifacts/dist.zip
+          ./ci/upload_create_nightly.sh "$BUCKET" macos-amd64 macos_artifacts/dist.zip
+          ./ci/upload_create_nightly.sh "$BUCKET" macos-arm64 macos_arm_artifacts/dist.zip
 
           echo Deleting old artifacts in B2
           python3 ci/delete_old_binaries.py "$BUCKET" "$DAYS_TO_KEEP"

.gitignore

@@ -28,6 +28,7 @@ tests/internal/test_map
 tests/internal/test_pow
 tests/internal/test_rtti
 tests/core/test_core_compress
+tests/core/test_core_container
 tests/core/test_core_filepath
 tests/core/test_core_fmt
 tests/core/test_core_i18n
@@ -39,7 +40,7 @@ tests/core/test_core_net
 tests/core/test_core_os_exit
 tests/core/test_core_reflect
 tests/core/test_core_strings
-tests/core/test_crypto_hash
+tests/core/test_crypto
 tests/core/test_hash
 tests/core/test_hxa
 tests/core/test_json
@@ -47,6 +48,9 @@ tests/core/test_linalg_glsl_math
 tests/core/test_noise
 tests/core/test_varint
 tests/core/test_xml
+tests/core/test_core_slice
+tests/core/test_core_thread
+tests/core/test_core_runtime
 tests/vendor/vendor_botan
 # Visual Studio 2015 cache/options directory
 .vs/
@@ -312,3 +316,6 @@ shared/
 examples/bug/
 build.sh
 !core/debug/
+
+# RAD debugger project file
+*.raddbg

base/builtin/builtin.odin

@@ -110,7 +110,7 @@ typeid_of    :: proc($T: typeid) -> typeid ---
 swizzle :: proc(x: [N]T, indices: ..int) -> [len(indices)]T ---
 
 complex    :: proc(real, imag: Float) -> Complex_Type ---
-quaternion :: proc(real, imag, jmag, kmag: Float) -> Quaternion_Type ---
+quaternion :: proc(imag, jmag, kmag, real: Float) -> Quaternion_Type --- // fields must be named
 real       :: proc(value: Complex_Or_Quaternion) -> Float ---
 imag       :: proc(value: Complex_Or_Quaternion) -> Float ---
 jmag       :: proc(value: Quaternion) -> Float ---

base/intrinsics/intrinsics.odin

@@ -5,6 +5,12 @@ package intrinsics
 // Package-Related
 is_package_imported :: proc(package_name: string) -> bool ---
 
+// Matrix Related Procedures
+transpose        :: proc(m: $T/matrix[$R, $C]$E)    -> matrix[C, R]E ---
+outer_product    :: proc(a: $A/[$X]$E, b: $B/[$Y]E) -> matrix[X, Y]E ---
+hadamard_product :: proc(a, b: $T/matrix[$R, $C]$E) -> T ---
+matrix_flatten   :: proc(m: $T/matrix[$R, $C]$E)    -> [R*C]E ---
+
 // Types
 soa_struct :: proc($N: int, $T: typeid) -> type/#soa[N]T
 

base/runtime/core.odin

@@ -18,9 +18,10 @@
 // This could change at a later date if the all these data structures are
 // implemented within the compiler rather than in this "preload" file
 //
+//+no-instrumentation
 package runtime
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 // NOTE(bill): This must match the compiler's
 Calling_Convention :: enum u8 {
@@ -180,6 +181,14 @@ Type_Info_Matrix :: struct {
 Type_Info_Soa_Pointer :: struct {
 	elem: ^Type_Info,
 }
+Type_Info_Bit_Field :: struct {
+	backing_type: ^Type_Info,
+	names:        []string,
+	types:        []^Type_Info,
+	bit_sizes:    []uintptr,
+	bit_offsets:  []uintptr,
+	tags:         []string,
+}
 
 Type_Info_Flag :: enum u8 {
 	Comparable     = 0,
@@ -222,6 +231,7 @@ Type_Info :: struct {
 		Type_Info_Relative_Multi_Pointer,
 		Type_Info_Matrix,
 		Type_Info_Soa_Pointer,
+		Type_Info_Bit_Field,
 	},
 }
 
@@ -255,6 +265,7 @@ Typeid_Kind :: enum u8 {
 	Relative_Multi_Pointer,
 	Matrix,
 	Soa_Pointer,
+	Bit_Field,
 }
 #assert(len(Typeid_Kind) < 32)
 
@@ -269,7 +280,7 @@ Typeid_Kind :: enum u8 {
 
 // NOTE(bill): only the ones that are needed (not all types)
 // This will be set by the compiler
-type_table: []Type_Info
+type_table: []^Type_Info
 
 args__: []cstring
 
@@ -295,6 +306,14 @@ Source_Code_Location :: struct {
 	procedure:    string,
 }
 
+/*
+	Used by the built-in directory `#load_directory(path: string) -> []Load_Directory_File`
+*/
+Load_Directory_File :: struct {
+	name: string,
+	data: []byte, // immutable data
+}
+
 Assertion_Failure_Proc :: #type proc(prefix, message: string, loc: Source_Code_Location) -> !
 
 // Allocation Stuff
@@ -306,6 +325,7 @@ Allocator_Mode :: enum byte {
 	Query_Features,
 	Query_Info,
 	Alloc_Non_Zeroed,
+	Resize_Non_Zeroed,
 }
 
 Allocator_Mode_Set :: distinct bit_set[Allocator_Mode]
@@ -589,7 +609,7 @@ __type_info_of :: proc "contextless" (id: typeid) -> ^Type_Info #no_bounds_check
 	if n < 0 || n >= len(type_table) {
 		n = 0
 	}
-	return &type_table[n]
+	return type_table[n]
 }
 
 when !ODIN_NO_RTTI {

base/runtime/core_builtin.odin

@@ -1,6 +1,6 @@
 package runtime
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 @builtin
 Maybe :: union($T: typeid) {T}
@@ -122,7 +122,7 @@ pop :: proc(array: ^$T/[dynamic]$E, loc := #caller_location) -> (res: E) #no_bou
 // `pop_safe` trys to remove and return the end value of dynamic array `array` and reduces the length of `array` by 1.
 // If the operation is not possible, it will return false.
 @builtin
-pop_safe :: proc(array: ^$T/[dynamic]$E) -> (res: E, ok: bool) #no_bounds_check {
+pop_safe :: proc "contextless" (array: ^$T/[dynamic]$E) -> (res: E, ok: bool) #no_bounds_check {
 	if len(array) == 0 {
 		return
 	}
@@ -148,7 +148,7 @@ pop_front :: proc(array: ^$T/[dynamic]$E, loc := #caller_location) -> (res: E) #
 // `pop_front_safe` trys to return and remove the first value of dynamic array `array` and reduces the length of `array` by 1.
 // If the operation is not possible, it will return false.
 @builtin
-pop_front_safe :: proc(array: ^$T/[dynamic]$E) -> (res: E, ok: bool) #no_bounds_check {
+pop_front_safe :: proc "contextless" (array: ^$T/[dynamic]$E) -> (res: E, ok: bool) #no_bounds_check {
 	if len(array) == 0 {
 		return
 	}
@@ -169,10 +169,16 @@ clear :: proc{clear_dynamic_array, clear_map}
 @builtin
 reserve :: proc{reserve_dynamic_array, reserve_map}
 
-// `resize` will try to resize memory of a passed dynamic array or map to the requested element count (setting the `len`, and possibly `cap`).
+@builtin
+non_zero_reserve :: proc{non_zero_reserve_dynamic_array}
+
+// `resize` will try to resize memory of a passed dynamic array to the requested element count (setting the `len`, and possibly `cap`).
 @builtin
 resize :: proc{resize_dynamic_array}
 
+@builtin
+non_zero_resize :: proc{non_zero_resize_dynamic_array}
+
 // Shrinks the capacity of a dynamic array or map down to the current length, or the given capacity.
 @builtin
 shrink :: proc{shrink_dynamic_array, shrink_map}
@@ -306,6 +312,7 @@ make_dynamic_array_len :: proc($T: typeid/[dynamic]$E, #any_int len: int, alloca
 @(builtin, require_results)
 make_dynamic_array_len_cap :: proc($T: typeid/[dynamic]$E, #any_int len: int, #any_int cap: int, allocator := context.allocator, loc := #caller_location) -> (array: T, err: Allocator_Error) #optional_allocator_error {
 	make_dynamic_array_error_loc(loc, len, cap)
+	array.allocator = allocator // initialize allocator before just in case it fails to allocate any memory
 	data := mem_alloc_bytes(size_of(E)*cap, align_of(E), allocator, loc) or_return
 	s := Raw_Dynamic_Array{raw_data(data), len, cap, allocator}
 	if data == nil && size_of(E) != 0 {
@@ -406,10 +413,7 @@ delete_key :: proc(m: ^$T/map[$K]$V, key: K) -> (deleted_key: K, deleted_value:
 	return
 }
 
-
-
-@builtin
-append_elem :: proc(array: ^$T/[dynamic]$E, arg: E, loc := #caller_location) -> (n: int, err: Allocator_Error) #optional_allocator_error {
+_append_elem :: #force_inline proc(array: ^$T/[dynamic]$E, arg: E, should_zero: bool, loc := #caller_location) -> (n: int, err: Allocator_Error) #optional_allocator_error {
 	if array == nil {
 		return 0, nil
 	}
@@ -420,7 +424,13 @@ append_elem :: proc(array: ^$T/[dynamic]$E, arg: E, loc := #caller_location) ->
 	} else {
 		if cap(array) < len(array)+1 {
 			cap := 2 * cap(array) + max(8, 1)
-			err = reserve(array, cap, loc) // do not 'or_return' here as it could be a partial success
+
+			// do not 'or_return' here as it could be a partial success
+			if should_zero {
+				err = reserve(array, cap, loc)
+			} else {
+				err = non_zero_reserve(array, cap, loc) 
+			}
 		}
 		if cap(array)-len(array) > 0 {
 			a := (^Raw_Dynamic_Array)(array)
@@ -437,7 +447,16 @@ append_elem :: proc(array: ^$T/[dynamic]$E, arg: E, loc := #caller_location) ->
 }
 
 @builtin
-append_elems :: proc(array: ^$T/[dynamic]$E, args: ..E, loc := #caller_location) -> (n: int, err: Allocator_Error) #optional_allocator_error {
+append_elem :: proc(array: ^$T/[dynamic]$E, arg: E, loc := #caller_location) -> (n: int, err: Allocator_Error) #optional_allocator_error {
+	return _append_elem(array, arg, true, loc=loc)
+}
+
+@builtin
+non_zero_append_elem :: proc(array: ^$T/[dynamic]$E, arg: E, loc := #caller_location) -> (n: int, err: Allocator_Error) #optional_allocator_error {
+	return _append_elem(array, arg, false, loc=loc)
+}
+
+_append_elems :: #force_inline proc(array: ^$T/[dynamic]$E, should_zero: bool, loc := #caller_location, args: ..E) -> (n: int, err: Allocator_Error) #optional_allocator_error {
 	if array == nil {
 		return 0, nil
 	}
@@ -454,7 +473,13 @@ append_elems :: proc(array: ^$T/[dynamic]$E, args: ..E, loc := #caller_location)
 	} else {
 		if cap(array) < len(array)+arg_len {
 			cap := 2 * cap(array) + max(8, arg_len)
-			err = reserve(array, cap, loc)  // do not 'or_return' here as it could be a partial success
+
+			// do not 'or_return' here as it could be a partial success
+			if should_zero {
+				err = reserve(array, cap, loc)
+			} else {
+				err = non_zero_reserve(array, cap, loc)
+			}
 		}
 		arg_len = min(cap(array)-len(array), arg_len)
 		if arg_len > 0 {
@@ -470,11 +495,33 @@ append_elems :: proc(array: ^$T/[dynamic]$E, args: ..E, loc := #caller_location)
 	}
 }
 
+@builtin
+append_elems :: proc(array: ^$T/[dynamic]$E, args: ..E, loc := #caller_location) -> (n: int, err: Allocator_Error) #optional_allocator_error {
+	return _append_elems(array, true, loc, ..args)
+}
+
+@builtin
+non_zero_append_elems :: proc(array: ^$T/[dynamic]$E, args: ..E, loc := #caller_location) -> (n: int, err: Allocator_Error) #optional_allocator_error {
+	return _append_elems(array, false, loc, ..args)
+}
+
 // The append_string built-in procedure appends a string to the end of a [dynamic]u8 like type
+_append_elem_string :: proc(array: ^$T/[dynamic]$E/u8, arg: $A/string, should_zero: bool, loc := #caller_location) -> (n: int, err: Allocator_Error) #optional_allocator_error {
+	args := transmute([]E)arg
+	if should_zero { 
+		return append_elems(array, ..args, loc=loc)
+	} else {
+		return non_zero_append_elems(array, ..args, loc=loc)
+	}
+}
+
 @builtin
 append_elem_string :: proc(array: ^$T/[dynamic]$E/u8, arg: $A/string, loc := #caller_location) -> (n: int, err: Allocator_Error) #optional_allocator_error {
-	args := transmute([]E)arg
-	return append_elems(array, ..args, loc=loc)
+	return _append_elem_string(array, arg, true, loc)
+}
+@builtin
+non_zero_append_elem_string :: proc(array: ^$T/[dynamic]$E/u8, arg: $A/string, loc := #caller_location) -> (n: int, err: Allocator_Error) #optional_allocator_error {
+	return _append_elem_string(array, arg, false, loc)
 }
 
 
@@ -494,6 +541,7 @@ append_string :: proc(array: ^$T/[dynamic]$E/u8, args: ..string, loc := #caller_
 
 // The append built-in procedure appends elements to the end of a dynamic array
 @builtin append :: proc{append_elem, append_elems, append_elem_string}
+@builtin non_zero_append :: proc{non_zero_append_elem, non_zero_append_elems, non_zero_append_elem_string}
 
 
 @builtin
@@ -638,8 +686,7 @@ clear_dynamic_array :: proc "contextless" (array: ^$T/[dynamic]$E) {
 // `reserve_dynamic_array` will try to reserve memory of a passed dynamic array or map to the requested element count (setting the `cap`).
 //
 // Note: Prefer the procedure group `reserve`.
-@builtin
-reserve_dynamic_array :: proc(array: ^$T/[dynamic]$E, capacity: int, loc := #caller_location) -> Allocator_Error {
+_reserve_dynamic_array :: #force_inline proc(array: ^$T/[dynamic]$E, capacity: int, should_zero: bool, loc := #caller_location) -> Allocator_Error {
 	if array == nil {
 		return nil
 	}
@@ -658,7 +705,12 @@ reserve_dynamic_array :: proc(array: ^$T/[dynamic]$E, capacity: int, loc := #cal
 	new_size  := capacity * size_of(E)
 	allocator := a.allocator
 
-	new_data := mem_resize(a.data, old_size, new_size, align_of(E), allocator, loc) or_return
+	new_data: []byte
+	if should_zero {
+		new_data = mem_resize(a.data, old_size, new_size, align_of(E), allocator, loc) or_return
+	} else {
+		new_data = non_zero_mem_resize(a.data, old_size, new_size, align_of(E), allocator, loc) or_return
+	}
 	if new_data == nil && new_size > 0 {
 		return .Out_Of_Memory
 	}
@@ -668,11 +720,20 @@ reserve_dynamic_array :: proc(array: ^$T/[dynamic]$E, capacity: int, loc := #cal
 	return nil
 }
 
+@builtin
+reserve_dynamic_array :: proc(array: ^$T/[dynamic]$E, capacity: int, loc := #caller_location) -> Allocator_Error {
+	return _reserve_dynamic_array(array, capacity, true, loc)
+}
+
+@builtin
+non_zero_reserve_dynamic_array :: proc(array: ^$T/[dynamic]$E, capacity: int, loc := #caller_location) -> Allocator_Error {
+	return _reserve_dynamic_array(array, capacity, false, loc)
+}
+
 // `resize_dynamic_array` will try to resize memory of a passed dynamic array or map to the requested element count (setting the `len`, and possibly `cap`).
 //
 // Note: Prefer the procedure group `resize`
-@builtin
-resize_dynamic_array :: proc(array: ^$T/[dynamic]$E, length: int, loc := #caller_location) -> Allocator_Error {
+_resize_dynamic_array :: #force_inline proc(array: ^$T/[dynamic]$E, length: int, should_zero: bool, loc := #caller_location) -> Allocator_Error {
 	if array == nil {
 		return nil
 	}
@@ -692,7 +753,12 @@ resize_dynamic_array :: proc(array: ^$T/[dynamic]$E, length: int, loc := #caller
 	new_size  := length * size_of(E)
 	allocator := a.allocator
 
-	new_data := mem_resize(a.data, old_size, new_size, align_of(E), allocator, loc) or_return
+	new_data : []byte
+	if should_zero {
+		new_data = mem_resize(a.data, old_size, new_size, align_of(E), allocator, loc) or_return
+	} else {
+		new_data = non_zero_mem_resize(a.data, old_size, new_size, align_of(E), allocator, loc) or_return
+	}
 	if new_data == nil && new_size > 0 {
 		return .Out_Of_Memory
 	}
@@ -703,6 +769,16 @@ resize_dynamic_array :: proc(array: ^$T/[dynamic]$E, length: int, loc := #caller
 	return nil
 }
 
+@builtin
+resize_dynamic_array :: proc(array: ^$T/[dynamic]$E, length: int, loc := #caller_location) -> Allocator_Error {
+	return _resize_dynamic_array(array, length, true, loc=loc)
+}
+
+@builtin
+non_zero_resize_dynamic_array :: proc(array: ^$T/[dynamic]$E, length: int, loc := #caller_location) -> Allocator_Error {
+	return _resize_dynamic_array(array, length, false, loc=loc)
+}
+
 /*
 	Shrinks the capacity of a dynamic array down to the current length, or the given capacity.
 
@@ -748,42 +824,23 @@ map_insert :: proc(m: ^$T/map[$K]$V, key: K, value: V, loc := #caller_location)
 	return (^V)(__dynamic_map_set_without_hash((^Raw_Map)(m), map_info(T), rawptr(&key), rawptr(&value), loc))
 }
 
-
-@builtin
-incl_elem :: proc(s: ^$S/bit_set[$E; $U], elem: E) {
-	s^ |= {elem}
+// Explicitly inserts a key and value into a map `m`, the same as `map_insert`, but the return values differ.
+// - `prev_key_ptr` will return the previous pointer of a key if it exists, and `nil` otherwise.
+// - `value_ptr` will return the pointer of the memory where the insertion happens, and `nil` if the map failed to resize
+// - `found_previous` will be true if `prev_key_ptr != nil`
+@(require_results)
+map_insert_and_check_for_previous :: proc(m: ^$T/map[$K]$V, key: K, value: V, loc := #caller_location) -> (prev_key_ptr: ^K, value_ptr: ^V, found_previous: bool) {
+	key, value := key, value
+	kp, vp := __dynamic_map_set_extra_without_hash((^Raw_Map)(m), map_info(T), rawptr(&key), rawptr(&value), loc)
+	prev_key_ptr   = (^K)(kp)
+	value_ptr      = (^V)(vp)
+	found_previous = kp != nil
+	return
 }
-@builtin
-incl_elems :: proc(s: ^$S/bit_set[$E; $U], elems: ..E) {
-	for elem in elems {
-		s^ |= {elem}
-	}
-}
-@builtin
-incl_bit_set :: proc(s: ^$S/bit_set[$E; $U], other: S) {
-	s^ |= other
-}
-@builtin
-excl_elem :: proc(s: ^$S/bit_set[$E; $U], elem: E) {
-	s^ &~= {elem}
-}
-@builtin
-excl_elems :: proc(s: ^$S/bit_set[$E; $U], elems: ..E) {
-	for elem in elems {
-		s^ &~= {elem}
-	}
-}
-@builtin
-excl_bit_set :: proc(s: ^$S/bit_set[$E; $U], other: S) {
-	s^ &~= other
-}
-
-@builtin incl :: proc{incl_elem, incl_elems, incl_bit_set}
-@builtin excl :: proc{excl_elem, excl_elems, excl_bit_set}
 
 
 @builtin
-card :: proc(s: $S/bit_set[$E; $U]) -> int {
+card :: proc "contextless" (s: $S/bit_set[$E; $U]) -> int {
 	when size_of(S) == 1 {
 		return int(intrinsics.count_ones(transmute(u8)s))
 	} else when size_of(S) == 2 {

base/runtime/core_builtin_soa.odin

@@ -1,6 +1,6 @@
 package runtime
 
-import "core:intrinsics"
+import "base:intrinsics"
 _ :: intrinsics
 
 /*

base/runtime/default_allocators_arena.odin

@@ -1,6 +1,6 @@
 package runtime
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 DEFAULT_ARENA_GROWING_MINIMUM_BLOCK_SIZE :: uint(DEFAULT_TEMP_ALLOCATOR_BACKING_SIZE)
 
@@ -195,7 +195,7 @@ arena_allocator_proc :: proc(allocator_data: rawptr, mode: Allocator_Mode,
 		err = .Mode_Not_Implemented
 	case .Free_All:
 		arena_free_all(arena, location)
-	case .Resize:
+	case .Resize, .Resize_Non_Zeroed:
 		old_data := ([^]byte)(old_memory)
 
 		switch {

base/runtime/default_allocators_general.odin

@@ -0,0 +1,12 @@
+package runtime
+
+when ODIN_DEFAULT_TO_NIL_ALLOCATOR {
+	default_allocator_proc :: nil_allocator_proc
+	default_allocator :: nil_allocator
+} else when ODIN_DEFAULT_TO_PANIC_ALLOCATOR {
+	default_allocator_proc :: panic_allocator_proc
+	default_allocator :: panic_allocator
+} else {
+	default_allocator :: heap_allocator
+	default_allocator_proc :: heap_allocator_proc
+}

base/runtime/default_allocators_nil.odin

@@ -10,7 +10,7 @@ nil_allocator_proc :: proc(allocator_data: rawptr, mode: Allocator_Mode,
 		return nil, .None
 	case .Free_All:
 		return nil, .Mode_Not_Implemented
-	case .Resize:
+	case .Resize, .Resize_Non_Zeroed:
 		if size == 0 {
 			return nil, .None
 		}
@@ -31,14 +31,6 @@ nil_allocator :: proc() -> Allocator {
 }
 
 
-
-when ODIN_OS == .Freestanding {
-	default_allocator_proc :: nil_allocator_proc
-	default_allocator :: nil_allocator
-}
-
-
-
 panic_allocator_proc :: proc(allocator_data: rawptr, mode: Allocator_Mode,
                              size, alignment: int,
                              old_memory: rawptr, old_size: int, loc := #caller_location) -> ([]byte, Allocator_Error) {
@@ -55,6 +47,10 @@ panic_allocator_proc :: proc(allocator_data: rawptr, mode: Allocator_Mode,
 		if size > 0 {
 			panic("panic allocator, .Resize called", loc=loc)
 		}
+	case .Resize_Non_Zeroed:
+		if size > 0 {
+			panic("panic allocator, .Alloc_Non_Zeroed called", loc=loc)
+		}
 	case .Free:
 		if old_memory != nil {
 			panic("panic allocator, .Free called", loc=loc)

base/runtime/docs.odin

@@ -44,7 +44,7 @@ memcpy
 memove
 
 
-## Procedures required by the LLVM backend
+## Procedures required by the LLVM backend if u128/i128 is used
 umodti3
 udivti3
 modti3
@@ -59,11 +59,12 @@ truncdfhf2
 gnu_h2f_ieee
 gnu_f2h_ieee
 extendhfsf2
+
+## Procedures required by the LLVM backend if f16 is used
 __ashlti3 // wasm specific
 __multi3  // wasm specific
 
 
-
 ## Required an entry point is defined (i.e. 'main')
 
 args__

base/runtime/dynamic_map_internal.odin

@@ -1,6 +1,6 @@
 package runtime
 
-import "core:intrinsics"
+import "base:intrinsics"
 _ :: intrinsics
 
 // High performance, cache-friendly, open-addressed Robin Hood hashing hash map
@@ -841,6 +841,33 @@ __dynamic_map_get :: proc "contextless" (#no_alias m: ^Raw_Map, #no_alias info:
 	}
 }
 
+__dynamic_map_get_key_and_value :: proc "contextless" (#no_alias m: ^Raw_Map, #no_alias info: ^Map_Info, h: Map_Hash, key: rawptr) -> (key_ptr, value_ptr: rawptr) {
+	if m.len == 0 {
+		return nil, nil
+	}
+	pos := map_desired_position(m^, h)
+	distance := uintptr(0)
+	mask := (uintptr(1) << map_log2_cap(m^)) - 1
+	ks, vs, hs, _, _ := map_kvh_data_dynamic(m^, info)
+	for {
+		element_hash := hs[pos]
+		if map_hash_is_empty(element_hash) {
+			return nil, nil
+		} else if distance > map_probe_distance(m^, element_hash, pos) {
+			return nil, nil
+		} else if element_hash == h {
+			other_key := rawptr(map_cell_index_dynamic(ks, info.ks, pos))
+			if info.key_equal(key, other_key) {
+				key_ptr   = other_key
+				value_ptr = rawptr(map_cell_index_dynamic(vs, info.vs, pos))
+				return
+			}
+		}
+		pos = (pos + 1) & mask
+		distance += 1
+	}
+}
+
 // IMPORTANT: USED WITHIN THE COMPILER
 __dynamic_map_check_grow :: proc "odin" (#no_alias m: ^Raw_Map, #no_alias info: ^Map_Info, loc := #caller_location) -> (err: Allocator_Error, has_grown: bool) {
 	if m.len >= map_resize_threshold(m^) {
@@ -874,6 +901,30 @@ __dynamic_map_set :: proc "odin" (#no_alias m: ^Raw_Map, #no_alias info: ^Map_In
 	m.len += 1
 	return rawptr(result)
 }
+__dynamic_map_set_extra_without_hash :: proc "odin" (#no_alias m: ^Raw_Map, #no_alias info: ^Map_Info, key, value: rawptr, loc := #caller_location) -> (prev_key_ptr, value_ptr: rawptr) {
+	return __dynamic_map_set_extra(m, info, info.key_hasher(key, map_seed(m^)), key, value, loc)
+}
+
+__dynamic_map_set_extra :: proc "odin" (#no_alias m: ^Raw_Map, #no_alias info: ^Map_Info, hash: Map_Hash, key, value: rawptr, loc := #caller_location) -> (prev_key_ptr, value_ptr: rawptr) {
+	if prev_key_ptr, value_ptr = __dynamic_map_get_key_and_value(m, info, hash, key); value_ptr != nil {
+		intrinsics.mem_copy_non_overlapping(value_ptr, value, info.vs.size_of_type)
+		return
+	}
+
+	hash := hash
+	err, has_grown := __dynamic_map_check_grow(m, info, loc)
+	if err != nil {
+		return nil, nil
+	}
+	if has_grown {
+		hash = info.key_hasher(key, map_seed(m^))
+	}
+
+	result := map_insert_hash_dynamic(m, info, hash, uintptr(key), uintptr(value))
+	m.len += 1
+	return nil, rawptr(result)
+}
+
 
 // IMPORTANT: USED WITHIN THE COMPILER
 @(private)

base/runtime/entry_unix.odin

@@ -1,8 +1,9 @@
 //+private
 //+build linux, darwin, freebsd, openbsd
+//+no-instrumentation
 package runtime
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 when ODIN_BUILD_MODE == .Dynamic {
 	@(link_name="_odin_entry_point", linkage="strong", require/*, link_section=".init"*/)

base/runtime/entry_wasm.odin

@@ -1,8 +1,9 @@
 //+private
 //+build wasm32, wasm64p32
+//+no-instrumentation
 package runtime
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 when !ODIN_TEST && !ODIN_NO_ENTRY_POINT {
 	@(link_name="_start", linkage="strong", require, export)

base/runtime/entry_windows.odin

@@ -1,12 +1,13 @@
 //+private
 //+build windows
+//+no-instrumentation
 package runtime
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 when ODIN_BUILD_MODE == .Dynamic {
 	@(link_name="DllMain", linkage="strong", require)
-	DllMain :: proc "stdcall" (hinstDLL: rawptr, fdwReason: u32, lpReserved: rawptr) -> b32 {
+	DllMain :: proc "system" (hinstDLL: rawptr, fdwReason: u32, lpReserved: rawptr) -> b32 {
 		context = default_context()
 
 		// Populate Windows DLL-specific global
@@ -28,7 +29,7 @@ when ODIN_BUILD_MODE == .Dynamic {
 } else when !ODIN_TEST && !ODIN_NO_ENTRY_POINT {
 	when ODIN_ARCH == .i386 || ODIN_NO_CRT {
 		@(link_name="mainCRTStartup", linkage="strong", require)
-		mainCRTStartup :: proc "stdcall" () -> i32 {
+		mainCRTStartup :: proc "system" () -> i32 {
 			context = default_context()
 			#force_no_inline _startup_runtime()
 			intrinsics.__entry_point()

base/runtime/error_checks.odin

@@ -1,5 +1,6 @@
 package runtime
 
+@(no_instrumentation)
 bounds_trap :: proc "contextless" () -> ! {
 	when ODIN_OS == .Windows {
 		windows_trap_array_bounds()
@@ -8,6 +9,7 @@ bounds_trap :: proc "contextless" () -> ! {
 	}
 }
 
+@(no_instrumentation)
 type_assertion_trap :: proc "contextless" () -> ! {
 	when ODIN_OS == .Windows {
 		windows_trap_type_assertion()
@@ -21,7 +23,7 @@ bounds_check_error :: proc "contextless" (file: string, line, column: i32, index
 	if uint(index) < uint(count) {
 		return
 	}
-	@(cold)
+	@(cold, no_instrumentation)
 	handle_error :: proc "contextless" (file: string, line, column: i32, index, count: int) -> ! {
 		print_caller_location(Source_Code_Location{file, line, column, ""})
 		print_string(" Index ")
@@ -34,6 +36,7 @@ bounds_check_error :: proc "contextless" (file: string, line, column: i32, index
 	handle_error(file, line, column, index, count)
 }
 
+@(no_instrumentation)
 slice_handle_error :: proc "contextless" (file: string, line, column: i32, lo, hi: int, len: int) -> ! {
 	print_caller_location(Source_Code_Location{file, line, column, ""})
 	print_string(" Invalid slice indices ")
@@ -46,6 +49,7 @@ slice_handle_error :: proc "contextless" (file: string, line, column: i32, lo, h
 	bounds_trap()
 }
 
+@(no_instrumentation)
 multi_pointer_slice_handle_error :: proc "contextless" (file: string, line, column: i32, lo, hi: int) -> ! {
 	print_caller_location(Source_Code_Location{file, line, column, ""})
 	print_string(" Invalid slice indices ")
@@ -82,7 +86,7 @@ dynamic_array_expr_error :: proc "contextless" (file: string, line, column: i32,
 	if 0 <= low && low <= high && high <= max {
 		return
 	}
-	@(cold)
+	@(cold, no_instrumentation)
 	handle_error :: proc "contextless" (file: string, line, column: i32, low, high, max: int) -> ! {
 		print_caller_location(Source_Code_Location{file, line, column, ""})
 		print_string(" Invalid dynamic array indices ")
@@ -103,7 +107,7 @@ matrix_bounds_check_error :: proc "contextless" (file: string, line, column: i32
 	   uint(column_index) < uint(column_count) {
 		return
 	}
-	@(cold)
+	@(cold, no_instrumentation)
 	handle_error :: proc "contextless" (file: string, line, column: i32, row_index, column_index, row_count, column_count: int) -> ! {
 		print_caller_location(Source_Code_Location{file, line, column, ""})
 		print_string(" Matrix indices [")
@@ -127,7 +131,7 @@ when ODIN_NO_RTTI {
 		if ok {
 			return
 		}
-		@(cold)
+		@(cold, no_instrumentation)
 		handle_error :: proc "contextless" (file: string, line, column: i32) -> ! {
 			print_caller_location(Source_Code_Location{file, line, column, ""})
 			print_string(" Invalid type assertion\n")
@@ -140,7 +144,7 @@ when ODIN_NO_RTTI {
 		if ok {
 			return
 		}
-		@(cold)
+		@(cold, no_instrumentation)
 		handle_error :: proc "contextless" (file: string, line, column: i32) -> ! {
 			print_caller_location(Source_Code_Location{file, line, column, ""})
 			print_string(" Invalid type assertion\n")
@@ -153,7 +157,7 @@ when ODIN_NO_RTTI {
 		if ok {
 			return
 		}
-		@(cold)
+		@(cold, no_instrumentation)
 		handle_error :: proc "contextless" (file: string, line, column: i32, from, to: typeid) -> ! {
 			print_caller_location(Source_Code_Location{file, line, column, ""})
 			print_string(" Invalid type assertion from ")
@@ -198,7 +202,7 @@ when ODIN_NO_RTTI {
 			return id
 		}
 
-		@(cold)
+		@(cold, no_instrumentation)
 		handle_error :: proc "contextless" (file: string, line, column: i32, from, to: typeid, from_data: rawptr) -> ! {
 
 			actual := variant_type(from, from_data)
@@ -224,7 +228,7 @@ make_slice_error_loc :: #force_inline proc "contextless" (loc := #caller_locatio
 	if 0 <= len {
 		return
 	}
-	@(cold)
+	@(cold, no_instrumentation)
 	handle_error :: proc "contextless" (loc: Source_Code_Location, len: int) -> ! {
 		print_caller_location(loc)
 		print_string(" Invalid slice length for make: ")
@@ -239,7 +243,7 @@ make_dynamic_array_error_loc :: #force_inline proc "contextless" (loc := #caller
 	if 0 <= len && len <= cap {
 		return
 	}
-	@(cold)
+	@(cold, no_instrumentation)
 	handle_error :: proc "contextless" (loc: Source_Code_Location, len, cap: int)  -> ! {
 		print_caller_location(loc)
 		print_string(" Invalid dynamic array parameters for make: ")
@@ -256,7 +260,7 @@ make_map_expr_error_loc :: #force_inline proc "contextless" (loc := #caller_loca
 	if 0 <= cap {
 		return
 	}
-	@(cold)
+	@(cold, no_instrumentation)
 	handle_error :: proc "contextless" (loc: Source_Code_Location, cap: int)  -> ! {
 		print_caller_location(loc)
 		print_string(" Invalid map capacity for make: ")

base/runtime/heap_allocator.odin

@@ -0,0 +1,110 @@
+package runtime
+
+import "base:intrinsics"
+
+heap_allocator :: proc() -> Allocator {
+	return Allocator{
+		procedure = heap_allocator_proc,
+		data = nil,
+	}
+}
+
+heap_allocator_proc :: proc(allocator_data: rawptr, mode: Allocator_Mode,
+                            size, alignment: int,
+                            old_memory: rawptr, old_size: int, loc := #caller_location) -> ([]byte, Allocator_Error) {
+	//
+	// NOTE(tetra, 2020-01-14): The heap doesn't respect alignment.
+	// Instead, we overallocate by `alignment + size_of(rawptr) - 1`, and insert
+	// padding. We also store the original pointer returned by heap_alloc right before
+	// the pointer we return to the user.
+	//
+
+	aligned_alloc :: proc(size, alignment: int, old_ptr: rawptr = nil, zero_memory := true) -> ([]byte, Allocator_Error) {
+		a := max(alignment, align_of(rawptr))
+		space := size + a - 1
+
+		allocated_mem: rawptr
+		if old_ptr != nil {
+			original_old_ptr := ([^]rawptr)(old_ptr)[-1]
+			allocated_mem = heap_resize(original_old_ptr, space+size_of(rawptr))
+		} else {
+			allocated_mem = heap_alloc(space+size_of(rawptr), zero_memory)
+		}
+		aligned_mem := rawptr(([^]u8)(allocated_mem)[size_of(rawptr):])
+
+		ptr := uintptr(aligned_mem)
+		aligned_ptr := (ptr - 1 + uintptr(a)) & -uintptr(a)
+		diff := int(aligned_ptr - ptr)
+		if (size + diff) > space || allocated_mem == nil {
+			return nil, .Out_Of_Memory
+		}
+
+		aligned_mem = rawptr(aligned_ptr)
+		([^]rawptr)(aligned_mem)[-1] = allocated_mem
+
+		return byte_slice(aligned_mem, size), nil
+	}
+
+	aligned_free :: proc(p: rawptr) {
+		if p != nil {
+			heap_free(([^]rawptr)(p)[-1])
+		}
+	}
+
+	aligned_resize :: proc(p: rawptr, old_size: int, new_size: int, new_alignment: int, zero_memory := true) -> (new_memory: []byte, err: Allocator_Error) {
+		if p == nil {
+			return nil, nil
+		}
+
+		new_memory = aligned_alloc(new_size, new_alignment, p, zero_memory) or_return
+
+		// NOTE: heap_resize does not zero the new memory, so we do it
+		if zero_memory && new_size > old_size {
+			new_region := raw_data(new_memory[old_size:])
+			intrinsics.mem_zero(new_region, new_size - old_size)
+		}
+		return
+	}
+
+	switch mode {
+	case .Alloc, .Alloc_Non_Zeroed:
+		return aligned_alloc(size, alignment, nil, mode == .Alloc)
+
+	case .Free:
+		aligned_free(old_memory)
+
+	case .Free_All:
+		return nil, .Mode_Not_Implemented
+
+	case .Resize, .Resize_Non_Zeroed:
+		if old_memory == nil {
+			return aligned_alloc(size, alignment, nil, mode == .Resize)
+		}
+		return aligned_resize(old_memory, old_size, size, alignment, mode == .Resize)
+
+	case .Query_Features:
+		set := (^Allocator_Mode_Set)(old_memory)
+		if set != nil {
+			set^ = {.Alloc, .Alloc_Non_Zeroed, .Free, .Resize, .Resize_Non_Zeroed, .Query_Features}
+		}
+		return nil, nil
+
+	case .Query_Info:
+		return nil, .Mode_Not_Implemented
+	}
+
+	return nil, nil
+}
+
+
+heap_alloc :: proc(size: int, zero_memory := true) -> rawptr {
+	return _heap_alloc(size, zero_memory)
+}
+
+heap_resize :: proc(ptr: rawptr, new_size: int) -> rawptr {
+	return _heap_resize(ptr, new_size)
+}
+
+heap_free :: proc(ptr: rawptr) {
+	_heap_free(ptr)
+}

base/runtime/heap_allocator_other.odin

@@ -0,0 +1,15 @@
+//+build js, wasi, freestanding, essence
+//+private
+package runtime
+
+_heap_alloc :: proc(size: int, zero_memory := true) -> rawptr {
+	unimplemented("base:runtime 'heap_alloc' procedure is not supported on this platform")
+}
+
+_heap_resize :: proc(ptr: rawptr, new_size: int) -> rawptr {
+	unimplemented("base:runtime 'heap_resize' procedure is not supported on this platform")
+}
+
+_heap_free :: proc(ptr: rawptr) {
+	unimplemented("base:runtime 'heap_free' procedure is not supported on this platform")
+}

base/runtime/heap_allocator_unix.odin

@@ -0,0 +1,38 @@
+//+build linux, darwin, freebsd, openbsd
+//+private
+package runtime
+
+when ODIN_OS == .Darwin {
+	foreign import libc "system:System.framework"
+} else {
+	foreign import libc "system:c"
+}
+
+@(default_calling_convention="c")
+foreign libc {
+	@(link_name="malloc")   _unix_malloc   :: proc(size: int) -> rawptr ---
+	@(link_name="calloc")   _unix_calloc   :: proc(num, size: int) -> rawptr ---
+	@(link_name="free")     _unix_free     :: proc(ptr: rawptr) ---
+	@(link_name="realloc")  _unix_realloc  :: proc(ptr: rawptr, size: int) -> rawptr ---
+}
+
+_heap_alloc :: proc(size: int, zero_memory := true) -> rawptr {
+	if size <= 0 {
+		return nil
+	}
+	if zero_memory {
+		return _unix_calloc(1, size)
+	} else {
+		return _unix_malloc(size)
+	}
+}
+
+_heap_resize :: proc(ptr: rawptr, new_size: int) -> rawptr {
+	// NOTE: _unix_realloc doesn't guarantee new memory will be zeroed on
+	// POSIX platforms. Ensure your caller takes this into account.
+	return _unix_realloc(ptr, new_size)
+}
+
+_heap_free :: proc(ptr: rawptr) {
+	_unix_free(ptr)
+}

base/runtime/heap_allocator_windows.odin

@@ -0,0 +1,39 @@
+package runtime
+
+foreign import kernel32 "system:Kernel32.lib"
+
+@(private="file")
+@(default_calling_convention="system")
+foreign kernel32 {
+	// NOTE(bill): The types are not using the standard names (e.g. DWORD and LPVOID) to just minimizing the dependency
+
+	// default_allocator
+	GetProcessHeap :: proc() -> rawptr ---
+	HeapAlloc      :: proc(hHeap: rawptr, dwFlags: u32, dwBytes: uint) -> rawptr ---
+	HeapReAlloc    :: proc(hHeap: rawptr, dwFlags: u32, lpMem: rawptr, dwBytes: uint) -> rawptr ---
+	HeapFree       :: proc(hHeap: rawptr, dwFlags: u32, lpMem: rawptr) -> b32 ---
+}
+
+_heap_alloc :: proc(size: int, zero_memory := true) -> rawptr {
+	HEAP_ZERO_MEMORY :: 0x00000008
+	return HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY if zero_memory else 0, uint(size))
+}
+_heap_resize :: proc(ptr: rawptr, new_size: int) -> rawptr {
+	if new_size == 0 {
+		_heap_free(ptr)
+		return nil
+	}
+	if ptr == nil {
+		return _heap_alloc(new_size)
+	}
+
+	HEAP_ZERO_MEMORY :: 0x00000008
+	return HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, ptr, uint(new_size))
+}
+_heap_free :: proc(ptr: rawptr) {
+	if ptr == nil {
+		return
+	}
+	HeapFree(GetProcessHeap(), 0, ptr)
+}
+

base/runtime/internal.odin

@@ -1,6 +1,6 @@
 package runtime
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 @(private="file")
 IS_WASM :: ODIN_ARCH == .wasm32 || ODIN_ARCH == .wasm64p32
@@ -11,7 +11,7 @@ RUNTIME_LINKAGE :: "strong" when (
 	ODIN_BUILD_MODE == .Dynamic ||
 	!ODIN_NO_CRT) &&
 	!IS_WASM) else "internal"
-RUNTIME_REQUIRE :: !ODIN_TILDE
+RUNTIME_REQUIRE :: false // !ODIN_TILDE
 
 @(private)
 __float16 :: f16 when __ODIN_LLVM_F16_SUPPORTED else u16
@@ -22,51 +22,7 @@ byte_slice :: #force_inline proc "contextless" (data: rawptr, len: int) -> []byt
 	return ([^]byte)(data)[:max(len, 0)]
 }
 
-bswap_16 :: proc "contextless" (x: u16) -> u16 {
-	return x>>8 | x<<8
-}
-
-bswap_32 :: proc "contextless" (x: u32) -> u32 {
-	return x>>24 | (x>>8)&0xff00 | (x<<8)&0xff0000 | x<<24
-}
-
-bswap_64 :: proc "contextless" (x: u64) -> u64 {
-	z := x
-	z = (z & 0x00000000ffffffff) << 32 | (z & 0xffffffff00000000) >> 32
-	z = (z & 0x0000ffff0000ffff) << 16 | (z & 0xffff0000ffff0000) >> 16
-	z = (z & 0x00ff00ff00ff00ff) << 8  | (z & 0xff00ff00ff00ff00) >> 8
-	return z
-}
-
-bswap_128 :: proc "contextless" (x: u128) -> u128 {
-	z := transmute([4]u32)x
-	z[0], z[3] = bswap_32(z[3]), bswap_32(z[0])
-	z[1], z[2] = bswap_32(z[2]), bswap_32(z[1])
-	return transmute(u128)z
-}
-
-bswap_f16 :: proc "contextless" (f: f16) -> f16 {
-	x := transmute(u16)f
-	z := bswap_16(x)
-	return transmute(f16)z
-
-}
-
-bswap_f32 :: proc "contextless" (f: f32) -> f32 {
-	x := transmute(u32)f
-	z := bswap_32(x)
-	return transmute(f32)z
-
-}
-
-bswap_f64 :: proc "contextless" (f: f64) -> f64 {
-	x := transmute(u64)f
-	z := bswap_64(x)
-	return transmute(f64)z
-}
-
-
-is_power_of_two_int :: #force_inline proc(x: int) -> bool {
+is_power_of_two_int :: #force_inline proc "contextless" (x: int) -> bool {
 	if x <= 0 {
 		return false
 	}
@@ -84,7 +40,7 @@ align_forward_int :: #force_inline proc(ptr, align: int) -> int {
 	return p
 }
 
-is_power_of_two_uintptr :: #force_inline proc(x: uintptr) -> bool {
+is_power_of_two_uintptr :: #force_inline proc "contextless" (x: uintptr) -> bool {
 	if x <= 0 {
 		return false
 	}
@@ -187,7 +143,7 @@ mem_free_all :: #force_inline proc(allocator := context.allocator, loc := #calle
 	return
 }
 
-mem_resize :: proc(ptr: rawptr, old_size, new_size: int, alignment: int = DEFAULT_ALIGNMENT, allocator := context.allocator, loc := #caller_location) -> (data: []byte, err: Allocator_Error) {
+_mem_resize :: #force_inline proc(ptr: rawptr, old_size, new_size: int, alignment: int = DEFAULT_ALIGNMENT, allocator := context.allocator, should_zero: bool, loc := #caller_location) -> (data: []byte, err: Allocator_Error) {
 	if allocator.procedure == nil {
 		return nil, nil
 	}
@@ -198,15 +154,27 @@ mem_resize :: proc(ptr: rawptr, old_size, new_size: int, alignment: int = DEFAUL
 		}
 		return
 	} else if ptr == nil {
-		return allocator.procedure(allocator.data, .Alloc, new_size, alignment, nil, 0, loc)
+		if should_zero {
+			return allocator.procedure(allocator.data, .Alloc, new_size, alignment, nil, 0, loc)
+		} else {
+			return allocator.procedure(allocator.data, .Alloc_Non_Zeroed, new_size, alignment, nil, 0, loc)
+		}
 	} else if old_size == new_size && uintptr(ptr) % uintptr(alignment) == 0 {
 		data = ([^]byte)(ptr)[:old_size]
 		return
 	}
 
-	data, err = allocator.procedure(allocator.data, .Resize, new_size, alignment, ptr, old_size, loc)
+	if should_zero {
+		data, err = allocator.procedure(allocator.data, .Resize, new_size, alignment, ptr, old_size, loc)
+	} else {
+		data, err = allocator.procedure(allocator.data, .Resize_Non_Zeroed, new_size, alignment, ptr, old_size, loc)
+	}
 	if err == .Mode_Not_Implemented {
-		data, err = allocator.procedure(allocator.data, .Alloc, new_size, alignment, nil, 0, loc)
+		if should_zero {
+			data, err = allocator.procedure(allocator.data, .Alloc, new_size, alignment, nil, 0, loc)
+		} else {
+			data, err = allocator.procedure(allocator.data, .Alloc_Non_Zeroed, new_size, alignment, nil, 0, loc)
+		}
 		if err != nil {
 			return
 		}
@@ -216,6 +184,13 @@ mem_resize :: proc(ptr: rawptr, old_size, new_size: int, alignment: int = DEFAUL
 	return
 }
 
+mem_resize :: proc(ptr: rawptr, old_size, new_size: int, alignment: int = DEFAULT_ALIGNMENT, allocator := context.allocator, loc := #caller_location) -> (data: []byte, err: Allocator_Error) {
+	return _mem_resize(ptr, old_size, new_size, alignment, allocator, true, loc)
+}
+non_zero_mem_resize :: proc(ptr: rawptr, old_size, new_size: int, alignment: int = DEFAULT_ALIGNMENT, allocator := context.allocator, loc := #caller_location) -> (data: []byte, err: Allocator_Error) {
+	return _mem_resize(ptr, old_size, new_size, alignment, allocator, false, loc)
+}
+
 memory_equal :: proc "contextless" (x, y: rawptr, n: int) -> bool {
 	switch {
 	case n == 0: return true
@@ -589,36 +564,6 @@ string_decode_last_rune :: proc "contextless" (s: string) -> (rune, int) {
 	return r, size
 }
 
-
-abs_f16 :: #force_inline proc "contextless" (x: f16) -> f16 {
-	return -x if x < 0 else x
-}
-abs_f32 :: #force_inline proc "contextless" (x: f32) -> f32 {
-	return -x if x < 0 else x
-}
-abs_f64 :: #force_inline proc "contextless" (x: f64) -> f64 {
-	return -x if x < 0 else x
-}
-
-min_f16 :: #force_inline proc "contextless" (a, b: f16) -> f16 {
-	return a if a < b else b
-}
-min_f32 :: #force_inline proc "contextless" (a, b: f32) -> f32 {
-	return a if a < b else b
-}
-min_f64 :: #force_inline proc "contextless" (a, b: f64) -> f64 {
-	return a if a < b else b
-}
-max_f16 :: #force_inline proc "contextless" (a, b: f16) -> f16 {
-	return a if a > b else b
-}
-max_f32 :: #force_inline proc "contextless" (a, b: f32) -> f32 {
-	return a if a > b else b
-}
-max_f64 :: #force_inline proc "contextless" (a, b: f64) -> f64 {
-	return a if a > b else b
-}
-
 abs_complex32 :: #force_inline proc "contextless" (x: complex32) -> f16 {
 	p, q := abs(real(x)), abs(imag(x))
 	if p < q {
@@ -730,7 +675,7 @@ mul_quaternion64 :: proc "contextless" (q, r: quaternion64) -> quaternion64 {
 	t2 := r0*q2 + r1*q3 + r2*q0 - r3*q1
 	t3 := r0*q3 - r1*q2 + r2*q1 + r3*q0
 
-	return quaternion(t0, t1, t2, t3)
+	return quaternion(w=t0, x=t1, y=t2, z=t3)
 }
 
 mul_quaternion128 :: proc "contextless" (q, r: quaternion128) -> quaternion128 {
@@ -742,7 +687,7 @@ mul_quaternion128 :: proc "contextless" (q, r: quaternion128) -> quaternion128 {
 	t2 := r0*q2 + r1*q3 + r2*q0 - r3*q1
 	t3 := r0*q3 - r1*q2 + r2*q1 + r3*q0
 
-	return quaternion(t0, t1, t2, t3)
+	return quaternion(w=t0, x=t1, y=t2, z=t3)
 }
 
 mul_quaternion256 :: proc "contextless" (q, r: quaternion256) -> quaternion256 {
@@ -754,7 +699,7 @@ mul_quaternion256 :: proc "contextless" (q, r: quaternion256) -> quaternion256 {
 	t2 := r0*q2 + r1*q3 + r2*q0 - r3*q1
 	t3 := r0*q3 - r1*q2 + r2*q1 + r3*q0
 
-	return quaternion(t0, t1, t2, t3)
+	return quaternion(w=t0, x=t1, y=t2, z=t3)
 }
 
 quo_quaternion64 :: proc "contextless" (q, r: quaternion64) -> quaternion64 {
@@ -768,7 +713,7 @@ quo_quaternion64 :: proc "contextless" (q, r: quaternion64) -> quaternion64 {
 	t2 := (r0*q2 - r1*q3 - r2*q0 + r3*q1) * invmag2
 	t3 := (r0*q3 + r1*q2 + r2*q1 - r3*q0) * invmag2
 
-	return quaternion(t0, t1, t2, t3)
+	return quaternion(w=t0, x=t1, y=t2, z=t3)
 }
 
 quo_quaternion128 :: proc "contextless" (q, r: quaternion128) -> quaternion128 {
@@ -782,7 +727,7 @@ quo_quaternion128 :: proc "contextless" (q, r: quaternion128) -> quaternion128 {
 	t2 := (r0*q2 - r1*q3 - r2*q0 + r3*q1) * invmag2
 	t3 := (r0*q3 + r1*q2 + r2*q1 - r3*q0) * invmag2
 
-	return quaternion(t0, t1, t2, t3)
+	return quaternion(w=t0, x=t1, y=t2, z=t3)
 }
 
 quo_quaternion256 :: proc "contextless" (q, r: quaternion256) -> quaternion256 {
@@ -796,7 +741,7 @@ quo_quaternion256 :: proc "contextless" (q, r: quaternion256) -> quaternion256 {
 	t2 := (r0*q2 - r1*q3 - r2*q0 + r3*q1) * invmag2
 	t3 := (r0*q3 + r1*q2 + r2*q1 - r3*q0) * invmag2
 
-	return quaternion(t0, t1, t2, t3)
+	return quaternion(w=t0, x=t1, y=t2, z=t3)
 }
 
 @(link_name="__truncsfhf2", linkage=RUNTIME_LINKAGE, require=RUNTIME_REQUIRE)
@@ -1089,3 +1034,25 @@ fixdfti :: proc(a: u64) -> i128 {
 	}
 
 }
+
+
+
+__write_bits :: proc "contextless" (dst, src: [^]byte, offset: uintptr, size: uintptr) {
+	for i in 0..<size {
+		j := offset+i
+		the_bit := byte((src[i/8]) & (1<<(i&7)) != 0)
+		b := the_bit<<(j&7)
+		dst[j/8] &~= b
+		dst[j/8] |=  b
+	}
+}
+
+__read_bits :: proc "contextless" (dst, src: [^]byte, offset: uintptr, size: uintptr) {
+	for j in 0..<size {
+		i := offset+j
+		the_bit := byte((src[i/8]) & (1<<(i&7)) != 0)
+		b := the_bit<<(j&7)
+		dst[j/8] &~= b
+		dst[j/8] |=  b
+	}
+}

base/runtime/os_specific.odin

@@ -0,0 +1,7 @@
+package runtime
+
+_OS_Errno :: distinct int
+
+stderr_write :: proc "contextless" (data: []byte) -> (int, _OS_Errno) {
+	return _stderr_write(data)
+}

base/runtime/os_specific_bsd.odin

@@ -0,0 +1,22 @@
+//+build freebsd, openbsd
+//+private
+package runtime
+
+foreign import libc "system:c"
+
+@(default_calling_convention="c")
+foreign libc {
+	@(link_name="write")
+	_unix_write :: proc(fd: i32, buf: rawptr, size: int) -> int ---
+
+	__error :: proc() -> ^i32 ---
+}
+
+_stderr_write :: proc "contextless" (data: []byte) -> (int, _OS_Errno) {
+	ret := _unix_write(2, raw_data(data), len(data))
+	if ret < len(data) {
+		err := __error()
+		return int(ret), _OS_Errno(err^ if err != nil else 0)
+	}
+	return int(ret), 0
+}

base/runtime/os_specific_darwin.odin

@@ -0,0 +1,15 @@
+//+build darwin
+//+private
+package runtime
+
+import "base:intrinsics"
+
+_stderr_write :: proc "contextless" (data: []byte) -> (int, _OS_Errno) {
+	WRITE  :: 0x20000004
+	STDERR :: 2
+	ret := intrinsics.syscall(WRITE, STDERR, uintptr(raw_data(data)), uintptr(len(data)))
+	if ret < 0 {
+		return 0, _OS_Errno(-ret)
+	}
+	return int(ret), 0
+}

base/runtime/os_specific_freestanding.odin

@@ -1,7 +1,8 @@
 //+build freestanding
+//+private
 package runtime
 
 // TODO(bill): reimplement `os.write`
-_os_write :: proc "contextless" (data: []byte) -> (int, _OS_Errno) {
+_stderr_write :: proc "contextless" (data: []byte) -> (int, _OS_Errno) {
 	return 0, -1
 }

base/runtime/os_specific_js.odin

@@ -1,9 +1,10 @@
 //+build js
+//+private
 package runtime
 
 foreign import "odin_env"
 
-_os_write :: proc "contextless" (data: []byte) -> (int, _OS_Errno) {
+_stderr_write :: proc "contextless" (data: []byte) -> (int, _OS_Errno) {
 	foreign odin_env {
 		write :: proc "contextless" (fd: u32, p: []byte) ---
 	}

base/runtime/os_specific_linux.odin

@@ -0,0 +1,24 @@
+//+private
+package runtime
+
+import "base:intrinsics"
+
+_stderr_write :: proc "contextless" (data: []byte) -> (int, _OS_Errno) {
+	when ODIN_ARCH == .amd64 {
+		SYS_write :: uintptr(1)
+	} else when ODIN_ARCH == .arm64 {
+		SYS_write :: uintptr(64)
+	} else when ODIN_ARCH == .i386 {
+		SYS_write :: uintptr(4)
+	} else when ODIN_ARCH == .arm32 {
+		SYS_write :: uintptr(4)
+	}
+
+	stderr :: 2
+
+	ret := int(intrinsics.syscall(SYS_write, uintptr(stderr), uintptr(raw_data(data)), uintptr(len(data))))
+	if ret < 0 && ret > -4096 {
+		return 0, _OS_Errno(-ret)
+	}
+	return ret, 0
+}

base/runtime/os_specific_wasi.odin

@@ -1,9 +1,10 @@
 //+build wasi
+//+private
 package runtime
 
 import "core:sys/wasm/wasi"
 
-_os_write :: proc "contextless" (data: []byte) -> (int, _OS_Errno) {
+_stderr_write :: proc "contextless" (data: []byte) -> (int, _OS_Errno) {
 	data := (wasi.ciovec_t)(data)
 	n, err := wasi.fd_write(1, {data})
 	return int(n), _OS_Errno(err)

base/runtime/os_specific_windows.odin

@@ -0,0 +1,51 @@
+//+build windows
+//+private
+package runtime
+
+foreign import kernel32 "system:Kernel32.lib"
+
+@(private="file")
+@(default_calling_convention="system")
+foreign kernel32 {
+	// NOTE(bill): The types are not using the standard names (e.g. DWORD and LPVOID) to just minimizing the dependency
+
+	// stderr_write
+	GetStdHandle         :: proc(which: u32) -> rawptr ---
+	SetHandleInformation :: proc(hObject: rawptr, dwMask: u32, dwFlags: u32) -> b32 ---
+	WriteFile            :: proc(hFile: rawptr, lpBuffer: rawptr, nNumberOfBytesToWrite: u32, lpNumberOfBytesWritten: ^u32, lpOverlapped: rawptr) -> b32 ---
+	GetLastError         :: proc() -> u32 ---
+}
+
+_stderr_write :: proc "contextless" (data: []byte) -> (n: int, err: _OS_Errno) #no_bounds_check {
+	if len(data) == 0 {
+		return 0, 0
+	}
+
+	STD_ERROR_HANDLE :: ~u32(0) -12 + 1
+	HANDLE_FLAG_INHERIT :: 0x00000001
+	MAX_RW :: 1<<30
+
+	h := GetStdHandle(STD_ERROR_HANDLE)
+	when size_of(uintptr) == 8 {
+		SetHandleInformation(h, HANDLE_FLAG_INHERIT, 0)
+	}
+
+	single_write_length: u32
+	total_write: i64
+	length := i64(len(data))
+
+	for total_write < length {
+		remaining := length - total_write
+		to_write := u32(min(i32(remaining), MAX_RW))
+
+		e := WriteFile(h, &data[total_write], to_write, &single_write_length, nil)
+		if single_write_length <= 0 || !e {
+			err = _OS_Errno(GetLastError())
+			n = int(total_write)
+			return
+		}
+		total_write += i64(single_write_length)
+	}
+	n = int(total_write)
+	return
+}

base/runtime/print.odin

@@ -123,13 +123,13 @@ encode_rune :: proc "contextless" (c: rune) -> ([4]u8, int) {
 }
 
 print_string :: proc "contextless" (str: string) -> (n: int) {
-	n, _ = os_write(transmute([]byte)str)
+	n, _ = stderr_write(transmute([]byte)str)
 	return
 }
 
 print_strings :: proc "contextless" (args: ..string) -> (n: int) {
 	for str in args {
-		m, err := os_write(transmute([]byte)str)
+		m, err := stderr_write(transmute([]byte)str)
 		n += m
 		if err != 0 {
 			break
@@ -139,7 +139,7 @@ print_strings :: proc "contextless" (args: ..string) -> (n: int) {
 }
 
 print_byte :: proc "contextless" (b: byte) -> (n: int) {
-	n, _ = os_write([]byte{b})
+	n, _ = stderr_write([]byte{b})
 	return
 }
 
@@ -178,7 +178,7 @@ print_rune :: proc "contextless" (r: rune) -> int #no_bounds_check {
 	}
 
 	b, n := encode_rune(r)
-	m, _ := os_write(b[:n])
+	m, _ := stderr_write(b[:n])
 	return m
 }
 
@@ -194,7 +194,7 @@ print_u64 :: proc "contextless" (x: u64) #no_bounds_check {
 	}
 	i -= 1; a[i] = _INTEGER_DIGITS_VAR[u % b]
 
-	os_write(a[i:])
+	stderr_write(a[i:])
 }
 
 
@@ -216,7 +216,7 @@ print_i64 :: proc "contextless" (x: i64) #no_bounds_check {
 		i -= 1; a[i] = '-'
 	}
 
-	os_write(a[i:])
+	stderr_write(a[i:])
 }
 
 print_uint    :: proc "contextless" (x: uint)    { print_u64(u64(x)) }
@@ -459,6 +459,20 @@ print_type :: proc "contextless" (ti: ^Type_Info) {
 		}
 		print_byte(']')
 
+	case Type_Info_Bit_Field:
+		print_string("bit_field ")
+		print_type(info.backing_type)
+		print_string(" {")
+		for name, i in info.names {
+			if i > 0 { print_string(", ") }
+			print_string(name)
+			print_string(": ")
+			print_type(info.types[i])
+			print_string(" | ")
+			print_u64(u64(info.bit_sizes[i]))
+		}
+		print_byte('}')
+
 
 	case Type_Info_Simd_Vector:
 		print_string("#simd[")

base/runtime/procs.odin

@@ -4,7 +4,7 @@ when ODIN_NO_CRT && ODIN_OS == .Windows {
 	foreign import lib "system:NtDll.lib"
 	
 	@(private="file")
-	@(default_calling_convention="stdcall")
+	@(default_calling_convention="system")
 	foreign lib {
 		RtlMoveMemory :: proc(dst, s: rawptr, length: int) ---
 		RtlFillMemory :: proc(dst: rawptr, length: int, fill: i32) ---

base/runtime/procs_darwin.odin

@@ -3,7 +3,7 @@ package runtime
 
 foreign import "system:Foundation.framework"
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 objc_id :: ^intrinsics.objc_object
 objc_Class :: ^intrinsics.objc_class

base/runtime/procs_windows_amd64.odin

@@ -1,11 +1,12 @@
 //+private
+//+no-instrumentation
 package runtime
 
 foreign import kernel32 "system:Kernel32.lib"
 
 @(private)
 foreign kernel32 {
-	RaiseException :: proc "stdcall" (dwExceptionCode, dwExceptionFlags, nNumberOfArguments: u32, lpArguments: ^uint) -> ! ---
+	RaiseException :: proc "system" (dwExceptionCode, dwExceptionFlags, nNumberOfArguments: u32, lpArguments: ^uint) -> ! ---
 }
 
 windows_trap_array_bounds :: proc "contextless" () -> ! {

base/runtime/procs_windows_i386.odin

@@ -1,4 +1,5 @@
 //+private
+//+no-instrumentation
 package runtime
 
 @require foreign import "system:int64.lib"
@@ -12,7 +13,7 @@ windows_trap_array_bounds :: proc "contextless" () -> ! {
 	EXCEPTION_ARRAY_BOUNDS_EXCEEDED :: 0xC000008C
 
 	foreign kernel32 {
-		RaiseException :: proc "stdcall" (dwExceptionCode, dwExceptionFlags, nNumberOfArguments: DWORD, lpArguments: ^ULONG_PTR) -> ! ---
+		RaiseException :: proc "system" (dwExceptionCode, dwExceptionFlags, nNumberOfArguments: DWORD, lpArguments: ^ULONG_PTR) -> ! ---
 	}
 
 	RaiseException(EXCEPTION_ARRAY_BOUNDS_EXCEEDED, 0, 0, nil)

base/runtime/udivmod128.odin

@@ -1,6 +1,6 @@
 package runtime
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 udivmod128 :: proc "c" (a, b: u128, rem: ^u128) -> u128 {
 	_ctz :: intrinsics.count_trailing_zeros

build_odin.sh

@@ -63,8 +63,7 @@ Darwin)
 	fi
 
 	CXXFLAGS="$CXXFLAGS $($LLVM_CONFIG --cxxflags --ldflags)"
-	LDFLAGS="$LDFLAGS -liconv -ldl -framework System"
-	LDFLAGS="$LDFLAGS -lLLVM-C"
+	LDFLAGS="$LDFLAGS -liconv -ldl -framework System -lLLVM"
 	;;
 FreeBSD)
 	CXXFLAGS="$CXXFLAGS $($LLVM_CONFIG --cxxflags --ldflags)"

ci/upload_create_nightly.sh

@@ -1,5 +1,7 @@
 #!/bin/bash
 
+set -e
+
 bucket=$1
 platform=$2
 artifact=$3
@@ -9,5 +11,15 @@ filename="odin-$platform-nightly+$now.zip"
 
 echo "Creating archive $filename from $artifact and uploading to $bucket"
 
-7z a -bd "output/$filename" -r "$artifact"
-b2 upload-file --noProgress "$bucket" "output/$filename" "nightly/$filename"
+# If this is already zipped up (done before artifact upload to keep permissions in tact), just move it.
+if [ "${artifact: -4}" == ".zip" ]
+then
+	echo "Artifact already a zip"
+	mkdir -p "output"
+	mv "$artifact" "output/$filename"
+else
+	echo "Artifact needs to be zipped"
+	7z a -bd "output/$filename" -r "$artifact"
+fi
+
+b2 upload-file --noProgress "$bucket" "output/$filename" "nightly/$filename"

core/bufio/scanner.odin

@@ -4,7 +4,7 @@ import "core:bytes"
 import "core:io"
 import "core:mem"
 import "core:unicode/utf8"
-import "core:intrinsics"
+import "base:intrinsics"
 
 // Extra errors returns by scanning procedures
 Scanner_Extra_Error :: enum i32 {

core/bytes/bytes.odin

@@ -895,7 +895,7 @@ split_multi_iterator :: proc(s: ^[]byte, substrs: [][]byte, skip_empty := false)
 
 
 
-// scrub scruvs invalid utf-8 characters and replaces them with the replacement string
+// Scrubs invalid utf-8 characters and replaces them with the replacement string
 // Adjacent invalid bytes are only replaced once
 scrub :: proc(s: []byte, replacement: []byte, allocator := context.allocator) -> []byte {
 	str := s

core/c/c.odin

@@ -1,6 +1,6 @@
 package c
 
-import builtin "core:builtin"
+import builtin "base:builtin"
 
 char           :: builtin.u8  // assuming -funsigned-char
 

core/c/libc/complex.odin

@@ -67,7 +67,7 @@ foreign libc {
 	crealf  :: proc(z: complex_float) -> float ---
 }
 
-import builtin "core:builtin"
+import builtin "base:builtin"
 
 complex_float  :: distinct builtin.complex64
 complex_double :: distinct builtin.complex128

core/c/libc/math.odin

@@ -2,7 +2,7 @@ package libc
 
 // 7.12 Mathematics
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 when ODIN_OS == .Windows {
 	foreign import libc "system:libucrt.lib"

core/c/libc/stdarg.odin

@@ -2,7 +2,7 @@ package libc
 
 // 7.16 Variable arguments
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 @(private="file")
 @(default_calling_convention="none")

core/c/libc/stdatomic.odin

@@ -2,7 +2,7 @@ package libc
 
 // 7.17 Atomics
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 ATOMIC_BOOL_LOCK_FREE     :: true
 ATOMIC_CHAR_LOCK_FREE     :: true

core/c/libc/stdio.odin

@@ -1,5 +1,7 @@
 package libc
 
+import "core:io"
+
 when ODIN_OS == .Windows {
 	foreign import libc {
 		"system:libucrt.lib",
@@ -218,3 +220,102 @@ foreign libc {
 	ferror    :: proc(stream: ^FILE) -> int ---
 	perror    :: proc(s: cstring) ---
 }
+
+to_stream :: proc(file: ^FILE) -> io.Stream {
+	stream_proc :: proc(stream_data: rawptr, mode: io.Stream_Mode, p: []byte, offset: i64, whence: io.Seek_From) -> (n: i64, err: io.Error) {
+		unknown_or_eof :: proc(f: ^FILE) -> io.Error {
+			switch {
+			case ferror(f) != 0:
+				return .Unknown
+			case feof(f) != 0:
+				return .EOF
+			case:
+				return nil
+			}
+		}
+
+		file := (^FILE)(stream_data)
+		switch mode {
+		case .Close:
+			if fclose(file) != 0 {
+				return 0, unknown_or_eof(file)
+			}
+
+		case .Flush:
+			if fflush(file) != 0 {
+				return 0, unknown_or_eof(file)
+			}
+
+		case .Read:
+			n = i64(fread(raw_data(p), size_of(byte), len(p), file))
+			if n == 0 { err = unknown_or_eof(file) }
+
+		case .Read_At:
+			curr := ftell(file)
+			if curr == -1 {
+				return 0, unknown_or_eof(file)
+			}
+
+			if fseek(file, long(offset), SEEK_SET) != 0 {
+				return 0, unknown_or_eof(file)
+			}
+
+			defer fseek(file, long(curr), SEEK_SET)
+
+			n = i64(fread(raw_data(p), size_of(byte), len(p), file))
+			if n == 0 { err = unknown_or_eof(file) }
+		
+		case .Write:
+			n = i64(fwrite(raw_data(p), size_of(byte), len(p), file))
+			if n == 0 { err = unknown_or_eof(file) }
+
+		case .Write_At:
+			curr := ftell(file)
+			if curr == -1 {
+				return 0, unknown_or_eof(file)
+			}
+
+			if fseek(file, long(offset), SEEK_SET) != 0 {
+				return 0, unknown_or_eof(file)
+			}
+
+			defer fseek(file, long(curr), SEEK_SET)
+
+			n = i64(fwrite(raw_data(p), size_of(byte), len(p), file))
+			if n == 0 { err = unknown_or_eof(file) }
+
+		case .Seek:
+			if fseek(file, long(offset), int(whence)) != 0 {
+				return 0, unknown_or_eof(file)
+			}
+		
+		case .Size:
+			curr := ftell(file)
+			if curr == -1 {
+				return 0, unknown_or_eof(file)
+			}
+			defer fseek(file, curr, SEEK_SET)
+
+			if fseek(file, 0, SEEK_END) != 0 {
+				return 0, unknown_or_eof(file)
+			}
+
+			n = i64(ftell(file))
+			if n == -1 {
+				return 0, unknown_or_eof(file)
+			}
+
+		case .Destroy:
+			return 0, .Empty
+		
+		case .Query:
+			return io.query_utility({ .Close, .Flush, .Read, .Read_At, .Write, .Write_At, .Seek, .Size })
+		}
+		return
+	}
+
+	return {
+		data      = file,
+		procedure = stream_proc,
+	}
+}

core/c/libc/string.odin

@@ -1,6 +1,6 @@
 package libc
 
-import "core:runtime"
+import "base:runtime"
 
 // 7.24 String handling
 

core/compress/common.odin

@@ -12,7 +12,7 @@ package compress
 
 import "core:io"
 import "core:bytes"
-import "core:runtime"
+import "base:runtime"
 
 /*
 	These settings bound how much compression algorithms will allocate for their output buffer.
@@ -20,10 +20,9 @@ import "core:runtime"
 
 */
 
-/*
-	When a decompression routine doesn't stream its output, but writes to a buffer,
-	we pre-allocate an output buffer to speed up decompression. The default is 1 MiB.
-*/
+
+// When a decompression routine doesn't stream its output, but writes to a buffer,
+// we pre-allocate an output buffer to speed up decompression. The default is 1 MiB.
 COMPRESS_OUTPUT_ALLOCATE_MIN :: int(#config(COMPRESS_OUTPUT_ALLOCATE_MIN, 1 << 20))
 
 /*
@@ -34,16 +33,14 @@ COMPRESS_OUTPUT_ALLOCATE_MIN :: int(#config(COMPRESS_OUTPUT_ALLOCATE_MIN, 1 << 2
 
 */
 when size_of(uintptr) == 8 {
-	/*
-		For 64-bit platforms, we set the default max buffer size to 4 GiB,
-		which is GZIP and PKZIP's max payload size.
-	*/	
+
+        // For 64-bit platforms, we set the default max buffer size to 4 GiB,
+        // which is GZIP and PKZIP's max payload size.
 	COMPRESS_OUTPUT_ALLOCATE_MAX :: int(#config(COMPRESS_OUTPUT_ALLOCATE_MAX, 1 << 32))
 } else {
-	/*
-		For 32-bit platforms, we set the default max buffer size to 512 MiB.
-	*/
-	COMPRESS_OUTPUT_ALLOCATE_MAX :: int(#config(COMPRESS_OUTPUT_ALLOCATE_MAX, 1 << 29))
+	
+	// For 32-bit platforms, we set the default max buffer size to 512 MiB.
+        COMPRESS_OUTPUT_ALLOCATE_MAX :: int(#config(COMPRESS_OUTPUT_ALLOCATE_MAX, 1 << 29))
 }
 
 
@@ -69,9 +66,8 @@ General_Error :: enum {
 	Incompatible_Options,
 	Unimplemented,
 
-	/*
-		Memory errors
-	*/
+	// Memory errors
+
 	Allocation_Failed,
 	Resize_Failed,
 }
@@ -86,17 +82,16 @@ GZIP_Error :: enum {
 	Payload_Length_Invalid,
 	Payload_CRC_Invalid,
 
-	/*
-		GZIP's payload can be a maximum of max(u32le), or 4 GiB.
-		If you tell it you expect it to contain more, that's obviously an error.
-	*/
-	Payload_Size_Exceeds_Max_Payload,
-	/*
-		For buffered instead of streamed output, the payload size can't exceed
-		the max set by the `COMPRESS_OUTPUT_ALLOCATE_MAX` switch in compress/common.odin.
+	// GZIP's payload can be a maximum of max(u32le), or 4 GiB.
+	// If you tell it you expect it to contain more, that's obviously an error.
+
+	Payload_Size_Exceeds_Max_Payload,
+
+	// For buffered instead of streamed output, the payload size can't exceed
+	// the max set by the `COMPRESS_OUTPUT_ALLOCATE_MAX` switch in compress/common.odin.
+	//
+	// You can tweak this setting using `-define:COMPRESS_OUTPUT_ALLOCATE_MAX=size_in_bytes`
 
-		You can tweak this setting using `-define:COMPRESS_OUTPUT_ALLOCATE_MAX=size_in_bytes`
-	*/
 	Output_Exceeds_COMPRESS_OUTPUT_ALLOCATE_MAX,
 
 }
@@ -137,9 +132,8 @@ Context_Memory_Input :: struct #packed {
 	code_buffer:       u64,
 	num_bits:          u64,
 
-	/*
-		If we know the data size, we can optimize the reads and writes.
-	*/
+	// If we know the data size, we can optimize the reads and writes.
+
 	size_packed:       i64,
 	size_unpacked:     i64,
 }
@@ -159,18 +153,16 @@ Context_Stream_Input :: struct #packed {
 	code_buffer:       u64,
 	num_bits:          u64,
 
-	/*
-		If we know the data size, we can optimize the reads and writes.
-	*/
+	// If we know the data size, we can optimize the reads and writes.
+
 	size_packed:       i64,
 	size_unpacked:     i64,
 
-	/*
-		Flags:
-			`input_fully_in_memory`
-				true  = This tells us we read input from `input_data` exclusively. [] = EOF.
-				false = Try to refill `input_data` from the `input` stream.
-	*/
+	// Flags:
+	// `input_fully_in_memory`
+	//   true  = This tells us we read input from `input_data` exclusively. [] = EOF.
+	//   false = Try to refill `input_data` from the `input` stream.
+
 	input_fully_in_memory: b8,
 
 	padding: [1]u8,
@@ -214,7 +206,7 @@ read_slice_from_memory :: #force_inline proc(z: ^Context_Memory_Input, size: int
 @(optimization_mode="speed")
 read_slice_from_stream :: #force_inline proc(z: ^Context_Stream_Input, size: int) -> (res: []u8, err: io.Error) {
 	// TODO: REMOVE ALL USE OF context.temp_allocator here
-	// the is literally no need for it
+	// there is literally no need for it
 	b := make([]u8, size, context.temp_allocator)
 	_ = io.read(z.input, b[:]) or_return
 	return b, nil
@@ -248,10 +240,8 @@ read_u8_from_stream :: #force_inline proc(z: ^Context_Stream_Input) -> (res: u8,
 
 read_u8 :: proc{read_u8_from_memory, read_u8_from_stream}
 
-/*
-	You would typically only use this at the end of Inflate, to drain bits from the code buffer
-	preferentially.
-*/
+// You would typically only use this at the end of Inflate, to drain bits from the code buffer
+// preferentially.
 @(optimization_mode="speed")
 read_u8_prefer_code_buffer_lsb :: #force_inline proc(z: ^$C) -> (res: u8, err: io.Error) {
 	if z.num_bits >= 8 {

core/compress/shoco/shoco.odin

@@ -11,7 +11,7 @@
 // package shoco is an implementation of the shoco short string compressor
 package shoco
 
-import "core:intrinsics"
+import "base:intrinsics"
 import "core:compress"
 
 Shoco_Pack :: struct {

core/container/avl/avl.odin

@@ -0,0 +1,678 @@
+/*
+package avl implements an AVL tree.
+
+The implementation is non-intrusive, and non-recursive.
+*/
+package container_avl
+
+import "base:intrinsics"
+import "base:runtime"
+import "core:slice"
+
+_ :: intrinsics
+_ :: runtime
+
+// Originally based on the CC0 implementation by Eric Biggers
+// See: https://github.com/ebiggers/avl_tree/
+
+// Direction specifies the traversal direction for a tree iterator.
+Direction :: enum i8 {
+	// Backward is the in-order backwards direction.
+	Backward = -1,
+	// Forward is the in-order forwards direction.
+	Forward  = 1,
+}
+
+// Ordering specifies order when inserting/finding values into the tree.
+Ordering :: slice.Ordering
+
+// Tree is an AVL tree.
+Tree :: struct($Value: typeid) {
+	// user_data is a parameter that will be passed to the on_remove
+	// callback.
+	user_data: rawptr,
+	// on_remove is an optional callback that can be called immediately
+	// after a node is removed from the tree.
+	on_remove: proc(value: Value, user_data: rawptr),
+
+	_root:           ^Node(Value),
+	_node_allocator: runtime.Allocator,
+	_cmp_fn:         proc(a, b: Value) -> Ordering,
+	_size:           int,
+}
+
+// Node is an AVL tree node.
+//
+// WARNING: It is unsafe to mutate value if the node is part of a tree
+// if doing so will alter the Node's sort position relative to other
+// elements in the tree.
+Node :: struct($Value: typeid) {
+	value: Value,
+
+	_parent:  ^Node(Value),
+	_left:    ^Node(Value),
+	_right:   ^Node(Value),
+	_balance: i8,
+}
+
+// Iterator is a tree iterator.
+//
+// WARNING: It is unsafe to modify the tree while iterating, except via
+// the iterator_remove method.
+Iterator :: struct($Value: typeid) {
+	_tree:        ^Tree(Value),
+	_cur:         ^Node(Value),
+	_next:        ^Node(Value),
+	_direction:   Direction,
+	_called_next: bool,
+}
+
+// init initializes a tree.
+init :: proc {
+	init_ordered,
+	init_cmp,
+}
+
+// init_cmp initializes a tree.
+init_cmp :: proc(
+	t: ^$T/Tree($Value),
+	cmp_fn: proc(a, b: Value) -> Ordering,
+	node_allocator := context.allocator,
+) {
+	t._root = nil
+	t._node_allocator = node_allocator
+	t._cmp_fn = cmp_fn
+	t._size = 0
+}
+
+// init_ordered initializes a tree containing ordered items, with
+// a comparison function that results in an ascending order sort.
+init_ordered :: proc(
+	t: ^$T/Tree($Value),
+	node_allocator := context.allocator,
+) where intrinsics.type_is_ordered_numeric(Value) {
+	init_cmp(t, slice.cmp_proc(Value), node_allocator)
+}
+
+// destroy de-initializes a tree.
+destroy :: proc(t: ^$T/Tree($Value), call_on_remove: bool = true) {
+	iter := iterator(t, Direction.Forward)
+	for _ in iterator_next(&iter) {
+		iterator_remove(&iter, call_on_remove)
+	}
+}
+
+// len returns the number of elements in the tree.
+len :: proc "contextless" (t: ^$T/Tree($Value)) -> int {
+	return t._size
+}
+
+// first returns the first node in the tree (in-order) or nil iff
+// the tree is empty.
+first :: proc "contextless" (t: ^$T/Tree($Value)) -> ^Node(Value) {
+	return tree_first_or_last_in_order(t, Direction.Backward)
+}
+
+// last returns the last element in the tree (in-order) or nil iff
+// the tree is empty.
+last :: proc "contextless" (t: ^$T/Tree($Value)) -> ^Node(Value) {
+	return tree_first_or_last_in_order(t, Direction.Forward)
+}
+
+// find finds the value in the tree, and returns the corresponding
+// node or nil iff the value is not present.
+find :: proc(t: ^$T/Tree($Value), value: Value) -> ^Node(Value) {
+	cur := t._root
+	descend_loop: for cur != nil {
+		switch t._cmp_fn(value, cur.value) {
+		case .Less:
+			cur = cur._left
+		case .Greater:
+			cur = cur._right
+		case .Equal:
+			break descend_loop
+		}
+	}
+
+	return cur
+}
+
+// find_or_insert attempts to insert the value into the tree, and returns
+// the node, a boolean indicating if the value was inserted, and the
+// node allocator error if relevant.  If the value is already
+// present, the existing node is returned un-altered.
+find_or_insert :: proc(
+	t: ^$T/Tree($Value),
+	value: Value,
+) -> (
+	n: ^Node(Value),
+	inserted: bool,
+	err: runtime.Allocator_Error,
+) {
+	n_ptr := &t._root
+	for n_ptr^ != nil {
+		n = n_ptr^
+		switch t._cmp_fn(value, n.value) {
+		case .Less:
+			n_ptr = &n._left
+		case .Greater:
+			n_ptr = &n._right
+		case .Equal:
+			return
+		}
+	}
+
+	parent := n
+	n = new(Node(Value), t._node_allocator) or_return
+	n.value = value
+	n._parent = parent
+	n_ptr^ = n
+	tree_rebalance_after_insert(t, n)
+
+	t._size += 1
+	inserted = true
+
+	return
+}
+
+// remove removes a node or value from the tree, and returns true iff the
+// removal was successful.  While the node's value will be left intact,
+// the node itself will be freed via the tree's node allocator.
+remove :: proc {
+	remove_value,
+	remove_node,
+}
+
+// remove_value removes a value from the tree, and returns true iff the
+// removal was successful.  While the node's value will be left intact,
+// the node itself will be freed via the tree's node allocator.
+remove_value :: proc(t: ^$T/Tree($Value), value: Value, call_on_remove: bool = true) -> bool {
+	n := find(t, value)
+	if n == nil {
+		return false
+	}
+	return remove_node(t, n, call_on_remove)
+}
+
+// remove_node removes a node from the tree, and returns true iff the
+// removal was successful.  While the node's value will be left intact,
+// the node itself will be freed via the tree's node allocator.
+remove_node :: proc(t: ^$T/Tree($Value), node: ^Node(Value), call_on_remove: bool = true) -> bool {
+	if node._parent == node || (node._parent == nil && t._root != node) {
+		return false
+	}
+	defer {
+		if call_on_remove && t.on_remove != nil {
+			t.on_remove(node.value, t.user_data)
+		}
+		free(node, t._node_allocator)
+	}
+
+	parent: ^Node(Value)
+	left_deleted: bool
+
+	t._size -= 1
+	if node._left != nil && node._right != nil {
+		parent, left_deleted = tree_swap_with_successor(t, node)
+	} else {
+		child := node._left
+		if child == nil {
+			child = node._right
+		}
+		parent = node._parent
+		if parent != nil {
+			if node == parent._left {
+				parent._left = child
+				left_deleted = true
+			} else {
+				parent._right = child
+				left_deleted = false
+			}
+			if child != nil {
+				child._parent = parent
+			}
+		} else {
+			if child != nil {
+				child._parent = parent
+			}
+			t._root = child
+			node_reset(node)
+			return true
+		}
+	}
+
+	for {
+		if left_deleted {
+			parent = tree_handle_subtree_shrink(t, parent, +1, &left_deleted)
+		} else {
+			parent = tree_handle_subtree_shrink(t, parent, -1, &left_deleted)
+		}
+		if parent == nil {
+			break
+		}
+	}
+	node_reset(node)
+
+	return true
+}
+
+// iterator returns a tree iterator in the specified direction.
+iterator :: proc "contextless" (t: ^$T/Tree($Value), direction: Direction) -> Iterator(Value) {
+	it: Iterator(Value)
+	it._tree = transmute(^Tree(Value))t
+	it._direction = direction
+
+	iterator_first(&it)
+
+	return it
+}
+
+// iterator_from_pos returns a tree iterator in the specified direction,
+// spanning the range [pos, last] (inclusive).
+iterator_from_pos :: proc "contextless" (
+	t: ^$T/Tree($Value),
+	pos: ^Node(Value),
+	direction: Direction,
+) -> Iterator(Value) {
+	it: Iterator(Value)
+	it._tree = transmute(^Tree(Value))t
+	it._direction = direction
+	it._next = nil
+	it._called_next = false
+
+	if it._cur = pos; pos != nil {
+		it._next = node_next_or_prev_in_order(it._cur, it._direction)
+	}
+
+	return it
+}
+
+// iterator_get returns the node currently pointed to by the iterator,
+// or nil iff the node has been removed, the tree is empty, or the end
+// of the tree has been reached.
+iterator_get :: proc "contextless" (it: ^$I/Iterator($Value)) -> ^Node(Value) {
+	return it._cur
+}
+
+// iterator_remove removes the node currently pointed to by the iterator,
+// and returns true iff the removal was successful.  Semantics are the
+// same as the Tree remove.
+iterator_remove :: proc(it: ^$I/Iterator($Value), call_on_remove: bool = true) -> bool {
+	if it._cur == nil {
+		return false
+	}
+
+	ok := remove_node(it._tree, it._cur, call_on_remove)
+	if ok {
+		it._cur = nil
+	}
+
+	return ok
+}
+
+// iterator_next advances the iterator and returns the (node, true) or
+// or (nil, false) iff the end of the tree has been reached.
+//
+// Note: The first call to iterator_next will return the first node instead
+// of advancing the iterator.
+iterator_next :: proc "contextless" (it: ^$I/Iterator($Value)) -> (^Node(Value), bool) {
+	// This check is needed so that the first element gets returned from
+	// a brand-new iterator, and so that the somewhat contrived case where
+	// iterator_remove is called before the first call to iterator_next
+	// returns the correct value.
+	if !it._called_next {
+		it._called_next = true
+
+		// There can be the contrived case where iterator_remove is
+		// called before ever calling iterator_next, which needs to be
+		// handled as an actual call to next.
+		//
+		// If this happens it._cur will be nil, so only return the
+		// first value, if it._cur is valid.
+		if it._cur != nil {
+			return it._cur, true
+		}
+	}
+
+	if it._next == nil {
+		return nil, false
+	}
+
+	it._cur = it._next
+	it._next = node_next_or_prev_in_order(it._cur, it._direction)
+
+	return it._cur, true
+}
+
+@(private)
+tree_first_or_last_in_order :: proc "contextless" (
+	t: ^$T/Tree($Value),
+	direction: Direction,
+) -> ^Node(Value) {
+	first, sign := t._root, i8(direction)
+	if first != nil {
+		for {
+			tmp := node_get_child(first, +sign)
+			if tmp == nil {
+				break
+			}
+			first = tmp
+		}
+	}
+
+	return first
+}
+
+@(private)
+tree_replace_child :: proc "contextless" (
+	t: ^$T/Tree($Value),
+	parent, old_child, new_child: ^Node(Value),
+) {
+	if parent != nil {
+		if old_child == parent._left {
+			parent._left = new_child
+		} else {
+			parent._right = new_child
+		}
+	} else {
+		t._root = new_child
+	}
+}
+
+@(private)
+tree_rotate :: proc "contextless" (t: ^$T/Tree($Value), a: ^Node(Value), sign: i8) {
+	b := node_get_child(a, -sign)
+	e := node_get_child(b, +sign)
+	p := a._parent
+
+	node_set_child(a, -sign, e)
+	a._parent = b
+
+	node_set_child(b, +sign, a)
+	b._parent = p
+
+	if e != nil {
+		e._parent = a
+	}
+
+	tree_replace_child(t, p, a, b)
+}
+
+@(private)
+tree_double_rotate :: proc "contextless" (
+	t: ^$T/Tree($Value),
+	b, a: ^Node(Value),
+	sign: i8,
+) -> ^Node(Value) {
+	e := node_get_child(b, +sign)
+	f := node_get_child(e, -sign)
+	g := node_get_child(e, +sign)
+	p := a._parent
+	e_bal := e._balance
+
+	node_set_child(a, -sign, g)
+	a_bal := -e_bal
+	if sign * e_bal >= 0 {
+		a_bal = 0
+	}
+	node_set_parent_balance(a, e, a_bal)
+
+	node_set_child(b, +sign, f)
+	b_bal := -e_bal
+	if sign * e_bal <= 0 {
+		b_bal = 0
+	}
+	node_set_parent_balance(b, e, b_bal)
+
+	node_set_child(e, +sign, a)
+	node_set_child(e, -sign, b)
+	node_set_parent_balance(e, p, 0)
+
+	if g != nil {
+		g._parent = a
+	}
+
+	if f != nil {
+		f._parent = b
+	}
+
+	tree_replace_child(t, p, a, e)
+
+	return e
+}
+
+@(private)
+tree_handle_subtree_growth :: proc "contextless" (
+	t: ^$T/Tree($Value),
+	node, parent: ^Node(Value),
+	sign: i8,
+) -> bool {
+	old_balance_factor := parent._balance
+	if old_balance_factor == 0 {
+		node_adjust_balance_factor(parent, sign)
+		return false
+	}
+
+	new_balance_factor := old_balance_factor + sign
+	if new_balance_factor == 0 {
+		node_adjust_balance_factor(parent, sign)
+		return true
+	}
+
+	if sign * node._balance > 0 {
+		tree_rotate(t, parent, -sign)
+		node_adjust_balance_factor(parent, -sign)
+		node_adjust_balance_factor(node, -sign)
+	} else {
+		tree_double_rotate(t, node, parent, -sign)
+	}
+
+	return true
+}
+
+@(private)
+tree_rebalance_after_insert :: proc "contextless" (t: ^$T/Tree($Value), inserted: ^Node(Value)) {
+	node, parent := inserted, inserted._parent
+	switch {
+	case parent == nil:
+		return
+	case node == parent._left:
+		node_adjust_balance_factor(parent, -1)
+	case:
+		node_adjust_balance_factor(parent, +1)
+	}
+
+	if parent._balance == 0 {
+		return
+	}
+
+	for done := false; !done; {
+		node = parent
+		if parent = node._parent; parent == nil {
+			return
+		}
+
+		if node == parent._left {
+			done = tree_handle_subtree_growth(t, node, parent, -1)
+		} else {
+			done = tree_handle_subtree_growth(t, node, parent, +1)
+		}
+	}
+}
+
+@(private)
+tree_swap_with_successor :: proc "contextless" (
+	t: ^$T/Tree($Value),
+	x: ^Node(Value),
+) -> (
+	^Node(Value),
+	bool,
+) {
+	ret: ^Node(Value)
+	left_deleted: bool
+
+	y := x._right
+	if y._left == nil {
+		ret = y
+	} else {
+		q: ^Node(Value)
+
+		for {
+			q = y
+			if y = y._left; y._left == nil {
+				break
+			}
+		}
+
+		if q._left = y._right; q._left != nil {
+			q._left._parent = q
+		}
+		y._right = x._right
+		x._right._parent = y
+		ret = q
+		left_deleted = true
+	}
+
+	y._left = x._left
+	x._left._parent = y
+
+	y._parent = x._parent
+	y._balance = x._balance
+
+	tree_replace_child(t, x._parent, x, y)
+
+	return ret, left_deleted
+}
+
+@(private)
+tree_handle_subtree_shrink :: proc "contextless" (
+	t: ^$T/Tree($Value),
+	parent: ^Node(Value),
+	sign: i8,
+	left_deleted: ^bool,
+) -> ^Node(Value) {
+	old_balance_factor := parent._balance
+	if old_balance_factor == 0 {
+		node_adjust_balance_factor(parent, sign)
+		return nil
+	}
+
+	node: ^Node(Value)
+	new_balance_factor := old_balance_factor + sign
+	if new_balance_factor == 0 {
+		node_adjust_balance_factor(parent, sign)
+		node = parent
+	} else {
+		node = node_get_child(parent, sign)
+		if sign * node._balance >= 0 {
+			tree_rotate(t, parent, -sign)
+			if node._balance == 0 {
+				node_adjust_balance_factor(node, -sign)
+				return nil
+			}
+			node_adjust_balance_factor(parent, -sign)
+			node_adjust_balance_factor(node, -sign)
+		} else {
+			node = tree_double_rotate(t, node, parent, -sign)
+		}
+	}
+
+	parent := parent
+	if parent = node._parent; parent != nil {
+		left_deleted^ = node == parent._left
+	}
+	return parent
+}
+
+@(private)
+node_reset :: proc "contextless" (n: ^Node($Value)) {
+	// Mostly pointless as n will be deleted after this is called, but
+	// attempt to be able to catch cases of n not being in the tree.
+	n._parent = n
+	n._left = nil
+	n._right = nil
+	n._balance = 0
+}
+
+@(private)
+node_set_parent_balance :: #force_inline proc "contextless" (
+	n, parent: ^Node($Value),
+	balance: i8,
+) {
+	n._parent = parent
+	n._balance = balance
+}
+
+@(private)
+node_get_child :: #force_inline proc "contextless" (n: ^Node($Value), sign: i8) -> ^Node(Value) {
+	if sign < 0 {
+		return n._left
+	}
+	return n._right
+}
+
+@(private)
+node_next_or_prev_in_order :: proc "contextless" (
+	n: ^Node($Value),
+	direction: Direction,
+) -> ^Node(Value) {
+	next, tmp: ^Node(Value)
+	sign := i8(direction)
+
+	if next = node_get_child(n, +sign); next != nil {
+		for {
+			tmp = node_get_child(next, -sign)
+			if tmp == nil {
+				break
+			}
+			next = tmp
+		}
+	} else {
+		tmp, next = n, n._parent
+		for next != nil && tmp == node_get_child(next, +sign) {
+			tmp, next = next, next._parent
+		}
+	}
+	return next
+}
+
+@(private)
+node_set_child :: #force_inline proc "contextless" (
+	n: ^Node($Value),
+	sign: i8,
+	child: ^Node(Value),
+) {
+	if sign < 0 {
+		n._left = child
+	} else {
+		n._right = child
+	}
+}
+
+@(private)
+node_adjust_balance_factor :: #force_inline proc "contextless" (n: ^Node($Value), amount: i8) {
+	n._balance += amount
+}
+
+@(private)
+iterator_first :: proc "contextless" (it: ^Iterator($Value)) {
+	// This is private because behavior when the user manually calls
+	// iterator_first followed by iterator_next is unintuitive, since
+	// the first call to iterator_next MUST return the first node
+	// instead of advancing so that `for node in iterator_next(&next)`
+	// works as expected.
+
+	switch it._direction {
+	case .Forward:
+		it._cur = tree_first_or_last_in_order(it._tree, .Backward)
+	case .Backward:
+		it._cur = tree_first_or_last_in_order(it._tree, .Forward)
+	}
+
+	it._next = nil
+	it._called_next = false
+
+	if it._cur != nil {
+		it._next = node_next_or_prev_in_order(it._cur, it._direction)
+	}
+}

core/container/bit_array/bit_array.odin

@@ -1,6 +1,6 @@
 package dynamic_bit_array
 
-import "core:intrinsics"
+import "base:intrinsics"
 import "core:mem"
 
 /*

core/container/intrusive/list/intrusive_list.odin

@@ -1,6 +1,6 @@
 package container_intrusive_list
 
-import "core:intrinsics"
+import "base:intrinsics"
 
 // An intrusive doubly-linked list
 //

core/container/lru/lru_cache.odin

@@ -1,7 +1,7 @@
 package container_lru
 
-import "core:runtime"
-import "core:intrinsics"
+import "base:runtime"
+import "base:intrinsics"
 _ :: runtime
 _ :: intrinsics
 

core/container/priority_queue/priority_queue.odin

@@ -1,6 +1,6 @@
 package container_priority_queue
 
-import "core:builtin"
+import "base:builtin"
 
 Priority_Queue :: struct($T: typeid) {
 	queue: [dynamic]T,
@@ -140,3 +140,18 @@ remove :: proc(pq: ^$Q/Priority_Queue($T), i: int) -> (value: T, ok: bool) {
 	return
 }
 
+peek_safe :: proc(pq: $Q/Priority_Queue($T), loc := #caller_location) -> (res: T, ok: bool) {
+	if builtin.len(pq.queue) > 0 {
+		return pq.queue[0], true
+	}
+	return
+}
+
+peek :: proc(pq: $Q/Priority_Queue($T), loc := #caller_location) -> (res: T) {
+	assert(condition=builtin.len(pq.queue)>0, loc=loc)
+
+	if builtin.len(pq.queue) > 0 {
+		return pq.queue[0]
+	}
+	return
+}

core/container/queue/queue.odin

@@ -1,7 +1,7 @@
 package container_queue
 
-import "core:builtin"
-import "core:runtime"
+import "base:builtin"
+import "base:runtime"
 _ :: runtime
 
 // Dynamically resizable double-ended queue/ring-buffer

core/container/small_array/small_array.odin

@@ -1,7 +1,7 @@
 package container_small_array
 
-import "core:builtin"
-import "core:runtime"
+import "base:builtin"
+import "base:runtime"
 _ :: runtime
 
 Small_Array :: struct($N: int, $T: typeid) where N >= 0 {

core/container/topological_sort/topological_sort.odin

@@ -3,8 +3,8 @@
 // map type is being used to accelerate lookups.
 package container_topological_sort
 
-import "core:intrinsics"
-import "core:runtime"
+import "base:intrinsics"
+import "base:runtime"
 _ :: intrinsics
 _ :: runtime
 

core/crypto/README.md

@@ -1,84 +1,22 @@
 # crypto
 
-A cryptography library for the Odin language
+A cryptography library for the Odin language.
 
 ## Supported
 
-This library offers various algorithms implemented in Odin.
-Please see the chart below for some of the options.
-
-## Hashing algorithms
-
-| Algorithm                                                                                                    |                  |
-|:-------------------------------------------------------------------------------------------------------------|:-----------------|
-| [BLAKE2B](https://datatracker.ietf.org/doc/html/rfc7693)                                                     | ✔️ |
-| [BLAKE2S](https://datatracker.ietf.org/doc/html/rfc7693)                                                     | ✔️ |
-| [SHA-2](https://csrc.nist.gov/csrc/media/publications/fips/180/2/archive/2002-08-01/documents/fips180-2.pdf) | ✔️ |
-| [SHA-3](https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.202.pdf)                                            | ✔️ |
-| [SHAKE](https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.202.pdf)                                            | ✔️ |
-| [SM3](https://datatracker.ietf.org/doc/html/draft-sca-cfrg-sm3-02)                                           | ✔️ |
-| legacy/[Keccak](https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.202.pdf)                                    | ✔️ |
-| legacy/[MD5](https://datatracker.ietf.org/doc/html/rfc1321)                                                  | ✔️ |
-| legacy/[SHA-1](https://datatracker.ietf.org/doc/html/rfc3174)                                                | ✔️ |
-
-#### High level API
-
-Each hash algorithm contains a procedure group named `hash`, or if the algorithm provides more than one digest size `hash_<size>`\*.
-Included in these groups are six procedures.
-- `hash_string` - Hash a given string and return the computed hash. Just calls `hash_bytes` internally
-- `hash_bytes` - Hash a given byte slice and return the computed hash
-- `hash_string_to_buffer` - Hash a given string and put the computed hash in the second proc parameter. Just calls `hash_bytes_to_buffer` internally
-- `hash_bytes_to_buffer` - Hash a given string and put the computed hash in the second proc parameter. The destination buffer has to be at least as big as the digest size of the hash
-- `hash_stream` - Takes a stream from io.Stream and returns the computed hash from it
-- `hash_file` - Takes a file handle and returns the computed hash from it. A second optional boolean parameter controls if the file is streamed (this is the default) or read at once (set to true)
-
-\* On some algorithms there is another part to the name, since they might offer control about additional parameters.
-For instance, `SHA-2` offers different sizes.
-Computing a 512-bit hash is therefore achieved by calling `sha2.hash_512(...)`.
-
-#### Low level API
-
-The above mentioned procedures internally call three procedures: `init`, `update` and `final`.
-You may also directly call them, if you wish.
-
-#### Example
-
-```odin
-package crypto_example
-
-// Import the desired package
-import "core:crypto/blake2b"
-
-main :: proc() {
-    input := "foo"
-
-    // Compute the hash, using the high level API
-    computed_hash := blake2b.hash(input)
-
-    // Variant that takes a destination buffer, instead of returning the computed hash
-    hash := make([]byte, sha2.DIGEST_SIZE) // @note: Destination buffer has to be at least as big as the digest size of the hash
-    blake2b.hash(input, hash[:])
-
-    // Compute the hash, using the low level API
-    ctx: blake2b.Context
-    computed_hash_low: [blake2b.DIGEST_SIZE]byte
-    blake2b.init(&ctx)
-    blake2b.update(&ctx, transmute([]byte)input)
-    blake2b.final(&ctx, computed_hash_low[:])
-}
-```
-For example uses of all available algorithms, please see the tests within `tests/core/crypto`.
+This package offers various algorithms implemented in Odin, along with
+useful helpers such as access to the system entropy source, and a
+constant-time byte comparison.
 
 ## Implementation considerations
 
 - The crypto packages are not thread-safe.
 - Best-effort is make to mitigate timing side-channels on reasonable
-  architectures. Architectures that are known to be unreasonable include
+  architectures.  Architectures that are known to be unreasonable include
   but are not limited to i386, i486, and WebAssembly.
-- Some but not all of the packages attempt to santize sensitive data,
-  however this is not done consistently through the library at the moment.
-  As Thomas Pornin puts it "In general, such memory cleansing is a fool's
-  quest."
+- The packages attempt to santize sensitive data, however this is, and
+  will remain a "best-effort" implementation decision.  As Thomas Pornin
+  puts it "In general, such memory cleansing is a fool's quest."
 - All of these packages have not received independent third party review.
 
 ## License

core/crypto/_blake2/blake2.odin

@@ -11,6 +11,7 @@ package _blake2
 */
 
 import "core:encoding/endian"
+import "core:mem"
 
 BLAKE2S_BLOCK_SIZE :: 64
 BLAKE2S_SIZE :: 32
@@ -28,7 +29,6 @@ Blake2s_Context :: struct {
 	is_keyed:     bool,
 	size:         byte,
 	is_last_node: bool,
-	cfg:          Blake2_Config,
 
 	is_initialized: bool,
 }
@@ -44,7 +44,6 @@ Blake2b_Context :: struct {
 	is_keyed:     bool,
 	size:         byte,
 	is_last_node: bool,
-	cfg:          Blake2_Config,
 
 	is_initialized: bool,
 }
@@ -83,62 +82,61 @@ BLAKE2B_IV := [8]u64 {
 	0x1f83d9abfb41bd6b, 0x5be0cd19137e2179,
 }
 
-init :: proc(ctx: ^$T) {
+init :: proc(ctx: ^$T, cfg: ^Blake2_Config) {
 	when T == Blake2s_Context {
-		block_size :: BLAKE2S_BLOCK_SIZE
 		max_size :: BLAKE2S_SIZE
 	} else when T == Blake2b_Context {
-		block_size :: BLAKE2B_BLOCK_SIZE
 		max_size :: BLAKE2B_SIZE
 	}
 
-	if ctx.cfg.size > max_size {
+	if cfg.size > max_size {
 		panic("blake2: requested output size exceeeds algorithm max")
 	}
 
-	p := make([]byte, block_size)
-	defer delete(p)
+	// To save having to allocate a scratch buffer, use the internal
+	// data buffer (`ctx.x`), as it is exactly the correct size.
+	p := ctx.x[:]
 
-	p[0] = ctx.cfg.size
-	p[1] = byte(len(ctx.cfg.key))
+	p[0] = cfg.size
+	p[1] = byte(len(cfg.key))
 
-	if ctx.cfg.salt != nil {
+	if cfg.salt != nil {
 		when T == Blake2s_Context {
-			copy(p[16:], ctx.cfg.salt)
+			copy(p[16:], cfg.salt)
 		} else when T == Blake2b_Context {
-			copy(p[32:], ctx.cfg.salt)
+			copy(p[32:], cfg.salt)
 		}
 	}
-	if ctx.cfg.person != nil {
+	if cfg.person != nil {
 		when T == Blake2s_Context {
-			copy(p[24:], ctx.cfg.person)
+			copy(p[24:], cfg.person)
 		} else when T == Blake2b_Context {
-			copy(p[48:], ctx.cfg.person)
+			copy(p[48:], cfg.person)
 		}
 	}
 
-	if ctx.cfg.tree != nil {
-		p[2] = ctx.cfg.tree.(Blake2_Tree).fanout
-		p[3] = ctx.cfg.tree.(Blake2_Tree).max_depth
-		endian.unchecked_put_u32le(p[4:], ctx.cfg.tree.(Blake2_Tree).leaf_size)
+	if cfg.tree != nil {
+		p[2] = cfg.tree.(Blake2_Tree).fanout
+		p[3] = cfg.tree.(Blake2_Tree).max_depth
+		endian.unchecked_put_u32le(p[4:], cfg.tree.(Blake2_Tree).leaf_size)
 		when T == Blake2s_Context {
-			p[8] = byte(ctx.cfg.tree.(Blake2_Tree).node_offset)
-			p[9] = byte(ctx.cfg.tree.(Blake2_Tree).node_offset >> 8)
-			p[10] = byte(ctx.cfg.tree.(Blake2_Tree).node_offset >> 16)
-			p[11] = byte(ctx.cfg.tree.(Blake2_Tree).node_offset >> 24)
-			p[12] = byte(ctx.cfg.tree.(Blake2_Tree).node_offset >> 32)
-			p[13] = byte(ctx.cfg.tree.(Blake2_Tree).node_offset >> 40)
-			p[14] = ctx.cfg.tree.(Blake2_Tree).node_depth
-			p[15] = ctx.cfg.tree.(Blake2_Tree).inner_hash_size
+			p[8] = byte(cfg.tree.(Blake2_Tree).node_offset)
+			p[9] = byte(cfg.tree.(Blake2_Tree).node_offset >> 8)
+			p[10] = byte(cfg.tree.(Blake2_Tree).node_offset >> 16)
+			p[11] = byte(cfg.tree.(Blake2_Tree).node_offset >> 24)
+			p[12] = byte(cfg.tree.(Blake2_Tree).node_offset >> 32)
+			p[13] = byte(cfg.tree.(Blake2_Tree).node_offset >> 40)
+			p[14] = cfg.tree.(Blake2_Tree).node_depth
+			p[15] = cfg.tree.(Blake2_Tree).inner_hash_size
 		} else when T == Blake2b_Context {
-			endian.unchecked_put_u64le(p[8:], ctx.cfg.tree.(Blake2_Tree).node_offset)
-			p[16] = ctx.cfg.tree.(Blake2_Tree).node_depth
-			p[17] = ctx.cfg.tree.(Blake2_Tree).inner_hash_size
+			endian.unchecked_put_u64le(p[8:], cfg.tree.(Blake2_Tree).node_offset)
+			p[16] = cfg.tree.(Blake2_Tree).node_depth
+			p[17] = cfg.tree.(Blake2_Tree).inner_hash_size
 		}
 	} else {
 		p[2], p[3] = 1, 1
 	}
-	ctx.size = ctx.cfg.size
+	ctx.size = cfg.size
 	for i := 0; i < 8; i += 1 {
 		when T == Blake2s_Context {
 			ctx.h[i] = BLAKE2S_IV[i] ~ endian.unchecked_get_u32le(p[i * 4:])
@@ -147,11 +145,14 @@ init :: proc(ctx: ^$T) {
 			ctx.h[i] = BLAKE2B_IV[i] ~ endian.unchecked_get_u64le(p[i * 8:])
 		}
 	}
-	if ctx.cfg.tree != nil && ctx.cfg.tree.(Blake2_Tree).is_last_node {
+
+	mem.zero(&ctx.x, size_of(ctx.x)) // Done with the scratch space, no barrier.
+
+	if cfg.tree != nil && cfg.tree.(Blake2_Tree).is_last_node {
 		ctx.is_last_node = true
 	}
-	if len(ctx.cfg.key) > 0 {
-		copy(ctx.padded_key[:], ctx.cfg.key)
+	if len(cfg.key) > 0 {
+		copy(ctx.padded_key[:], cfg.key)
 		update(ctx, ctx.padded_key[:])
 		ctx.is_keyed = true
 	}
@@ -194,22 +195,40 @@ update :: proc(ctx: ^$T, p: []byte) {
 	ctx.nx += copy(ctx.x[ctx.nx:], p)
 }
 
-final :: proc(ctx: ^$T, hash: []byte) {
+final :: proc(ctx: ^$T, hash: []byte, finalize_clone: bool = false) {
 	assert(ctx.is_initialized)
 
+	ctx := ctx
+	if finalize_clone {
+		tmp_ctx: T
+		clone(&tmp_ctx, ctx)
+		ctx = &tmp_ctx
+	}
+	defer(reset(ctx))
+
 	when T == Blake2s_Context {
-		if len(hash) < int(ctx.cfg.size) {
+		if len(hash) < int(ctx.size) {
 			panic("crypto/blake2s: invalid destination digest size")
 		}
 		blake2s_final(ctx, hash)
 	} else when T == Blake2b_Context {
-		if len(hash) < int(ctx.cfg.size) {
+		if len(hash) < int(ctx.size) {
 			panic("crypto/blake2b: invalid destination digest size")
 		}
 		blake2b_final(ctx, hash)
 	}
+}
 
-	ctx.is_initialized = false
+clone :: proc(ctx, other: ^$T) {
+	ctx^ = other^
+}
+
+reset :: proc(ctx: ^$T) {
+	if !ctx.is_initialized {
+		return
+	}
+
+	mem.zero_explicit(ctx, size_of(ctx^))
 }
 
 @(private)

core/crypto/_sha3/sha3.odin

@@ -12,10 +12,16 @@ package _sha3
 */
 
 import "core:math/bits"
+import "core:mem"
 
 ROUNDS :: 24
 
-Sha3_Context :: struct {
+RATE_224 :: 1152 / 8
+RATE_256 :: 1088 / 8
+RATE_384 :: 832 / 8
+RATE_512 :: 576 / 8
+
+Context :: struct {
 	st:        struct #raw_union {
 		b: [200]u8,
 		q: [25]u64,
@@ -103,81 +109,100 @@ keccakf :: proc "contextless" (st: ^[25]u64) {
 	}
 }
 
-init :: proc(c: ^Sha3_Context) {
+init :: proc(ctx: ^Context) {
 	for i := 0; i < 25; i += 1 {
-		c.st.q[i] = 0
+		ctx.st.q[i] = 0
 	}
-	c.rsiz = 200 - 2 * c.mdlen
-	c.pt = 0
+	ctx.rsiz = 200 - 2 * ctx.mdlen
+	ctx.pt = 0
 
-	c.is_initialized = true
-	c.is_finalized = false
+	ctx.is_initialized = true
+	ctx.is_finalized = false
 }
 
-update :: proc(c: ^Sha3_Context, data: []byte) {
-	assert(c.is_initialized)
-	assert(!c.is_finalized)
+update :: proc(ctx: ^Context, data: []byte) {
+	assert(ctx.is_initialized)
+	assert(!ctx.is_finalized)
 
-	j := c.pt
+	j := ctx.pt
 	for i := 0; i < len(data); i += 1 {
-		c.st.b[j] ~= data[i]
+		ctx.st.b[j] ~= data[i]
 		j += 1
-		if j >= c.rsiz {
-			keccakf(&c.st.q)
+		if j >= ctx.rsiz {
+			keccakf(&ctx.st.q)
 			j = 0
 		}
 	}
-	c.pt = j
+	ctx.pt = j
 }
 
-final :: proc(c: ^Sha3_Context, hash: []byte) {
-	assert(c.is_initialized)
+final :: proc(ctx: ^Context, hash: []byte, finalize_clone: bool = false) {
+	assert(ctx.is_initialized)
 
-	if len(hash) < c.mdlen {
-		if c.is_keccak {
+	if len(hash) < ctx.mdlen {
+		if ctx.is_keccak {
 			panic("crypto/keccac: invalid destination digest size")
 		}
 		panic("crypto/sha3: invalid destination digest size")
 	}
-	if c.is_keccak {
-		c.st.b[c.pt] ~= 0x01
+
+	ctx := ctx
+	if finalize_clone {
+		tmp_ctx: Context
+		clone(&tmp_ctx, ctx)
+		ctx = &tmp_ctx
+	}
+	defer(reset(ctx))
+
+	if ctx.is_keccak {
+		ctx.st.b[ctx.pt] ~= 0x01
 	} else {
-		c.st.b[c.pt] ~= 0x06
+		ctx.st.b[ctx.pt] ~= 0x06
 	}
 
-	c.st.b[c.rsiz - 1] ~= 0x80
-	keccakf(&c.st.q)
-	for i := 0; i < c.mdlen; i += 1 {
-		hash[i] = c.st.b[i]
+	ctx.st.b[ctx.rsiz - 1] ~= 0x80
+	keccakf(&ctx.st.q)
+	for i := 0; i < ctx.mdlen; i += 1 {
+		hash[i] = ctx.st.b[i]
+	}
+}
+
+clone :: proc(ctx, other: ^Context) {
+	ctx^ = other^
+}
+
+reset :: proc(ctx: ^Context) {
+	if !ctx.is_initialized {
+		return
 	}
 
-	c.is_initialized = false // No more absorb, no more squeeze.
+	mem.zero_explicit(ctx, size_of(ctx^))
 }
 
-shake_xof :: proc(c: ^Sha3_Context) {
-	assert(c.is_initialized)
-	assert(!c.is_finalized)
+shake_xof :: proc(ctx: ^Context) {
+	assert(ctx.is_initialized)
+	assert(!ctx.is_finalized)
 
-	c.st.b[c.pt] ~= 0x1F
-	c.st.b[c.rsiz - 1] ~= 0x80
-	keccakf(&c.st.q)
-	c.pt = 0
+	ctx.st.b[ctx.pt] ~= 0x1F
+	ctx.st.b[ctx.rsiz - 1] ~= 0x80
+	keccakf(&ctx.st.q)
+	ctx.pt = 0
 
-	c.is_finalized = true // No more absorb, unlimited squeeze.
+	ctx.is_finalized = true // No more absorb, unlimited squeeze.
 }
 
-shake_out :: proc(c: ^Sha3_Context, hash: []byte) {
-	assert(c.is_initialized)
-	assert(c.is_finalized)
+shake_out :: proc(ctx: ^Context, hash: []byte) {
+	assert(ctx.is_initialized)
+	assert(ctx.is_finalized)
 
-	j := c.pt
+	j := ctx.pt
 	for i := 0; i < len(hash); i += 1 {
-		if j >= c.rsiz {
-			keccakf(&c.st.q)
+		if j >= ctx.rsiz {
+			keccakf(&ctx.st.q)
 			j = 0
 		}
-		hash[i] = c.st.b[j]
+		hash[i] = ctx.st.b[j]
 		j += 1
 	}
-	c.pt = j
+	ctx.pt = j
 }

core/crypto/blake2b/blake2b.odin

@@ -1,3 +1,10 @@
+/*
+package blake2b implements the BLAKE2b hash algorithm.
+
+See:
+- https://datatracker.ietf.org/doc/html/rfc7693
+- https://www.blake2.net
+*/
 package blake2b
 
 /*
@@ -6,122 +13,47 @@ package blake2b
 
     List of contributors:
         zhibog, dotbmp:  Initial implementation.
-
-    Interface for the BLAKE2b hashing algorithm.
-    BLAKE2b and BLAKE2s share the implementation in the _blake2 package.
 */
 
-import "core:io"
-import "core:os"
-
 import "../_blake2"
 
-/*
-    High level API
-*/
-
+// DIGEST_SIZE is the BLAKE2b digest size in bytes.
 DIGEST_SIZE :: 64
 
-// hash_string will hash the given input and return the
-// computed hash
-hash_string :: proc(data: string) -> [DIGEST_SIZE]byte {
-	return hash_bytes(transmute([]byte)(data))
-}
-
-// hash_bytes will hash the given input and return the
-// computed hash
-hash_bytes :: proc(data: []byte) -> [DIGEST_SIZE]byte {
-	hash: [DIGEST_SIZE]byte
-	ctx: Context
-	cfg: _blake2.Blake2_Config
-	cfg.size = _blake2.BLAKE2B_SIZE
-	ctx.cfg = cfg
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-	return hash
-}
-
-// hash_string_to_buffer will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer :: proc(data: string, hash: []byte) {
-	hash_bytes_to_buffer(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer :: proc(data, hash: []byte) {
-	ctx: Context
-	cfg: _blake2.Blake2_Config
-	cfg.size = _blake2.BLAKE2B_SIZE
-	ctx.cfg = cfg
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash)
-}
-
-// hash_stream will read the stream in chunks and compute a
-// hash from its contents
-hash_stream :: proc(s: io.Stream) -> ([DIGEST_SIZE]byte, bool) {
-	hash: [DIGEST_SIZE]byte
-	ctx: Context
-	cfg: _blake2.Blake2_Config
-	cfg.size = _blake2.BLAKE2B_SIZE
-	ctx.cfg = cfg
-	init(&ctx)
-
-	buf := make([]byte, 512)
-	defer delete(buf)
-
-	read := 1
-	for read > 0 {
-		read, _ = io.read(s, buf)
-		if read > 0 {
-			update(&ctx, buf[:read])
-		}
-	}
-	final(&ctx, hash[:])
-	return hash, true
-}
-
-// hash_file will read the file provided by the given handle
-// and compute a hash
-hash_file :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE]byte, bool) {
-	if !load_at_once {
-		return hash_stream(os.stream_from_handle(hd))
-	} else {
-		if buf, ok := os.read_entire_file(hd); ok {
-			return hash_bytes(buf[:]), ok
-		}
-	}
-	return [DIGEST_SIZE]byte{}, false
-}
-
-hash :: proc {
-	hash_stream,
-	hash_file,
-	hash_bytes,
-	hash_string,
-	hash_bytes_to_buffer,
-	hash_string_to_buffer,
-}
-
-/*
-    Low level API
-*/
+// BLOCK_SIZE is the BLAKE2b block size in bytes.
+BLOCK_SIZE :: _blake2.BLAKE2B_BLOCK_SIZE
 
+// Context is a BLAKE2b instance.
 Context :: _blake2.Blake2b_Context
 
+// init initializes a Context with the default BLAKE2b config.
 init :: proc(ctx: ^Context) {
-	_blake2.init(ctx)
+	cfg: _blake2.Blake2_Config
+	cfg.size = _blake2.BLAKE2B_SIZE
+	_blake2.init(ctx, &cfg)
 }
 
+// update adds more data to the Context.
 update :: proc(ctx: ^Context, data: []byte) {
 	_blake2.update(ctx, data)
 }
 
-final :: proc(ctx: ^Context, hash: []byte) {
-	_blake2.final(ctx, hash)
+// final finalizes the Context, writes the digest to hash, and calls
+// reset on the Context.
+//
+// Iff finalize_clone is set, final will work on a copy of the Context,
+// which is useful for for calculating rolling digests.
+final :: proc(ctx: ^Context, hash: []byte, finalize_clone: bool = false) {
+	_blake2.final(ctx, hash, finalize_clone)
+}
+
+// clone clones the Context other into ctx.
+clone :: proc(ctx, other: ^Context) {
+	_blake2.clone(ctx, other)
+}
+
+// reset sanitizes the Context.  The Context must be re-initialized to
+// be used again.
+reset :: proc(ctx: ^Context) {
+	_blake2.reset(ctx)
 }

core/crypto/blake2s/blake2s.odin

@@ -1,3 +1,10 @@
+/*
+package blake2s implements the BLAKE2s hash algorithm.
+
+See:
+- https://datatracker.ietf.org/doc/html/rfc7693
+- https://www.blake2.net/
+*/
 package blake2s
 
 /*
@@ -6,122 +13,47 @@ package blake2s
 
     List of contributors:
         zhibog, dotbmp:  Initial implementation.
-
-    Interface for the BLAKE2s hashing algorithm.
-    BLAKE2s and BLAKE2b share the implementation in the _blake2 package.
 */
 
-import "core:io"
-import "core:os"
-
 import "../_blake2"
 
-/*
-    High level API
-*/
-
+// DIGEST_SIZE is the BLAKE2s digest size in bytes.
 DIGEST_SIZE :: 32
 
-// hash_string will hash the given input and return the
-// computed hash
-hash_string :: proc(data: string) -> [DIGEST_SIZE]byte {
-	return hash_bytes(transmute([]byte)(data))
-}
-
-// hash_bytes will hash the given input and return the
-// computed hash
-hash_bytes :: proc(data: []byte) -> [DIGEST_SIZE]byte {
-	hash: [DIGEST_SIZE]byte
-	ctx: Context
-	cfg: _blake2.Blake2_Config
-	cfg.size = _blake2.BLAKE2S_SIZE
-	ctx.cfg = cfg
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-	return hash
-}
-
-// hash_string_to_buffer will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer :: proc(data: string, hash: []byte) {
-	hash_bytes_to_buffer(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer :: proc(data, hash: []byte) {
-	ctx: Context
-	cfg: _blake2.Blake2_Config
-	cfg.size = _blake2.BLAKE2S_SIZE
-	ctx.cfg = cfg
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash)
-}
-
-// hash_stream will read the stream in chunks and compute a
-// hash from its contents
-hash_stream :: proc(s: io.Stream) -> ([DIGEST_SIZE]byte, bool) {
-	hash: [DIGEST_SIZE]byte
-	ctx: Context
-	cfg: _blake2.Blake2_Config
-	cfg.size = _blake2.BLAKE2S_SIZE
-	ctx.cfg = cfg
-	init(&ctx)
-
-	buf := make([]byte, 512)
-	defer delete(buf)
-
-	read := 1
-	for read > 0 {
-		read, _ = io.read(s, buf)
-		if read > 0 {
-			update(&ctx, buf[:read])
-		}
-	}
-	final(&ctx, hash[:])
-	return hash, true
-}
-
-// hash_file will read the file provided by the given handle
-// and compute a hash
-hash_file :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE]byte, bool) {
-	if !load_at_once {
-		return hash_stream(os.stream_from_handle(hd))
-	} else {
-		if buf, ok := os.read_entire_file(hd); ok {
-			return hash_bytes(buf[:]), ok
-		}
-	}
-	return [DIGEST_SIZE]byte{}, false
-}
-
-hash :: proc {
-	hash_stream,
-	hash_file,
-	hash_bytes,
-	hash_string,
-	hash_bytes_to_buffer,
-	hash_string_to_buffer,
-}
-
-/*
-    Low level API
-*/
+// BLOCK_SIZE is the BLAKE2s block size in bytes.
+BLOCK_SIZE :: _blake2.BLAKE2S_BLOCK_SIZE
 
+// Context is a BLAKE2s instance.
 Context :: _blake2.Blake2s_Context
 
+// init initializes a Context with the default BLAKE2s config.
 init :: proc(ctx: ^Context) {
-	_blake2.init(ctx)
+	cfg: _blake2.Blake2_Config
+	cfg.size = _blake2.BLAKE2S_SIZE
+	_blake2.init(ctx, &cfg)
 }
 
+// update adds more data to the Context.
 update :: proc(ctx: ^Context, data: []byte) {
 	_blake2.update(ctx, data)
 }
 
-final :: proc(ctx: ^Context, hash: []byte) {
-	_blake2.final(ctx, hash)
+// final finalizes the Context, writes the digest to hash, and calls
+// reset on the Context.
+//
+// Iff finalize_clone is set, final will work on a copy of the Context,
+// which is useful for for calculating rolling digests.
+final :: proc(ctx: ^Context, hash: []byte, finalize_clone: bool = false) {
+	_blake2.final(ctx, hash, finalize_clone)
+}
+
+// clone clones the Context other into ctx.
+clone :: proc(ctx, other: ^Context) {
+	_blake2.clone(ctx, other)
+}
+
+// reset sanitizes the Context.  The Context must be re-initialized to
+// be used again.
+reset :: proc(ctx: ^Context) {
+	_blake2.reset(ctx)
 }

core/crypto/hash/doc.odin

@@ -0,0 +1,62 @@
+/*
+package hash provides a generic interface to the supported hash algorithms.
+
+A high-level convenience procedure group `hash` is provided to easily
+accomplish common tasks.
+- `hash_string` - Hash a given string and return the digest.
+- `hash_bytes` - Hash a given byte slice and return the digest.
+- `hash_string_to_buffer` - Hash a given string and put the digest in
+  the third parameter.  It requires that the destination buffer
+  is at least as big as the digest size.
+- `hash_bytes_to_buffer` - Hash a given string and put the computed
+  digest in the third parameter.  It requires that the destination
+  buffer is at least as big as the digest size.
+- `hash_stream` - Incrementally fully consume a `io.Stream`, and return
+  the computed digest.
+- `hash_file` - Takes a file handle and returns the computed digest.
+  A third optional boolean parameter controls if the file is streamed
+  (default), or or read at once.
+
+```odin
+package hash_example
+
+import "core:crypto/hash"
+
+main :: proc() {
+	input := "Feed the fire."
+
+	// Compute the digest, using the high level API.
+	returned_digest := hash.hash(hash.Algorithm.SHA512_256, input)
+	defer delete(returned_digest)
+
+	// Variant that takes a destination buffer, instead of returning
+	// the digest.
+	digest := make([]byte, hash.DIGEST_SIZES[hash.Algorithm.BLAKE2B]) // @note: Destination buffer has to be at least as big as the digest size of the hash.
+	defer delete(digest)
+	hash.hash(hash.Algorithm.BLAKE2B, input, digest)
+}
+```
+
+A generic low level API is provided supporting the init/update/final interface
+that is typical with cryptographic hash function implementations.
+
+```odin
+package hash_example
+
+import "core:crypto/hash"
+
+main :: proc() {
+    input := "Let the cinders burn."
+
+    // Compute the digest, using the low level API.
+    ctx: hash.Context
+    digest := make([]byte, hash.DIGEST_SIZES[hash.Algorithm.SHA3_512])
+    defer delete(digest)
+
+    hash.init(&ctx, hash.Algorithm.SHA3_512)
+    hash.update(&ctx, transmute([]byte)input)
+    hash.final(&ctx, digest)
+}
+```
+*/
+package crypto_hash

core/crypto/hash/hash.odin

@@ -0,0 +1,116 @@
+package crypto_hash
+
+/*
+    Copyright 2021 zhibog
+    Made available under the BSD-3 license.
+
+    List of contributors:
+        zhibog, dotbmp:  Initial implementation.
+*/
+
+import "core:io"
+import "core:mem"
+import "core:os"
+
+// hash_bytes will hash the given input and return the computed digest
+// in a newly allocated slice.
+hash_string :: proc(algorithm: Algorithm, data: string, allocator := context.allocator) -> []byte {
+	return hash_bytes(algorithm, transmute([]byte)(data), allocator)
+}
+
+// hash_bytes will hash the given input and return the computed digest
+// in a newly allocated slice.
+hash_bytes :: proc(algorithm: Algorithm, data: []byte, allocator := context.allocator) -> []byte {
+	dst := make([]byte, DIGEST_SIZES[algorithm], allocator)
+	hash_bytes_to_buffer(algorithm, data, dst)
+	return dst
+}
+
+// hash_string_to_buffer will hash the given input and assign the
+// computed digest to the third parameter.  It requires that the
+// destination buffer is at least as big as the digest size.
+hash_string_to_buffer :: proc(algorithm: Algorithm, data: string, hash: []byte) {
+	hash_bytes_to_buffer(algorithm, transmute([]byte)(data), hash)
+}
+
+// hash_bytes_to_buffer will hash the given input and write the
+// computed digest into the third parameter.  It requires that the
+// destination buffer is at least as big as the digest size.
+hash_bytes_to_buffer :: proc(algorithm: Algorithm, data, hash: []byte) {
+	ctx: Context
+
+	init(&ctx, algorithm)
+	update(&ctx, data)
+	final(&ctx, hash)
+}
+
+// hash_stream will incrementally fully consume a stream, and return the
+// computed digest in a newly allocated slice.
+hash_stream :: proc(
+	algorithm: Algorithm,
+	s: io.Stream,
+	allocator := context.allocator,
+) -> (
+	[]byte,
+	io.Error,
+) {
+	ctx: Context
+
+	buf: [MAX_BLOCK_SIZE * 4]byte
+	defer mem.zero_explicit(&buf, size_of(buf))
+
+	init(&ctx, algorithm)
+
+	loop: for {
+		n, err := io.read(s, buf[:])
+		if n > 0 {
+			// XXX/yawning: Can io.read return n > 0 and EOF?
+			update(&ctx, buf[:n])
+		}
+		#partial switch err {
+		case .None:
+		case .EOF:
+			break loop
+		case:
+			return nil, err
+		}
+	}
+
+	dst := make([]byte, DIGEST_SIZES[algorithm], allocator)
+	final(&ctx, dst)
+
+	return dst, io.Error.None
+}
+
+// hash_file will read the file provided by the given handle and return the
+// computed digest in a newly allocated slice.
+hash_file :: proc(
+	algorithm: Algorithm,
+	hd: os.Handle,
+	load_at_once := false,
+	allocator := context.allocator,
+) -> (
+	[]byte,
+	io.Error,
+) {
+	if !load_at_once {
+		return hash_stream(algorithm, os.stream_from_handle(hd), allocator)
+	}
+
+	buf, ok := os.read_entire_file(hd, allocator)
+	if !ok {
+		return nil, io.Error.Unknown
+	}
+	defer delete(buf, allocator)
+
+	return hash_bytes(algorithm, buf, allocator), io.Error.None
+}
+
+hash :: proc {
+	hash_stream,
+	hash_file,
+	hash_bytes,
+	hash_string,
+	hash_bytes_to_buffer,
+	hash_string_to_buffer,
+}

core/crypto/hash/low_level.odin

@@ -0,0 +1,353 @@
+package crypto_hash
+
+import "core:crypto/blake2b"
+import "core:crypto/blake2s"
+import "core:crypto/sha2"
+import "core:crypto/sha3"
+import "core:crypto/sm3"
+import "core:crypto/legacy/keccak"
+import "core:crypto/legacy/md5"
+import "core:crypto/legacy/sha1"
+
+import "core:reflect"
+
+// MAX_DIGEST_SIZE is the maximum size digest that can be returned by any
+// of the Algorithms supported via this package.
+MAX_DIGEST_SIZE :: 64
+// MAX_BLOCK_SIZE is the maximum block size used by any of Algorithms
+// supported by this package.
+MAX_BLOCK_SIZE :: sha3.BLOCK_SIZE_224
+
+// Algorithm is the algorithm identifier associated with a given Context.
+Algorithm :: enum {
+	Invalid,
+	BLAKE2B,
+	BLAKE2S,
+	SHA224,
+	SHA256,
+	SHA384,
+	SHA512,
+	SHA512_256,
+	SHA3_224,
+	SHA3_256,
+	SHA3_384,
+	SHA3_512,
+	SM3,
+	Legacy_KECCAK_224,
+	Legacy_KECCAK_256,
+	Legacy_KECCAK_384,
+	Legacy_KECCAK_512,
+	Insecure_MD5,
+	Insecure_SHA1,
+}
+
+// ALGORITHM_NAMES is the Algorithm to algorithm name string.
+ALGORITHM_NAMES := [Algorithm]string {
+	.Invalid           = "Invalid",
+	.BLAKE2B           = "BLAKE2b",
+	.BLAKE2S           = "BLAKE2s",
+	.SHA224            = "SHA-224",
+	.SHA256            = "SHA-256",
+	.SHA384            = "SHA-384",
+	.SHA512            = "SHA-512",
+	.SHA512_256        = "SHA-512/256",
+	.SHA3_224          = "SHA3-224",
+	.SHA3_256          = "SHA3-256",
+	.SHA3_384          = "SHA3-384",
+	.SHA3_512          = "SHA3-512",
+	.SM3               = "SM3",
+	.Legacy_KECCAK_224 = "Keccak-224",
+	.Legacy_KECCAK_256 = "Keccak-256",
+	.Legacy_KECCAK_384 = "Keccak-384",
+	.Legacy_KECCAK_512 = "Keccak-512",
+	.Insecure_MD5      = "MD5",
+	.Insecure_SHA1     = "SHA-1",
+}
+
+// DIGEST_SIZES is the Algorithm to digest size in bytes.
+DIGEST_SIZES := [Algorithm]int {
+	.Invalid           = 0,
+	.BLAKE2B           = blake2b.DIGEST_SIZE,
+	.BLAKE2S           = blake2s.DIGEST_SIZE,
+	.SHA224            = sha2.DIGEST_SIZE_224,
+	.SHA256            = sha2.DIGEST_SIZE_256,
+	.SHA384            = sha2.DIGEST_SIZE_384,
+	.SHA512            = sha2.DIGEST_SIZE_512,
+	.SHA512_256        = sha2.DIGEST_SIZE_512_256,
+	.SHA3_224          = sha3.DIGEST_SIZE_224,
+	.SHA3_256          = sha3.DIGEST_SIZE_256,
+	.SHA3_384          = sha3.DIGEST_SIZE_384,
+	.SHA3_512          = sha3.DIGEST_SIZE_512,
+	.SM3               = sm3.DIGEST_SIZE,
+	.Legacy_KECCAK_224 = keccak.DIGEST_SIZE_224,
+	.Legacy_KECCAK_256 = keccak.DIGEST_SIZE_256,
+	.Legacy_KECCAK_384 = keccak.DIGEST_SIZE_384,
+	.Legacy_KECCAK_512 = keccak.DIGEST_SIZE_512,
+	.Insecure_MD5      = md5.DIGEST_SIZE,
+	.Insecure_SHA1     = sha1.DIGEST_SIZE,
+}
+
+// BLOCK_SIZES is the Algoritm to block size in bytes.
+BLOCK_SIZES := [Algorithm]int {
+	.Invalid           = 0,
+	.BLAKE2B           = blake2b.BLOCK_SIZE,
+	.BLAKE2S           = blake2s.BLOCK_SIZE,
+	.SHA224            = sha2.BLOCK_SIZE_256,
+	.SHA256            = sha2.BLOCK_SIZE_256,
+	.SHA384            = sha2.BLOCK_SIZE_512,
+	.SHA512            = sha2.BLOCK_SIZE_512,
+	.SHA512_256        = sha2.BLOCK_SIZE_512,
+	.SHA3_224          = sha3.BLOCK_SIZE_224,
+	.SHA3_256          = sha3.BLOCK_SIZE_256,
+	.SHA3_384          = sha3.BLOCK_SIZE_384,
+	.SHA3_512          = sha3.BLOCK_SIZE_512,
+	.SM3               = sm3.BLOCK_SIZE,
+	.Legacy_KECCAK_224 = keccak.BLOCK_SIZE_224,
+	.Legacy_KECCAK_256 = keccak.BLOCK_SIZE_256,
+	.Legacy_KECCAK_384 = keccak.BLOCK_SIZE_384,
+	.Legacy_KECCAK_512 = keccak.BLOCK_SIZE_512,
+	.Insecure_MD5      = md5.BLOCK_SIZE,
+	.Insecure_SHA1     = sha1.BLOCK_SIZE,
+}
+
+// Context is a concrete instantiation of a specific hash algorithm.
+Context :: struct {
+	_algo: Algorithm,
+	_impl: union {
+		blake2b.Context,
+		blake2s.Context,
+		sha2.Context_256,
+		sha2.Context_512,
+		sha3.Context,
+		sm3.Context,
+		keccak.Context,
+		md5.Context,
+		sha1.Context,
+	},
+}
+
+@(private)
+_IMPL_IDS := [Algorithm]typeid {
+	.Invalid           = nil,
+	.BLAKE2B           = typeid_of(blake2b.Context),
+	.BLAKE2S           = typeid_of(blake2s.Context),
+	.SHA224            = typeid_of(sha2.Context_256),
+	.SHA256            = typeid_of(sha2.Context_256),
+	.SHA384            = typeid_of(sha2.Context_512),
+	.SHA512            = typeid_of(sha2.Context_512),
+	.SHA512_256        = typeid_of(sha2.Context_512),
+	.SHA3_224          = typeid_of(sha3.Context),
+	.SHA3_256          = typeid_of(sha3.Context),
+	.SHA3_384          = typeid_of(sha3.Context),
+	.SHA3_512          = typeid_of(sha3.Context),
+	.SM3               = typeid_of(sm3.Context),
+	.Legacy_KECCAK_224 = typeid_of(keccak.Context),
+	.Legacy_KECCAK_256 = typeid_of(keccak.Context),
+	.Legacy_KECCAK_384 = typeid_of(keccak.Context),
+	.Legacy_KECCAK_512 = typeid_of(keccak.Context),
+	.Insecure_MD5      = typeid_of(md5.Context),
+	.Insecure_SHA1     = typeid_of(sha1.Context),
+}
+
+// init initializes a Context with a specific hash Algorithm.
+init :: proc(ctx: ^Context, algorithm: Algorithm) {
+	if ctx._impl != nil {
+		reset(ctx)
+	}
+
+	// Directly specialize the union by setting the type ID (save a copy).
+	reflect.set_union_variant_typeid(
+		ctx._impl,
+		_IMPL_IDS[algorithm],
+	)
+	switch algorithm {
+	case .BLAKE2B:
+		blake2b.init(&ctx._impl.(blake2b.Context))
+	case .BLAKE2S:
+		blake2s.init(&ctx._impl.(blake2s.Context))
+	case .SHA224:
+		sha2.init_224(&ctx._impl.(sha2.Context_256))
+	case .SHA256:
+		sha2.init_256(&ctx._impl.(sha2.Context_256))
+	case .SHA384:
+		sha2.init_384(&ctx._impl.(sha2.Context_512))
+	case .SHA512:
+		sha2.init_512(&ctx._impl.(sha2.Context_512))
+	case .SHA512_256:
+		sha2.init_512_256(&ctx._impl.(sha2.Context_512))
+	case .SHA3_224:
+		sha3.init_224(&ctx._impl.(sha3.Context))
+	case .SHA3_256:
+		sha3.init_256(&ctx._impl.(sha3.Context))
+	case .SHA3_384:
+		sha3.init_384(&ctx._impl.(sha3.Context))
+	case .SHA3_512:
+		sha3.init_512(&ctx._impl.(sha3.Context))
+	case .SM3:
+		sm3.init(&ctx._impl.(sm3.Context))
+	case .Legacy_KECCAK_224:
+		keccak.init_224(&ctx._impl.(keccak.Context))
+	case .Legacy_KECCAK_256:
+		keccak.init_256(&ctx._impl.(keccak.Context))
+	case .Legacy_KECCAK_384:
+		keccak.init_384(&ctx._impl.(keccak.Context))
+	case .Legacy_KECCAK_512:
+		keccak.init_512(&ctx._impl.(keccak.Context))
+	case .Insecure_MD5:
+		md5.init(&ctx._impl.(md5.Context))
+	case .Insecure_SHA1:
+		sha1.init(&ctx._impl.(sha1.Context))
+	case .Invalid:
+		panic("crypto/hash: uninitialized algorithm")
+	case:
+		panic("crypto/hash: invalid algorithm")
+	}
+
+	ctx._algo = algorithm
+}
+
+// update adds more data to the Context.
+update :: proc(ctx: ^Context, data: []byte) {
+	switch &impl in ctx._impl {
+	case blake2b.Context:
+		blake2b.update(&impl, data)
+	case blake2s.Context:
+		blake2s.update(&impl, data)
+	case sha2.Context_256:
+		sha2.update(&impl, data)
+	case sha2.Context_512:
+		sha2.update(&impl, data)
+	case sha3.Context:
+		sha3.update(&impl, data)
+	case sm3.Context:
+		sm3.update(&impl, data)
+	case keccak.Context:
+		keccak.update(&impl, data)
+	case md5.Context:
+		md5.update(&impl, data)
+	case sha1.Context:
+		sha1.update(&impl, data)
+	case:
+		panic("crypto/hash: uninitialized algorithm")
+	}
+}
+
+// final finalizes the Context, writes the digest to hash, and calls
+// reset on the Context.
+//
+// Iff finalize_clone is set, final will work on a copy of the Context,
+// which is useful for for calculating rolling digests.
+final :: proc(ctx: ^Context, hash: []byte, finalize_clone: bool = false) {
+	switch &impl in ctx._impl {
+	case blake2b.Context:
+		blake2b.final(&impl, hash, finalize_clone)
+	case blake2s.Context:
+		blake2s.final(&impl, hash, finalize_clone)
+	case sha2.Context_256:
+		sha2.final(&impl, hash, finalize_clone)
+	case sha2.Context_512:
+		sha2.final(&impl, hash, finalize_clone)
+	case sha3.Context:
+		sha3.final(&impl, hash, finalize_clone)
+	case sm3.Context:
+		sm3.final(&impl, hash, finalize_clone)
+	case keccak.Context:
+		keccak.final(&impl, hash, finalize_clone)
+	case md5.Context:
+		md5.final(&impl, hash, finalize_clone)
+	case sha1.Context:
+		sha1.final(&impl, hash, finalize_clone)
+	case:
+		panic("crypto/hash: uninitialized algorithm")
+	}
+
+	if !finalize_clone {
+		reset(ctx)
+	}
+}
+
+// clone clones the Context other into ctx.
+clone :: proc(ctx, other: ^Context) {
+	// XXX/yawning: Maybe these cases should panic, because both cases,
+	// are probably bugs.
+	if ctx == other {
+		return
+	}
+	if ctx._impl != nil {
+		reset(ctx)
+	}
+
+	ctx._algo = other._algo
+
+	reflect.set_union_variant_typeid(
+		ctx._impl,
+		reflect.union_variant_typeid(other._impl),
+	)
+	switch &src_impl in other._impl {
+	case blake2b.Context:
+		blake2b.clone(&ctx._impl.(blake2b.Context), &src_impl)
+	case blake2s.Context:
+		blake2s.clone(&ctx._impl.(blake2s.Context), &src_impl)
+	case sha2.Context_256:
+		sha2.clone(&ctx._impl.(sha2.Context_256), &src_impl)
+	case sha2.Context_512:
+		sha2.clone(&ctx._impl.(sha2.Context_512), &src_impl)
+	case sha3.Context:
+		sha3.clone(&ctx._impl.(sha3.Context), &src_impl)
+	case sm3.Context:
+		sm3.clone(&ctx._impl.(sm3.Context), &src_impl)
+	case keccak.Context:
+		keccak.clone(&ctx._impl.(keccak.Context), &src_impl)
+	case md5.Context:
+		md5.clone(&ctx._impl.(md5.Context), &src_impl)
+	case sha1.Context:
+		sha1.clone(&ctx._impl.(sha1.Context), &src_impl)
+	case:
+		panic("crypto/hash: uninitialized algorithm")
+	}
+}
+
+// reset sanitizes the Context.  The Context must be re-initialized to
+// be used again.
+reset :: proc(ctx: ^Context) {
+	switch &impl in ctx._impl {
+	case blake2b.Context:
+		blake2b.reset(&impl)
+	case blake2s.Context:
+		blake2s.reset(&impl)
+	case sha2.Context_256:
+		sha2.reset(&impl)
+	case sha2.Context_512:
+		sha2.reset(&impl)
+	case sha3.Context:
+		sha3.reset(&impl)
+	case sm3.Context:
+		sm3.reset(&impl)
+	case keccak.Context:
+		keccak.reset(&impl)
+	case md5.Context:
+		md5.reset(&impl)
+	case sha1.Context:
+		sha1.reset(&impl)
+	case:
+	// Unlike clone, calling reset repeatedly is fine.
+	}
+
+	ctx._algo = .Invalid
+	ctx._impl = nil
+}
+
+// algorithm returns the Algorithm used by a Context instance.
+algorithm :: proc(ctx: ^Context) -> Algorithm {
+	return ctx._algo
+}
+
+// digest_size returns the digest size of a Context instance in bytes.
+digest_size :: proc(ctx: ^Context) -> int {
+	return DIGEST_SIZES[ctx._algo]
+}
+
+// block_size returns the block size of a Context instance in bytes.
+block_size :: proc(ctx: ^Context) -> int {
+	return BLOCK_SIZES[ctx._algo]
+}

core/crypto/hmac/hmac.odin

@@ -0,0 +1,162 @@
+/*
+package hmac implements the HMAC MAC algorithm.
+
+See:
+- https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.198-1.pdf
+*/
+package hmac
+
+import "core:crypto"
+import "core:crypto/hash"
+import "core:mem"
+
+// sum will compute the HMAC with the specified algorithm and key
+// over msg, and write the computed digest to dst.  It requires that
+// the dst buffer is the tag size.
+sum :: proc(algorithm: hash.Algorithm, dst, msg, key: []byte) {
+	ctx: Context
+
+	init(&ctx, algorithm, key)
+	update(&ctx, msg)
+	final(&ctx, dst)
+}
+
+// verify will verify the HMAC tag computed with the specified algorithm
+// and key over msg and return true iff the tag is valid.  It requires
+// that the tag is correctly sized.
+verify :: proc(algorithm: hash.Algorithm, tag, msg, key: []byte) -> bool {
+	tag_buf: [hash.MAX_DIGEST_SIZE]byte
+
+	derived_tag := tag_buf[:hash.DIGEST_SIZES[algorithm]]
+	sum(algorithm, derived_tag, msg, key)
+
+	return crypto.compare_constant_time(derived_tag, tag) == 1
+}
+
+// Context is a concrete instantiation of HMAC with a specific hash
+// algorithm.
+Context :: struct {
+	_o_hash:         hash.Context, // H(k ^ ipad) (not finalized)
+	_i_hash:         hash.Context, // H(k ^ opad) (not finalized)
+	_tag_sz:         int,
+	_is_initialized: bool,
+}
+
+// init initializes a Context with a specific hash Algorithm and key.
+init :: proc(ctx: ^Context, algorithm: hash.Algorithm, key: []byte) {
+	if ctx._is_initialized {
+		reset(ctx)
+	}
+
+	_init_hashes(ctx, algorithm, key)
+
+	ctx._tag_sz = hash.DIGEST_SIZES[algorithm]
+	ctx._is_initialized = true
+}
+
+// update adds more data to the Context.
+update :: proc(ctx: ^Context, data: []byte) {
+	assert(ctx._is_initialized)
+
+	hash.update(&ctx._i_hash, data)
+}
+
+// final finalizes the Context, writes the tag to dst, and calls
+// reset on the Context.
+final :: proc(ctx: ^Context, dst: []byte) {
+	assert(ctx._is_initialized)
+
+	defer (reset(ctx))
+
+	if len(dst) != ctx._tag_sz {
+		panic("crypto/hmac: invalid destination tag size")
+	}
+
+	hash.final(&ctx._i_hash, dst) // H((k ^ ipad) || text)
+
+	hash.update(&ctx._o_hash, dst) // H((k ^ opad) || H((k ^ ipad) || text))
+	hash.final(&ctx._o_hash, dst)
+}
+
+// reset sanitizes the Context.  The Context must be re-initialized to
+// be used again.
+reset :: proc(ctx: ^Context) {
+	if !ctx._is_initialized {
+		return
+	}
+
+	hash.reset(&ctx._o_hash)
+	hash.reset(&ctx._i_hash)
+	ctx._tag_sz = 0
+	ctx._is_initialized = false
+}
+
+// algorithm returns the Algorithm used by a Context instance.
+algorithm :: proc(ctx: ^Context) -> hash.Algorithm {
+	assert(ctx._is_initialized)
+
+	return hash.algorithm(&ctx._i_hash)
+}
+
+// tag_size returns the tag size of a Context instance in bytes.
+tag_size :: proc(ctx: ^Context) -> int {
+	assert(ctx._is_initialized)
+
+	return ctx._tag_sz
+}
+
+@(private)
+_I_PAD :: 0x36
+_O_PAD :: 0x5c
+
+@(private)
+_init_hashes :: proc(ctx: ^Context, algorithm: hash.Algorithm, key: []byte) {
+	K0_buf: [hash.MAX_BLOCK_SIZE]byte
+	kPad_buf: [hash.MAX_BLOCK_SIZE]byte
+
+	kLen := len(key)
+	B := hash.BLOCK_SIZES[algorithm]
+	K0 := K0_buf[:B]
+	defer mem.zero_explicit(raw_data(K0), B)
+
+	switch {
+	case kLen == B, kLen < B:
+		// If the length of K = B: set K0 = K.
+		//
+		// If the length of K < B: append zeros to the end of K to
+		// create a B-byte string K0 (e.g., if K is 20 bytes in
+		// length and B = 64, then K will be appended with 44 zero
+		// bytes x’00’).
+		//
+		// K0 is zero-initialized, so the copy handles both cases.
+		copy(K0, key)
+	case kLen > B:
+		// If the length of K > B: hash K to obtain an L byte string,
+		// then append (B-L) zeros to create a B-byte string K0
+		// (i.e., K0 = H(K) || 00...00).
+		tmpCtx := &ctx._o_hash // Saves allocating a hash.Context.
+		hash.init(tmpCtx, algorithm)
+		hash.update(tmpCtx, key)
+		hash.final(tmpCtx, K0)
+	}
+
+	// Initialize the hashes, and write the padded keys:
+	// - ctx._i_hash -> H(K0 ^ ipad)
+	// - ctx._o_hash -> H(K0 ^ opad)
+
+	hash.init(&ctx._o_hash, algorithm)
+	hash.init(&ctx._i_hash, algorithm)
+
+	kPad := kPad_buf[:B]
+	defer mem.zero_explicit(raw_data(kPad), B)
+
+	for v, i in K0 {
+		kPad[i] = v ~ _I_PAD
+	}
+	hash.update(&ctx._i_hash, kPad)
+
+	for v, i in K0 {
+		kPad[i] = v ~ _O_PAD
+	}
+	hash.update(&ctx._o_hash, kPad)
+}

core/crypto/legacy/keccak/keccak.odin

@@ -1,3 +1,11 @@
+/*
+package keccak implements the Keccak hash algorithm family.
+
+During the SHA-3 standardization process, the padding scheme was changed
+thus Keccac and SHA-3 produce different outputs.  Most users should use
+SHA-3 and/or SHAKE instead, however the legacy algorithm is provided for
+backward compatibility purposes.
+*/
 package keccak
 
 /*
@@ -6,372 +14,82 @@ package keccak
 
     List of contributors:
         zhibog, dotbmp:  Initial implementation.
-
-    Interface for the Keccak hashing algorithm.
-    This is done because the padding in the SHA3 standard was changed by the NIST, resulting in a different output.
 */
 
-import "core:io"
-import "core:os"
-
 import "../../_sha3"
 
-/*
-    High level API
-*/
-
+// DIGEST_SIZE_224 is the Keccak-224 digest size.
 DIGEST_SIZE_224 :: 28
+// DIGEST_SIZE_256 is the Keccak-256 digest size.
 DIGEST_SIZE_256 :: 32
+// DIGEST_SIZE_384 is the Keccak-384 digest size.
 DIGEST_SIZE_384 :: 48
+// DIGEST_SIZE_512 is the Keccak-512 digest size.
 DIGEST_SIZE_512 :: 64
 
-// hash_string_224 will hash the given input and return the
-// computed hash
-hash_string_224 :: proc(data: string) -> [DIGEST_SIZE_224]byte {
-	return hash_bytes_224(transmute([]byte)(data))
-}
+// BLOCK_SIZE_224 is the Keccak-224 block size in bytes.
+BLOCK_SIZE_224 :: _sha3.RATE_224
+// BLOCK_SIZE_256 is the Keccak-256 block size in bytes.
+BLOCK_SIZE_256 :: _sha3.RATE_256
+// BLOCK_SIZE_384 is the Keccak-384 block size in bytes.
+BLOCK_SIZE_384 :: _sha3.RATE_384
+// BLOCK_SIZE_512 is the Keccak-512 block size in bytes.
+BLOCK_SIZE_512 :: _sha3.RATE_512
 
-// hash_bytes_224 will hash the given input and return the
-// computed hash
-hash_bytes_224 :: proc(data: []byte) -> [DIGEST_SIZE_224]byte {
-	hash: [DIGEST_SIZE_224]byte
-	ctx: Context
+// Context is a Keccak instance.
+Context :: distinct _sha3.Context
+
+// init_224 initializes a Context for Keccak-224.
+init_224 :: proc(ctx: ^Context) {
 	ctx.mdlen = DIGEST_SIZE_224
-	ctx.is_keccak = true
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-	return hash
+	_init(ctx)
 }
 
-// hash_string_to_buffer_224 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_224 :: proc(data: string, hash: []byte) {
-	hash_bytes_to_buffer_224(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_224 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_224 :: proc(data, hash: []byte) {
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_224
-	ctx.is_keccak = true
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash)
-}
-
-// hash_stream_224 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_224 :: proc(s: io.Stream) -> ([DIGEST_SIZE_224]byte, bool) {
-	hash: [DIGEST_SIZE_224]byte
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_224
-	ctx.is_keccak = true
-	init(&ctx)
-
-	buf := make([]byte, 512)
-	defer delete(buf)
-
-	read := 1
-	for read > 0 {
-		read, _ = io.read(s, buf)
-		if read > 0 {
-			update(&ctx, buf[:read])
-		}
-	}
-	final(&ctx, hash[:])
-	return hash, true
-}
-
-// hash_file_224 will read the file provided by the given handle
-// and compute a hash
-hash_file_224 :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE_224]byte, bool) {
-	if !load_at_once {
-		return hash_stream_224(os.stream_from_handle(hd))
-	} else {
-		if buf, ok := os.read_entire_file(hd); ok {
-			return hash_bytes_224(buf[:]), ok
-		}
-	}
-	return [DIGEST_SIZE_224]byte{}, false
-}
-
-hash_224 :: proc {
-	hash_stream_224,
-	hash_file_224,
-	hash_bytes_224,
-	hash_string_224,
-	hash_bytes_to_buffer_224,
-	hash_string_to_buffer_224,
-}
-
-// hash_string_256 will hash the given input and return the
-// computed hash
-hash_string_256 :: proc(data: string) -> [DIGEST_SIZE_256]byte {
-	return hash_bytes_256(transmute([]byte)(data))
-}
-
-// hash_bytes_256 will hash the given input and return the
-// computed hash
-hash_bytes_256 :: proc(data: []byte) -> [DIGEST_SIZE_256]byte {
-	hash: [DIGEST_SIZE_256]byte
-	ctx: Context
+// init_256 initializes a Context for Keccak-256.
+init_256 :: proc(ctx: ^Context) {
 	ctx.mdlen = DIGEST_SIZE_256
-	ctx.is_keccak = true
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-	return hash
+	_init(ctx)
 }
 
-// hash_string_to_buffer_256 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_256 :: proc(data: string, hash: []byte) {
-	hash_bytes_to_buffer_256(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_256 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_256 :: proc(data, hash: []byte) {
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_256
-	ctx.is_keccak = true
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash)
-}
-
-// hash_stream_256 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_256 :: proc(s: io.Stream) -> ([DIGEST_SIZE_256]byte, bool) {
-	hash: [DIGEST_SIZE_256]byte
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_256
-	ctx.is_keccak = true
-	init(&ctx)
-
-	buf := make([]byte, 512)
-	defer delete(buf)
-
-	read := 1
-	for read > 0 {
-		read, _ = io.read(s, buf)
-		if read > 0 {
-			update(&ctx, buf[:read])
-		}
-	}
-	final(&ctx, hash[:])
-	return hash, true
-}
-
-// hash_file_256 will read the file provided by the given handle
-// and compute a hash
-hash_file_256 :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE_256]byte, bool) {
-	if !load_at_once {
-		return hash_stream_256(os.stream_from_handle(hd))
-	} else {
-		if buf, ok := os.read_entire_file(hd); ok {
-			return hash_bytes_256(buf[:]), ok
-		}
-	}
-	return [DIGEST_SIZE_256]byte{}, false
-}
-
-hash_256 :: proc {
-	hash_stream_256,
-	hash_file_256,
-	hash_bytes_256,
-	hash_string_256,
-	hash_bytes_to_buffer_256,
-	hash_string_to_buffer_256,
-}
-
-// hash_string_384 will hash the given input and return the
-// computed hash
-hash_string_384 :: proc(data: string) -> [DIGEST_SIZE_384]byte {
-	return hash_bytes_384(transmute([]byte)(data))
-}
-
-// hash_bytes_384 will hash the given input and return the
-// computed hash
-hash_bytes_384 :: proc(data: []byte) -> [DIGEST_SIZE_384]byte {
-	hash: [DIGEST_SIZE_384]byte
-	ctx: Context
+// init_384 initializes a Context for Keccak-384.
+init_384 :: proc(ctx: ^Context) {
 	ctx.mdlen = DIGEST_SIZE_384
-	ctx.is_keccak = true
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-	return hash
+	_init(ctx)
 }
 
-// hash_string_to_buffer_384 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_384 :: proc(data: string, hash: []byte) {
-	hash_bytes_to_buffer_384(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_384 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_384 :: proc(data, hash: []byte) {
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_384
-	ctx.is_keccak = true
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash)
-}
-
-// hash_stream_384 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_384 :: proc(s: io.Stream) -> ([DIGEST_SIZE_384]byte, bool) {
-	hash: [DIGEST_SIZE_384]byte
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_384
-	ctx.is_keccak = true
-	init(&ctx)
-
-	buf := make([]byte, 512)
-	defer delete(buf)
-
-	read := 1
-	for read > 0 {
-		read, _ = io.read(s, buf)
-		if read > 0 {
-			update(&ctx, buf[:read])
-		}
-	}
-	final(&ctx, hash[:])
-	return hash, true
-}
-
-// hash_file_384 will read the file provided by the given handle
-// and compute a hash
-hash_file_384 :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE_384]byte, bool) {
-	if !load_at_once {
-		return hash_stream_384(os.stream_from_handle(hd))
-	} else {
-		if buf, ok := os.read_entire_file(hd); ok {
-			return hash_bytes_384(buf[:]), ok
-		}
-	}
-	return [DIGEST_SIZE_384]byte{}, false
-}
-
-hash_384 :: proc {
-	hash_stream_384,
-	hash_file_384,
-	hash_bytes_384,
-	hash_string_384,
-	hash_bytes_to_buffer_384,
-	hash_string_to_buffer_384,
-}
-
-// hash_string_512 will hash the given input and return the
-// computed hash
-hash_string_512 :: proc(data: string) -> [DIGEST_SIZE_512]byte {
-	return hash_bytes_512(transmute([]byte)(data))
-}
-
-// hash_bytes_512 will hash the given input and return the
-// computed hash
-hash_bytes_512 :: proc(data: []byte) -> [DIGEST_SIZE_512]byte {
-	hash: [DIGEST_SIZE_512]byte
-	ctx: Context
+// init_512 initializes a Context for Keccak-512.
+init_512 :: proc(ctx: ^Context) {
 	ctx.mdlen = DIGEST_SIZE_512
+	_init(ctx)
+}
+
+@(private)
+_init :: proc(ctx: ^Context) {
 	ctx.is_keccak = true
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-	return hash
-}
-
-// hash_string_to_buffer_512 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_512 :: proc(data: string, hash: []byte) {
-	hash_bytes_to_buffer_512(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_512 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_512 :: proc(data, hash: []byte) {
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_512
-	ctx.is_keccak = true
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash)
-}
-
-// hash_stream_512 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_512 :: proc(s: io.Stream) -> ([DIGEST_SIZE_512]byte, bool) {
-	hash: [DIGEST_SIZE_512]byte
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_512
-	ctx.is_keccak = true
-	init(&ctx)
-
-	buf := make([]byte, 512)
-	defer delete(buf)
-
-	read := 1
-	for read > 0 {
-		read, _ = io.read(s, buf)
-		if read > 0 {
-			update(&ctx, buf[:read])
-		}
-	}
-	final(&ctx, hash[:])
-	return hash, true
-}
-
-// hash_file_512 will read the file provided by the given handle
-// and compute a hash
-hash_file_512 :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE_512]byte, bool) {
-	if !load_at_once {
-		return hash_stream_512(os.stream_from_handle(hd))
-	} else {
-		if buf, ok := os.read_entire_file(hd); ok {
-			return hash_bytes_512(buf[:]), ok
-		}
-	}
-	return [DIGEST_SIZE_512]byte{}, false
-}
-
-hash_512 :: proc {
-	hash_stream_512,
-	hash_file_512,
-	hash_bytes_512,
-	hash_string_512,
-	hash_bytes_to_buffer_512,
-	hash_string_to_buffer_512,
-}
-
-/*
-    Low level API
-*/
-
-Context :: _sha3.Sha3_Context
-
-init :: proc(ctx: ^Context) {
-	ctx.is_keccak = true
-	_sha3.init(ctx)
+	_sha3.init(transmute(^_sha3.Context)(ctx))
 }
 
+// update adds more data to the Context.
 update :: proc(ctx: ^Context, data: []byte) {
-	_sha3.update(ctx, data)
+	_sha3.update(transmute(^_sha3.Context)(ctx), data)
 }
 
-final :: proc(ctx: ^Context, hash: []byte) {
-	_sha3.final(ctx, hash)
+// final finalizes the Context, writes the digest to hash, and calls
+// reset on the Context.
+//
+// Iff finalize_clone is set, final will work on a copy of the Context,
+// which is useful for for calculating rolling digests.
+final :: proc(ctx: ^Context, hash: []byte, finalize_clone: bool = false) {
+	_sha3.final(transmute(^_sha3.Context)(ctx), hash, finalize_clone)
+}
+
+// clone clones the Context other into ctx.
+clone :: proc(ctx, other: ^Context) {
+	_sha3.clone(transmute(^_sha3.Context)(ctx), transmute(^_sha3.Context)(other))
+}
+
+// reset sanitizes the Context.  The Context must be re-initialized to
+// be used again.
+reset :: proc(ctx: ^Context) {
+	_sha3.reset(transmute(^_sha3.Context)(ctx))
 }

core/crypto/legacy/md5/md5.odin

@@ -1,3 +1,13 @@
+/*
+package md5 implements the MD5 hash algorithm.
+
+WARNING: The MD5 algorithm is known to be insecure and should only be
+used for interoperating with legacy applications.
+
+See:
+- https://eprint.iacr.org/2005/075
+- https://datatracker.ietf.org/doc/html/rfc1321
+*/
 package md5
 
 /*
@@ -6,103 +16,29 @@ package md5
 
     List of contributors:
         zhibog, dotbmp:  Initial implementation.
-
-    Implementation of the MD5 hashing algorithm, as defined in RFC 1321 <https://datatracker.ietf.org/doc/html/rfc1321>
 */
 
 import "core:encoding/endian"
-import "core:io"
 import "core:math/bits"
 import "core:mem"
-import "core:os"
-
-/*
-    High level API
-*/
 
+// DIGEST_SIZE is the MD5 digest size in bytes.
 DIGEST_SIZE :: 16
 
-// hash_string will hash the given input and return the
-// computed hash
-hash_string :: proc(data: string) -> [DIGEST_SIZE]byte {
-	return hash_bytes(transmute([]byte)(data))
+// BLOCK_SIZE is the MD5 block size in bytes.
+BLOCK_SIZE :: 64
+
+// Context is a MD5 instance.
+Context :: struct {
+	data:    [BLOCK_SIZE]byte,
+	state:   [4]u32,
+	bitlen:  u64,
+	datalen: u32,
+
+	is_initialized: bool,
 }
 
-// hash_bytes will hash the given input and return the
-// computed hash
-hash_bytes :: proc(data: []byte) -> [DIGEST_SIZE]byte {
-	hash: [DIGEST_SIZE]byte
-	ctx: Context
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-	return hash
-}
-
-// hash_string_to_buffer will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer :: proc(data: string, hash: []byte) {
-	hash_bytes_to_buffer(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer :: proc(data, hash: []byte) {
-	ctx: Context
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash)
-}
-
-// hash_stream will read the stream in chunks and compute a
-// hash from its contents
-hash_stream :: proc(s: io.Stream) -> ([DIGEST_SIZE]byte, bool) {
-	hash: [DIGEST_SIZE]byte
-	ctx: Context
-	init(&ctx)
-
-	buf := make([]byte, 512)
-	defer delete(buf)
-
-	read := 1
-	for read > 0 {
-		read, _ = io.read(s, buf)
-		if read > 0 {
-			update(&ctx, buf[:read])
-		}
-	}
-	final(&ctx, hash[:])
-	return hash, true
-}
-
-// hash_file will read the file provided by the given handle
-// and compute a hash
-hash_file :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE]byte, bool) {
-	if !load_at_once {
-		return hash_stream(os.stream_from_handle(hd))
-	} else {
-		if buf, ok := os.read_entire_file(hd); ok {
-			return hash_bytes(buf[:]), ok
-		}
-	}
-	return [DIGEST_SIZE]byte{}, false
-}
-
-hash :: proc {
-	hash_stream,
-	hash_file,
-	hash_bytes,
-	hash_string,
-	hash_bytes_to_buffer,
-	hash_string_to_buffer,
-}
-
-/*
-    Low level API
-*/
-
+// init initializes a Context.
 init :: proc(ctx: ^Context) {
 	ctx.state[0] = 0x67452301
 	ctx.state[1] = 0xefcdab89
@@ -115,6 +51,7 @@ init :: proc(ctx: ^Context) {
 	ctx.is_initialized = true
 }
 
+// update adds more data to the Context.
 update :: proc(ctx: ^Context, data: []byte) {
 	assert(ctx.is_initialized)
 
@@ -129,13 +66,26 @@ update :: proc(ctx: ^Context, data: []byte) {
 	}
 }
 
-final :: proc(ctx: ^Context, hash: []byte) {
+// final finalizes the Context, writes the digest to hash, and calls
+// reset on the Context.
+//
+// Iff finalize_clone is set, final will work on a copy of the Context,
+// which is useful for for calculating rolling digests.
+final :: proc(ctx: ^Context, hash: []byte, finalize_clone: bool = false) {
 	assert(ctx.is_initialized)
 
 	if len(hash) < DIGEST_SIZE {
 		panic("crypto/md5: invalid destination digest size")
 	}
 
+	ctx := ctx
+	if finalize_clone {
+		tmp_ctx: Context
+		clone(&tmp_ctx, ctx)
+		ctx = &tmp_ctx
+	}
+	defer(reset(ctx))
+
 	i := ctx.datalen
 
 	if ctx.datalen < 56 {
@@ -163,25 +113,27 @@ final :: proc(ctx: ^Context, hash: []byte) {
 	for i = 0; i < DIGEST_SIZE / 4; i += 1 {
 		endian.unchecked_put_u32le(hash[i * 4:], ctx.state[i])
 	}
+}
 
-	ctx.is_initialized = false
+// clone clones the Context other into ctx.
+clone :: proc(ctx, other: ^$T) {
+	ctx^ = other^
+}
+
+// reset sanitizes the Context.  The Context must be re-initialized to
+// be used again.
+reset :: proc(ctx: ^$T) {
+	if !ctx.is_initialized {
+		return
+	}
+
+	mem.zero_explicit(ctx, size_of(ctx^))
 }
 
 /*
     MD5 implementation
 */
 
-BLOCK_SIZE :: 64
-
-Context :: struct {
-	data:    [BLOCK_SIZE]byte,
-	state:   [4]u32,
-	bitlen:  u64,
-	datalen: u32,
-
-	is_initialized: bool,
-}
-
 /*
     @note(zh): F, G, H and I, as mentioned in the RFC, have been inlined into FF, GG, HH
     and II respectively, instead of declaring them separately.

core/crypto/legacy/sha1/sha1.odin

@@ -1,3 +1,14 @@
+/*
+package sha1 implements the SHA1 hash algorithm.
+
+WARNING: The SHA1 algorithm is known to be insecure and should only be
+used for interoperating with legacy applications.
+
+See:
+- https://eprint.iacr.org/2017/190
+- https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.180-4.pdf
+- https://datatracker.ietf.org/doc/html/rfc3174
+*/
 package sha1
 
 /*
@@ -6,103 +17,30 @@ package sha1
 
     List of contributors:
         zhibog, dotbmp:  Initial implementation.
-
-    Implementation of the SHA1 hashing algorithm, as defined in RFC 3174 <https://datatracker.ietf.org/doc/html/rfc3174>
 */
 
 import "core:encoding/endian"
-import "core:io"
 import "core:math/bits"
 import "core:mem"
-import "core:os"
-
-/*
-    High level API
-*/
 
+// DIGEST_SIZE is the SHA1 digest size in bytes.
 DIGEST_SIZE :: 20
 
-// hash_string will hash the given input and return the
-// computed hash
-hash_string :: proc(data: string) -> [DIGEST_SIZE]byte {
-	return hash_bytes(transmute([]byte)(data))
+// BLOCK_SIZE is the SHA1 block size in bytes.
+BLOCK_SIZE :: 64
+
+// Context is a SHA1 instance.
+Context :: struct {
+	data:    [BLOCK_SIZE]byte,
+	state:   [5]u32,
+	k:       [4]u32,
+	bitlen:  u64,
+	datalen: u32,
+
+	is_initialized: bool,
 }
 
-// hash_bytes will hash the given input and return the
-// computed hash
-hash_bytes :: proc(data: []byte) -> [DIGEST_SIZE]byte {
-	hash: [DIGEST_SIZE]byte
-	ctx: Context
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-	return hash
-}
-
-// hash_string_to_buffer will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer :: proc(data: string, hash: []byte) {
-	hash_bytes_to_buffer(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer :: proc(data, hash: []byte) {
-	ctx: Context
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash)
-}
-
-// hash_stream will read the stream in chunks and compute a
-// hash from its contents
-hash_stream :: proc(s: io.Stream) -> ([DIGEST_SIZE]byte, bool) {
-	hash: [DIGEST_SIZE]byte
-	ctx: Context
-	init(&ctx)
-
-	buf := make([]byte, 512)
-	defer delete(buf)
-
-	read := 1
-	for read > 0 {
-		read, _ = io.read(s, buf)
-		if read > 0 {
-			update(&ctx, buf[:read])
-		}
-	}
-	final(&ctx, hash[:])
-	return hash, true
-}
-
-// hash_file will read the file provided by the given handle
-// and compute a hash
-hash_file :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE]byte, bool) {
-	if !load_at_once {
-		return hash_stream(os.stream_from_handle(hd))
-	} else {
-		if buf, ok := os.read_entire_file(hd); ok {
-			return hash_bytes(buf[:]), ok
-		}
-	}
-	return [DIGEST_SIZE]byte{}, false
-}
-
-hash :: proc {
-	hash_stream,
-	hash_file,
-	hash_bytes,
-	hash_string,
-	hash_bytes_to_buffer,
-	hash_string_to_buffer,
-}
-
-/*
-    Low level API
-*/
-
+// init initializes a Context.
 init :: proc(ctx: ^Context) {
 	ctx.state[0] = 0x67452301
 	ctx.state[1] = 0xefcdab89
@@ -120,6 +58,7 @@ init :: proc(ctx: ^Context) {
 	ctx.is_initialized = true
 }
 
+// update adds more data to the Context.
 update :: proc(ctx: ^Context, data: []byte) {
 	assert(ctx.is_initialized)
 
@@ -134,13 +73,26 @@ update :: proc(ctx: ^Context, data: []byte) {
 	}
 }
 
-final :: proc(ctx: ^Context, hash: []byte) {
+// final finalizes the Context, writes the digest to hash, and calls
+// reset on the Context.
+//
+// Iff finalize_clone is set, final will work on a copy of the Context,
+// which is useful for for calculating rolling digests.
+final :: proc(ctx: ^Context, hash: []byte, finalize_clone: bool = false) {
 	assert(ctx.is_initialized)
 
 	if len(hash) < DIGEST_SIZE {
 		panic("crypto/sha1: invalid destination digest size")
 	}
 
+	ctx := ctx
+	if finalize_clone {
+		tmp_ctx: Context
+		clone(&tmp_ctx, ctx)
+		ctx = &tmp_ctx
+	}
+	defer(reset(ctx))
+
 	i := ctx.datalen
 
 	if ctx.datalen < 56 {
@@ -168,26 +120,27 @@ final :: proc(ctx: ^Context, hash: []byte) {
 	for i = 0; i < DIGEST_SIZE / 4; i += 1 {
 		endian.unchecked_put_u32be(hash[i * 4:], ctx.state[i])
 	}
+}
 
-	ctx.is_initialized = false
+// clone clones the Context other into ctx.
+clone :: proc(ctx, other: ^$T) {
+	ctx^ = other^
+}
+
+// reset sanitizes the Context.  The Context must be re-initialized to
+// be used again.
+reset :: proc(ctx: ^$T) {
+	if !ctx.is_initialized {
+		return
+	}
+
+	mem.zero_explicit(ctx, size_of(ctx^))
 }
 
 /*
     SHA1 implementation
 */
 
-BLOCK_SIZE :: 64
-
-Context :: struct {
-	data:    [BLOCK_SIZE]byte,
-	datalen: u32,
-	bitlen:  u64,
-	state:   [5]u32,
-	k:       [4]u32,
-
-	is_initialized: bool,
-}
-
 @(private)
 transform :: proc "contextless" (ctx: ^Context, data: []byte) {
 	a, b, c, d, e, i, t: u32

core/crypto/poly1305/poly1305.odin

@@ -23,10 +23,6 @@ verify :: proc (tag, msg, key: []byte) -> bool {
 	ctx: Context = ---
 	derived_tag: [16]byte = ---
 
-	if len(tag) != TAG_SIZE {
-		panic("crypto/poly1305: invalid tag size")
-	}
-
 	init(&ctx, key)
 	update(&ctx, msg)
 	final(&ctx, derived_tag[:])

core/crypto/sha2/sha2.odin

@@ -1,3 +1,10 @@
+/*
+package sha2 implements the SHA2 hash algorithm family.
+
+See:
+- https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.180-4.pdf
+- https://datatracker.ietf.org/doc/html/rfc3874
+*/
 package sha2
 
 /*
@@ -6,431 +13,83 @@ package sha2
 
     List of contributors:
         zhibog, dotbmp:  Initial implementation.
-
-    Implementation of the SHA2 hashing algorithm, as defined in <https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.180-4.pdf>
-    and in RFC 3874 <https://datatracker.ietf.org/doc/html/rfc3874>
 */
 
 import "core:encoding/endian"
-import "core:io"
 import "core:math/bits"
-import "core:os"
-
-/*
-    High level API
-*/
+import "core:mem"
 
+// DIGEST_SIZE_224 is the SHA-224 digest size in bytes.
 DIGEST_SIZE_224 :: 28
+// DIGEST_SIZE_256 is the SHA-256 digest size in bytes.
 DIGEST_SIZE_256 :: 32
+// DIGEST_SIZE_384 is the SHA-384 digest size in bytes.
 DIGEST_SIZE_384 :: 48
+// DIGEST_SIZE_512 is the SHA-512 digest size in bytes.
 DIGEST_SIZE_512 :: 64
+// DIGEST_SIZE_512_256 is the SHA-512/256 digest size in bytes.
 DIGEST_SIZE_512_256 :: 32
 
-// hash_string_224 will hash the given input and return the
-// computed hash
-hash_string_224 :: proc(data: string) -> [DIGEST_SIZE_224]byte {
-	return hash_bytes_224(transmute([]byte)(data))
+// BLOCK_SIZE_256 is the SHA-224 and SHA-256 block size in bytes.
+BLOCK_SIZE_256 :: 64
+// BLOCK_SIZE_512 is the SHA-384, SHA-512, and SHA-512/256 block size
+// in bytes.
+BLOCK_SIZE_512 :: 128
+
+// Context_256 is a SHA-224 or SHA-256 instance.
+Context_256 :: struct {
+	block:     [BLOCK_SIZE_256]byte,
+	h:         [8]u32,
+	bitlength: u64,
+	length:    u64,
+	md_bits:   int,
+
+	is_initialized: bool,
 }
 
-// hash_bytes_224 will hash the given input and return the
-// computed hash
-hash_bytes_224 :: proc(data: []byte) -> [DIGEST_SIZE_224]byte {
-	hash: [DIGEST_SIZE_224]byte
-	ctx: Context_256
+// Context_512 is a SHA-384, SHA-512 or SHA-512/256 instance.
+Context_512 :: struct {
+	block:     [BLOCK_SIZE_512]byte,
+	h:         [8]u64,
+	bitlength: u64,
+	length:    u64,
+	md_bits:   int,
+
+	is_initialized: bool,
+}
+
+// init_224 initializes a Context_256 for SHA-224.
+init_224 :: proc(ctx: ^Context_256) {
 	ctx.md_bits = 224
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-	return hash
+	_init(ctx)
 }
 
-// hash_string_to_buffer_224 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_224 :: proc(data: string, hash: []byte) {
-	hash_bytes_to_buffer_224(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_224 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_224 :: proc(data, hash: []byte) {
-	ctx: Context_256
-	ctx.md_bits = 224
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash)
-}
-
-// hash_stream_224 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_224 :: proc(s: io.Stream) -> ([DIGEST_SIZE_224]byte, bool) {
-	hash: [DIGEST_SIZE_224]byte
-	ctx: Context_256
-	ctx.md_bits = 224
-	init(&ctx)
-
-	buf := make([]byte, 512)
-	defer delete(buf)
-
-	read := 1
-	for read > 0 {
-		read, _ = io.read(s, buf)
-		if read > 0 {
-			update(&ctx, buf[:read])
-		}
-	}
-	final(&ctx, hash[:])
-	return hash, true
-}
-
-// hash_file_224 will read the file provided by the given handle
-// and compute a hash
-hash_file_224 :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE_224]byte, bool) {
-	if !load_at_once {
-		return hash_stream_224(os.stream_from_handle(hd))
-	} else {
-		if buf, ok := os.read_entire_file(hd); ok {
-			return hash_bytes_224(buf[:]), ok
-		}
-	}
-	return [DIGEST_SIZE_224]byte{}, false
-}
-
-hash_224 :: proc {
-	hash_stream_224,
-	hash_file_224,
-	hash_bytes_224,
-	hash_string_224,
-	hash_bytes_to_buffer_224,
-	hash_string_to_buffer_224,
-}
-
-// hash_string_256 will hash the given input and return the
-// computed hash
-hash_string_256 :: proc(data: string) -> [DIGEST_SIZE_256]byte {
-	return hash_bytes_256(transmute([]byte)(data))
-}
-
-// hash_bytes_256 will hash the given input and return the
-// computed hash
-hash_bytes_256 :: proc(data: []byte) -> [DIGEST_SIZE_256]byte {
-	hash: [DIGEST_SIZE_256]byte
-	ctx: Context_256
+// init_256 initializes a Context_256 for SHA-256.
+init_256 :: proc(ctx: ^Context_256) {
 	ctx.md_bits = 256
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-	return hash
+	_init(ctx)
 }
 
-// hash_string_to_buffer_256 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_256 :: proc(data: string, hash: []byte) {
-	hash_bytes_to_buffer_256(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_256 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_256 :: proc(data, hash: []byte) {
-	ctx: Context_256
-	ctx.md_bits = 256
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash)
-}
-
-// hash_stream_256 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_256 :: proc(s: io.Stream) -> ([DIGEST_SIZE_256]byte, bool) {
-	hash: [DIGEST_SIZE_256]byte
-	ctx: Context_256
-	ctx.md_bits = 256
-	init(&ctx)
-
-	buf := make([]byte, 512)
-	defer delete(buf)
-
-	read := 1
-	for read > 0 {
-		read, _ = io.read(s, buf)
-		if read > 0 {
-			update(&ctx, buf[:read])
-		}
-	}
-	final(&ctx, hash[:])
-	return hash, true
-}
-
-// hash_file_256 will read the file provided by the given handle
-// and compute a hash
-hash_file_256 :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE_256]byte, bool) {
-	if !load_at_once {
-		return hash_stream_256(os.stream_from_handle(hd))
-	} else {
-		if buf, ok := os.read_entire_file(hd); ok {
-			return hash_bytes_256(buf[:]), ok
-		}
-	}
-	return [DIGEST_SIZE_256]byte{}, false
-}
-
-hash_256 :: proc {
-	hash_stream_256,
-	hash_file_256,
-	hash_bytes_256,
-	hash_string_256,
-	hash_bytes_to_buffer_256,
-	hash_string_to_buffer_256,
-}
-
-// hash_string_384 will hash the given input and return the
-// computed hash
-hash_string_384 :: proc(data: string) -> [DIGEST_SIZE_384]byte {
-	return hash_bytes_384(transmute([]byte)(data))
-}
-
-// hash_bytes_384 will hash the given input and return the
-// computed hash
-hash_bytes_384 :: proc(data: []byte) -> [DIGEST_SIZE_384]byte {
-	hash: [DIGEST_SIZE_384]byte
-	ctx: Context_512
+// init_384 initializes a Context_512 for SHA-384.
+init_384 :: proc(ctx: ^Context_512) {
 	ctx.md_bits = 384
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-	return hash
+	_init(ctx)
 }
 
-// hash_string_to_buffer_384 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_384 :: proc(data: string, hash: []byte) {
-	hash_bytes_to_buffer_384(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_384 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_384 :: proc(data, hash: []byte) {
-	ctx: Context_512
-	ctx.md_bits = 384
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash)
-}
-
-// hash_stream_384 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_384 :: proc(s: io.Stream) -> ([DIGEST_SIZE_384]byte, bool) {
-	hash: [DIGEST_SIZE_384]byte
-	ctx: Context_512
-	ctx.md_bits = 384
-	init(&ctx)
-
-	buf := make([]byte, 512)
-	defer delete(buf)
-
-	read := 1
-	for read > 0 {
-		read, _ = io.read(s, buf)
-		if read > 0 {
-			update(&ctx, buf[:read])
-		}
-	}
-	final(&ctx, hash[:])
-	return hash, true
-}
-
-// hash_file_384 will read the file provided by the given handle
-// and compute a hash
-hash_file_384 :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE_384]byte, bool) {
-	if !load_at_once {
-		return hash_stream_384(os.stream_from_handle(hd))
-	} else {
-		if buf, ok := os.read_entire_file(hd); ok {
-			return hash_bytes_384(buf[:]), ok
-		}
-	}
-	return [DIGEST_SIZE_384]byte{}, false
-}
-
-hash_384 :: proc {
-	hash_stream_384,
-	hash_file_384,
-	hash_bytes_384,
-	hash_string_384,
-	hash_bytes_to_buffer_384,
-	hash_string_to_buffer_384,
-}
-
-// hash_string_512 will hash the given input and return the
-// computed hash
-hash_string_512 :: proc(data: string) -> [DIGEST_SIZE_512]byte {
-	return hash_bytes_512(transmute([]byte)(data))
-}
-
-// hash_bytes_512 will hash the given input and return the
-// computed hash
-hash_bytes_512 :: proc(data: []byte) -> [DIGEST_SIZE_512]byte {
-	hash: [DIGEST_SIZE_512]byte
-	ctx: Context_512
+// init_512 initializes a Context_512 for SHA-512.
+init_512 :: proc(ctx: ^Context_512) {
 	ctx.md_bits = 512
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-	return hash
+	_init(ctx)
 }
 
-// hash_string_to_buffer_512 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_512 :: proc(data: string, hash: []byte) {
-	hash_bytes_to_buffer_512(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_512 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_512 :: proc(data, hash: []byte) {
-	ctx: Context_512
-	ctx.md_bits = 512
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash)
-}
-
-// hash_stream_512 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_512 :: proc(s: io.Stream) -> ([DIGEST_SIZE_512]byte, bool) {
-	hash: [DIGEST_SIZE_512]byte
-	ctx: Context_512
-	ctx.md_bits = 512
-	init(&ctx)
-
-	buf := make([]byte, 512)
-	defer delete(buf)
-
-	read := 1
-	for read > 0 {
-		read, _ = io.read(s, buf)
-		if read > 0 {
-			update(&ctx, buf[:read])
-		}
-	}
-	final(&ctx, hash[:])
-	return hash, true
-}
-
-// hash_file_512 will read the file provided by the given handle
-// and compute a hash
-hash_file_512 :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE_512]byte, bool) {
-	if !load_at_once {
-		return hash_stream_512(os.stream_from_handle(hd))
-	} else {
-		if buf, ok := os.read_entire_file(hd); ok {
-			return hash_bytes_512(buf[:]), ok
-		}
-	}
-	return [DIGEST_SIZE_512]byte{}, false
-}
-
-hash_512 :: proc {
-	hash_stream_512,
-	hash_file_512,
-	hash_bytes_512,
-	hash_string_512,
-	hash_bytes_to_buffer_512,
-	hash_string_to_buffer_512,
-}
-
-// hash_string_512_256 will hash the given input and return the
-// computed hash
-hash_string_512_256 :: proc(data: string) -> [DIGEST_SIZE_512_256]byte {
-	return hash_bytes_512_256(transmute([]byte)(data))
-}
-
-// hash_bytes_512_256 will hash the given input and return the
-// computed hash
-hash_bytes_512_256 :: proc(data: []byte) -> [DIGEST_SIZE_512_256]byte {
-	hash: [DIGEST_SIZE_512_256]byte
-	ctx: Context_512
+// init_512_256 initializes a Context_512 for SHA-512/256.
+init_512_256 :: proc(ctx: ^Context_512) {
 	ctx.md_bits = 256
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-	return hash
+	_init(ctx)
 }
 
-// hash_string_to_buffer_512_256 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_512_256 :: proc(data: string, hash: []byte) {
-	hash_bytes_to_buffer_512_256(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_512_256 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_512_256 :: proc(data, hash: []byte) {
-	ctx: Context_512
-	ctx.md_bits = 256
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash)
-}
-
-// hash_stream_512_256 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_512_256 :: proc(s: io.Stream) -> ([DIGEST_SIZE_512_256]byte, bool) {
-	hash: [DIGEST_SIZE_512_256]byte
-	ctx: Context_512
-	ctx.md_bits = 256
-	init(&ctx)
-
-	buf := make([]byte, 512)
-	defer delete(buf)
-
-	read := 1
-	for read > 0 {
-		read, _ = io.read(s, buf)
-		if read > 0 {
-			update(&ctx, buf[:read])
-		}
-	}
-	final(&ctx, hash[:])
-	return hash, true
-}
-
-// hash_file_512_256 will read the file provided by the given handle
-// and compute a hash
-hash_file_512_256 :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE_512_256]byte, bool) {
-	if !load_at_once {
-		return hash_stream_512_256(os.stream_from_handle(hd))
-	} else {
-		if buf, ok := os.read_entire_file(hd); ok {
-			return hash_bytes_512_256(buf[:]), ok
-		}
-	}
-	return [DIGEST_SIZE_512_256]byte{}, false
-}
-
-hash_512_256 :: proc {
-	hash_stream_512_256,
-	hash_file_512_256,
-	hash_bytes_512_256,
-	hash_string_512_256,
-	hash_bytes_to_buffer_512_256,
-	hash_string_to_buffer_512_256,
-}
-
-/*
-    Low level API
-*/
-
-init :: proc(ctx: ^$T) {
+@(private)
+_init :: proc(ctx: ^$T) {
 	when T == Context_256 {
 		switch ctx.md_bits {
 		case 224:
@@ -497,13 +156,14 @@ init :: proc(ctx: ^$T) {
 	ctx.is_initialized = true
 }
 
+// update adds more data to the Context.
 update :: proc(ctx: ^$T, data: []byte) {
 	assert(ctx.is_initialized)
 
 	when T == Context_256 {
-		CURR_BLOCK_SIZE :: SHA256_BLOCK_SIZE
+		CURR_BLOCK_SIZE :: BLOCK_SIZE_256
 	} else when T == Context_512 {
-		CURR_BLOCK_SIZE :: SHA512_BLOCK_SIZE
+		CURR_BLOCK_SIZE :: BLOCK_SIZE_512
 	}
 
 	data := data
@@ -528,21 +188,34 @@ update :: proc(ctx: ^$T, data: []byte) {
 	}
 }
 
-final :: proc(ctx: ^$T, hash: []byte) {
+// final finalizes the Context, writes the digest to hash, and calls
+// reset on the Context.
+//
+// Iff finalize_clone is set, final will work on a copy of the Context,
+// which is useful for for calculating rolling digests.
+final :: proc(ctx: ^$T, hash: []byte, finalize_clone: bool = false) {
 	assert(ctx.is_initialized)
 
 	if len(hash) * 8 < ctx.md_bits {
 		panic("crypto/sha2: invalid destination digest size")
 	}
 
+	ctx := ctx
+	if finalize_clone {
+		tmp_ctx: T
+		clone(&tmp_ctx, ctx)
+		ctx = &tmp_ctx
+	}
+	defer(reset(ctx))
+
 	length := ctx.length
 
-	raw_pad: [SHA512_BLOCK_SIZE]byte
+	raw_pad: [BLOCK_SIZE_512]byte
 	when T == Context_256 {
-		CURR_BLOCK_SIZE :: SHA256_BLOCK_SIZE
+		CURR_BLOCK_SIZE :: BLOCK_SIZE_256
 		pm_len := 8 // 64-bits for length
 	} else when T == Context_512 {
-		CURR_BLOCK_SIZE :: SHA512_BLOCK_SIZE
+		CURR_BLOCK_SIZE :: BLOCK_SIZE_512
 		pm_len := 16 // 128-bits for length
 	}
 	pad := raw_pad[:CURR_BLOCK_SIZE]
@@ -576,37 +249,27 @@ final :: proc(ctx: ^$T, hash: []byte) {
 			endian.unchecked_put_u64be(hash[i * 8:], ctx.h[i])
 		}
 	}
+}
 
-	ctx.is_initialized = false
+// clone clones the Context other into ctx.
+clone :: proc(ctx, other: ^$T) {
+	ctx^ = other^
+}
+
+// reset sanitizes the Context.  The Context must be re-initialized to
+// be used again.
+reset :: proc(ctx: ^$T) {
+	if !ctx.is_initialized {
+		return
+	}
+
+	mem.zero_explicit(ctx, size_of(ctx^))
 }
 
 /*
     SHA2 implementation
 */
 
-SHA256_BLOCK_SIZE :: 64
-SHA512_BLOCK_SIZE :: 128
-
-Context_256 :: struct {
-	block:     [SHA256_BLOCK_SIZE]byte,
-	h:         [8]u32,
-	bitlength: u64,
-	length:    u64,
-	md_bits:   int,
-
-	is_initialized: bool,
-}
-
-Context_512 :: struct {
-	block:     [SHA512_BLOCK_SIZE]byte,
-	h:         [8]u64,
-	bitlength: u64,
-	length:    u64,
-	md_bits:   int,
-
-	is_initialized: bool,
-}
-
 @(private)
 sha256_k := [64]u32 {
 	0x428a2f98, 0x71374491, 0xb5c0fbcf, 0xe9b5dba5,
@@ -737,12 +400,12 @@ sha2_transf :: proc "contextless" (ctx: ^$T, data: []byte) {
 		w: [64]u32
 		wv: [8]u32
 		t1, t2: u32
-		CURR_BLOCK_SIZE :: SHA256_BLOCK_SIZE
+		CURR_BLOCK_SIZE :: BLOCK_SIZE_256
 	} else when T == Context_512 {
 		w: [80]u64
 		wv: [8]u64
 		t1, t2: u64
-		CURR_BLOCK_SIZE :: SHA512_BLOCK_SIZE
+		CURR_BLOCK_SIZE :: BLOCK_SIZE_512
 	}
 
 	data := data

core/crypto/sha3/sha3.odin

@@ -1,3 +1,13 @@
+/*
+package sha3 implements the SHA3 hash algorithm family.
+
+The SHAKE XOF can be found in crypto/shake.  While discouraged if the
+pre-standardization Keccak algorithm is required, it can be found in
+crypto/legacy/keccak.
+
+See:
+- https://nvlpubs.nist.gov/nistpubs/fips/nist.fips.202.pdf
+*/
 package sha3
 
 /*
@@ -6,359 +16,81 @@ package sha3
 
     List of contributors:
         zhibog, dotbmp:  Initial implementation.
-
-    Interface for the SHA3 hashing algorithm. The SHAKE functionality can be found in package shake.
-    If you wish to compute a Keccak hash, you can use the keccak package, it will use the original padding.
 */
 
-import "core:io"
-import "core:os"
-
 import "../_sha3"
 
-/*
-    High level API
-*/
-
+// DIGEST_SIZE_224 is the SHA3-224 digest size.
 DIGEST_SIZE_224 :: 28
+// DIGEST_SIZE_256 is the SHA3-256 digest size.
 DIGEST_SIZE_256 :: 32
+// DIGEST_SIZE_384 is the SHA3-384 digest size.
 DIGEST_SIZE_384 :: 48
+// DIGEST_SIZE_512 is the SHA3-512 digest size.
 DIGEST_SIZE_512 :: 64
 
-// hash_string_224 will hash the given input and return the
-// computed hash
-hash_string_224 :: proc(data: string) -> [DIGEST_SIZE_224]byte {
-	return hash_bytes_224(transmute([]byte)(data))
-}
+// BLOCK_SIZE_224 is the SHA3-224 block size in bytes.
+BLOCK_SIZE_224 :: _sha3.RATE_224
+// BLOCK_SIZE_256 is the SHA3-256 block size in bytes.
+BLOCK_SIZE_256 :: _sha3.RATE_256
+// BLOCK_SIZE_384 is the SHA3-384 block size in bytes.
+BLOCK_SIZE_384 :: _sha3.RATE_384
+// BLOCK_SIZE_512 is the SHA3-512 block size in bytes.
+BLOCK_SIZE_512 :: _sha3.RATE_512
 
-// hash_bytes_224 will hash the given input and return the
-// computed hash
-hash_bytes_224 :: proc(data: []byte) -> [DIGEST_SIZE_224]byte {
-	hash: [DIGEST_SIZE_224]byte
-	ctx: Context
+// Context is a SHA3 instance.
+Context :: distinct _sha3.Context
+
+// init_224 initializes a Context for SHA3-224.
+init_224 :: proc(ctx: ^Context) {
 	ctx.mdlen = DIGEST_SIZE_224
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-	return hash
+	_init(ctx)
 }
 
-// hash_string_to_buffer_224 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_224 :: proc(data: string, hash: []byte) {
-	hash_bytes_to_buffer_224(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_224 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_224 :: proc(data, hash: []byte) {
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_224
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash)
-}
-
-// hash_stream_224 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_224 :: proc(s: io.Stream) -> ([DIGEST_SIZE_224]byte, bool) {
-	hash: [DIGEST_SIZE_224]byte
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_224
-	init(&ctx)
-
-	buf := make([]byte, 512)
-	defer delete(buf)
-
-	read := 1
-	for read > 0 {
-		read, _ = io.read(s, buf)
-		if read > 0 {
-			update(&ctx, buf[:read])
-		}
-	}
-	final(&ctx, hash[:])
-	return hash, true
-}
-
-// hash_file_224 will read the file provided by the given handle
-// and compute a hash
-hash_file_224 :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE_224]byte, bool) {
-	if !load_at_once {
-		return hash_stream_224(os.stream_from_handle(hd))
-	} else {
-		if buf, ok := os.read_entire_file(hd); ok {
-			return hash_bytes_224(buf[:]), ok
-		}
-	}
-	return [DIGEST_SIZE_224]byte{}, false
-}
-
-hash_224 :: proc {
-	hash_stream_224,
-	hash_file_224,
-	hash_bytes_224,
-	hash_string_224,
-	hash_bytes_to_buffer_224,
-	hash_string_to_buffer_224,
-}
-
-// hash_string_256 will hash the given input and return the
-// computed hash
-hash_string_256 :: proc(data: string) -> [DIGEST_SIZE_256]byte {
-	return hash_bytes_256(transmute([]byte)(data))
-}
-
-// hash_bytes_256 will hash the given input and return the
-// computed hash
-hash_bytes_256 :: proc(data: []byte) -> [DIGEST_SIZE_256]byte {
-	hash: [DIGEST_SIZE_256]byte
-	ctx: Context
+// init_256 initializes a Context for SHA3-256.
+init_256 :: proc(ctx: ^Context) {
 	ctx.mdlen = DIGEST_SIZE_256
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-	return hash
+	_init(ctx)
 }
 
-// hash_string_to_buffer_256 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_256 :: proc(data: string, hash: []byte) {
-	hash_bytes_to_buffer_256(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_256 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_256 :: proc(data, hash: []byte) {
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_256
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash)
-}
-
-// hash_stream_256 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_256 :: proc(s: io.Stream) -> ([DIGEST_SIZE_256]byte, bool) {
-	hash: [DIGEST_SIZE_256]byte
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_256
-	init(&ctx)
-
-	buf := make([]byte, 512)
-	defer delete(buf)
-
-	read := 1
-	for read > 0 {
-		read, _ = io.read(s, buf)
-		if read > 0 {
-			update(&ctx, buf[:read])
-		}
-	}
-	final(&ctx, hash[:])
-	return hash, true
-}
-
-// hash_file_256 will read the file provided by the given handle
-// and compute a hash
-hash_file_256 :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE_256]byte, bool) {
-	if !load_at_once {
-		return hash_stream_256(os.stream_from_handle(hd))
-	} else {
-		if buf, ok := os.read_entire_file(hd); ok {
-			return hash_bytes_256(buf[:]), ok
-		}
-	}
-	return [DIGEST_SIZE_256]byte{}, false
-}
-
-hash_256 :: proc {
-	hash_stream_256,
-	hash_file_256,
-	hash_bytes_256,
-	hash_string_256,
-	hash_bytes_to_buffer_256,
-	hash_string_to_buffer_256,
-}
-
-// hash_string_384 will hash the given input and return the
-// computed hash
-hash_string_384 :: proc(data: string) -> [DIGEST_SIZE_384]byte {
-	return hash_bytes_384(transmute([]byte)(data))
-}
-
-// hash_bytes_384 will hash the given input and return the
-// computed hash
-hash_bytes_384 :: proc(data: []byte) -> [DIGEST_SIZE_384]byte {
-	hash: [DIGEST_SIZE_384]byte
-	ctx: Context
+// init_384 initializes a Context for SHA3-384.
+init_384 :: proc(ctx: ^Context) {
 	ctx.mdlen = DIGEST_SIZE_384
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-	return hash
+	_init(ctx)
 }
 
-// hash_string_to_buffer_384 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_384 :: proc(data: string, hash: []byte) {
-	hash_bytes_to_buffer_384(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_384 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_384 :: proc(data, hash: []byte) {
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_384
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash)
-}
-
-// hash_stream_384 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_384 :: proc(s: io.Stream) -> ([DIGEST_SIZE_384]byte, bool) {
-	hash: [DIGEST_SIZE_384]byte
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_384
-	init(&ctx)
-
-	buf := make([]byte, 512)
-	defer delete(buf)
-
-	read := 1
-	for read > 0 {
-		read, _ = io.read(s, buf)
-		if read > 0 {
-			update(&ctx, buf[:read])
-		}
-	}
-	final(&ctx, hash[:])
-	return hash, true
-}
-
-// hash_file_384 will read the file provided by the given handle
-// and compute a hash
-hash_file_384 :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE_384]byte, bool) {
-	if !load_at_once {
-		return hash_stream_384(os.stream_from_handle(hd))
-	} else {
-		if buf, ok := os.read_entire_file(hd); ok {
-			return hash_bytes_384(buf[:]), ok
-		}
-	}
-	return [DIGEST_SIZE_384]byte{}, false
-}
-
-hash_384 :: proc {
-	hash_stream_384,
-	hash_file_384,
-	hash_bytes_384,
-	hash_string_384,
-	hash_bytes_to_buffer_384,
-	hash_string_to_buffer_384,
-}
-
-// hash_string_512 will hash the given input and return the
-// computed hash
-hash_string_512 :: proc(data: string) -> [DIGEST_SIZE_512]byte {
-	return hash_bytes_512(transmute([]byte)(data))
-}
-
-// hash_bytes_512 will hash the given input and return the
-// computed hash
-hash_bytes_512 :: proc(data: []byte) -> [DIGEST_SIZE_512]byte {
-	hash: [DIGEST_SIZE_512]byte
-	ctx: Context
+// init_512 initializes a Context for SHA3-512.
+init_512 :: proc(ctx: ^Context) {
 	ctx.mdlen = DIGEST_SIZE_512
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-	return hash
+	_init(ctx)
 }
 
-// hash_string_to_buffer_512 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_512 :: proc(data: string, hash: []byte) {
-	hash_bytes_to_buffer_512(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_512 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_512 :: proc(data, hash: []byte) {
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_512
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash)
-}
-
-// hash_stream_512 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_512 :: proc(s: io.Stream) -> ([DIGEST_SIZE_512]byte, bool) {
-	hash: [DIGEST_SIZE_512]byte
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_512
-	init(&ctx)
-
-	buf := make([]byte, 512)
-	defer delete(buf)
-
-	read := 1
-	for read > 0 {
-		read, _ = io.read(s, buf)
-		if read > 0 {
-			update(&ctx, buf[:read])
-		}
-	}
-	final(&ctx, hash[:])
-	return hash, true
-}
-
-// hash_file_512 will read the file provided by the given handle
-// and compute a hash
-hash_file_512 :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE_512]byte, bool) {
-	if !load_at_once {
-		return hash_stream_512(os.stream_from_handle(hd))
-	} else {
-		if buf, ok := os.read_entire_file(hd); ok {
-			return hash_bytes_512(buf[:]), ok
-		}
-	}
-	return [DIGEST_SIZE_512]byte{}, false
-}
-
-hash_512 :: proc {
-	hash_stream_512,
-	hash_file_512,
-	hash_bytes_512,
-	hash_string_512,
-	hash_bytes_to_buffer_512,
-	hash_string_to_buffer_512,
-}
-
-/*
-    Low level API
-*/
-
-Context :: _sha3.Sha3_Context
-
-init :: proc(ctx: ^Context) {
-	_sha3.init(ctx)
+@(private)
+_init :: proc(ctx: ^Context) {
+	_sha3.init(transmute(^_sha3.Context)(ctx))
 }
 
+// update adds more data to the Context.
 update :: proc(ctx: ^Context, data: []byte) {
-	_sha3.update(ctx, data)
+	_sha3.update(transmute(^_sha3.Context)(ctx), data)
 }
 
-final :: proc(ctx: ^Context, hash: []byte) {
-	_sha3.final(ctx, hash)
+// final finalizes the Context, writes the digest to hash, and calls
+// reset on the Context.
+//
+// Iff finalize_clone is set, final will work on a copy of the Context,
+// which is useful for for calculating rolling digests.
+final :: proc(ctx: ^Context, hash: []byte, finalize_clone: bool = false) {
+	_sha3.final(transmute(^_sha3.Context)(ctx), hash, finalize_clone)
+}
+
+// clone clones the Context other into ctx.
+clone :: proc(ctx, other: ^Context) {
+	_sha3.clone(transmute(^_sha3.Context)(ctx), transmute(^_sha3.Context)(other))
+}
+
+// reset sanitizes the Context.  The Context must be re-initialized to
+// be used again.
+reset :: proc(ctx: ^Context) {
+	_sha3.reset(transmute(^_sha3.Context)(ctx))
 }

core/crypto/shake/shake.odin

@@ -1,3 +1,11 @@
+/*
+package shake implements the SHAKE XOF algorithm family.
+
+The SHA3 hash algorithm can be found in the crypto/sha3.
+
+See:
+- https://nvlpubs.nist.gov/nistpubs/fips/nist.fips.202.pdf
+*/
 package shake
 
 /*
@@ -6,201 +14,55 @@ package shake
 
     List of contributors:
         zhibog, dotbmp:  Initial implementation.
-
-    Interface for the SHAKE hashing algorithm.
-    The SHA3 functionality can be found in package sha3.
-
-    TODO: This should provide an incremental squeeze interface, in addition
-    to the one-shot final call.
 */
 
-import "core:io"
-import "core:os"
-
 import "../_sha3"
 
-/*
-    High level API
-*/
+// Context is a SHAKE128 or SHAKE256 instance.
+Context :: distinct _sha3.Context
 
-DIGEST_SIZE_128 :: 16
-DIGEST_SIZE_256 :: 32
-
-// hash_string_128 will hash the given input and return the
-// computed hash
-hash_string_128 :: proc(data: string) -> [DIGEST_SIZE_128]byte {
-	return hash_bytes_128(transmute([]byte)(data))
+// init_128 initializes a Context for SHAKE128.
+init_128 :: proc(ctx: ^Context) {
+	ctx.mdlen = 128 / 8
+	_init(ctx)
 }
 
-// hash_bytes_128 will hash the given input and return the
-// computed hash
-hash_bytes_128 :: proc(data: []byte) -> [DIGEST_SIZE_128]byte {
-	hash: [DIGEST_SIZE_128]byte
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_128
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-	return hash
+// init_256 initializes a Context for SHAKE256.
+init_256 :: proc(ctx: ^Context) {
+	ctx.mdlen = 256 / 8
+	_init(ctx)
 }
 
-// hash_string_to_buffer_128 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_128 :: proc(data: string, hash: []byte) {
-	hash_bytes_to_buffer_128(transmute([]byte)(data), hash)
+@(private)
+_init :: proc(ctx: ^Context) {
+	_sha3.init(transmute(^_sha3.Context)(ctx))
 }
 
-// hash_bytes_to_buffer_128 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_128 :: proc(data, hash: []byte) {
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_128
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash)
+// write writes more data into the SHAKE instance.  This MUST not be called
+// after any reads have been done, and attempts to do so will panic.
+write :: proc(ctx: ^Context, data: []byte) {
+	_sha3.update(transmute(^_sha3.Context)(ctx), data)
 }
 
-// hash_stream_128 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_128 :: proc(s: io.Stream) -> ([DIGEST_SIZE_128]byte, bool) {
-	hash: [DIGEST_SIZE_128]byte
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_128
-	init(&ctx)
-
-	buf := make([]byte, 512)
-	defer delete(buf)
-
-	read := 1
-	for read > 0 {
-		read, _ = io.read(s, buf)
-		if read > 0 {
-			update(&ctx, buf[:read])
-		}
+// read reads output from the SHAKE instance.  There is no practical upper
+// limit to the amount of data that can be read from SHAKE.  After read has
+// been called one or more times, further calls to write will panic.
+read :: proc(ctx: ^Context, dst: []byte) {
+	ctx_ := transmute(^_sha3.Context)(ctx)
+	if !ctx.is_finalized {
+		_sha3.shake_xof(ctx_)
 	}
-	final(&ctx, hash[:])
-	return hash, true
+
+	_sha3.shake_out(ctx_, dst)
 }
 
-// hash_file_128 will read the file provided by the given handle
-// and compute a hash
-hash_file_128 :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE_128]byte, bool) {
-	if !load_at_once {
-		return hash_stream_128(os.stream_from_handle(hd))
-	} else {
-		if buf, ok := os.read_entire_file(hd); ok {
-			return hash_bytes_128(buf[:]), ok
-		}
-	}
-	return [DIGEST_SIZE_128]byte{}, false
+// clone clones the Context other into ctx.
+clone :: proc(ctx, other: ^Context) {
+	_sha3.clone(transmute(^_sha3.Context)(ctx), transmute(^_sha3.Context)(other))
 }
 
-hash_128 :: proc {
-	hash_stream_128,
-	hash_file_128,
-	hash_bytes_128,
-	hash_string_128,
-	hash_bytes_to_buffer_128,
-	hash_string_to_buffer_128,
-}
-
-// hash_string_256 will hash the given input and return the
-// computed hash
-hash_string_256 :: proc(data: string) -> [DIGEST_SIZE_256]byte {
-	return hash_bytes_256(transmute([]byte)(data))
-}
-
-// hash_bytes_256 will hash the given input and return the
-// computed hash
-hash_bytes_256 :: proc(data: []byte) -> [DIGEST_SIZE_256]byte {
-	hash: [DIGEST_SIZE_256]byte
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_256
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-	return hash
-}
-
-// hash_string_to_buffer_256 will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer_256 :: proc(data: string, hash: []byte) {
-	hash_bytes_to_buffer_256(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer_256 will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer_256 :: proc(data, hash: []byte) {
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_256
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-}
-
-// hash_stream_256 will read the stream in chunks and compute a
-// hash from its contents
-hash_stream_256 :: proc(s: io.Stream) -> ([DIGEST_SIZE_256]byte, bool) {
-	hash: [DIGEST_SIZE_256]byte
-	ctx: Context
-	ctx.mdlen = DIGEST_SIZE_256
-	init(&ctx)
-
-	buf := make([]byte, 512)
-	defer delete(buf)
-
-	read := 1
-	for read > 0 {
-		read, _ = io.read(s, buf)
-		if read > 0 {
-			update(&ctx, buf[:read])
-		}
-	}
-	final(&ctx, hash[:])
-	return hash, true
-}
-
-// hash_file_256 will read the file provided by the given handle
-// and compute a hash
-hash_file_256 :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE_256]byte, bool) {
-	if !load_at_once {
-		return hash_stream_256(os.stream_from_handle(hd))
-	} else {
-		if buf, ok := os.read_entire_file(hd); ok {
-			return hash_bytes_256(buf[:]), ok
-		}
-	}
-	return [DIGEST_SIZE_256]byte{}, false
-}
-
-hash_256 :: proc {
-	hash_stream_256,
-	hash_file_256,
-	hash_bytes_256,
-	hash_string_256,
-	hash_bytes_to_buffer_256,
-	hash_string_to_buffer_256,
-}
-
-/*
-    Low level API
-*/
-
-Context :: _sha3.Sha3_Context
-
-init :: proc(ctx: ^Context) {
-	_sha3.init(ctx)
-}
-
-update :: proc(ctx: ^Context, data: []byte) {
-	_sha3.update(ctx, data)
-}
-
-final :: proc(ctx: ^Context, hash: []byte) {
-	_sha3.shake_xof(ctx)
-	_sha3.shake_out(ctx, hash[:])
+// reset sanitizes the Context.  The Context must be re-initialized to
+// be used again.
+reset :: proc(ctx: ^Context) {
+	_sha3.reset(transmute(^_sha3.Context)(ctx))
 }

core/crypto/sm3/sm3.odin

@@ -1,3 +1,9 @@
+/*
+package sm3 implements the SM3 hash algorithm.
+
+See:
+- https://datatracker.ietf.org/doc/html/draft-sca-cfrg-sm3-02
+*/
 package sm3
 
 /*
@@ -6,102 +12,29 @@ package sm3
 
     List of contributors:
         zhibog, dotbmp:  Initial implementation.
-
-    Implementation of the SM3 hashing algorithm, as defined in <https://datatracker.ietf.org/doc/html/draft-sca-cfrg-sm3-02>
 */
 
 import "core:encoding/endian"
-import "core:io"
 import "core:math/bits"
-import "core:os"
-
-/*
-    High level API
-*/
+import "core:mem"
 
+// DIGEST_SIZE is the SM3 digest size in bytes.
 DIGEST_SIZE :: 32
 
-// hash_string will hash the given input and return the
-// computed hash
-hash_string :: proc(data: string) -> [DIGEST_SIZE]byte {
-	return hash_bytes(transmute([]byte)(data))
+// BLOCK_SIZE is the SM3 block size in bytes.
+BLOCK_SIZE :: 64
+
+// Context is a SM3 instance.
+Context :: struct {
+	state:     [8]u32,
+	x:         [BLOCK_SIZE]byte,
+	bitlength: u64,
+	length:    u64,
+
+	is_initialized: bool,
 }
 
-// hash_bytes will hash the given input and return the
-// computed hash
-hash_bytes :: proc(data: []byte) -> [DIGEST_SIZE]byte {
-	hash: [DIGEST_SIZE]byte
-	ctx: Context
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash[:])
-	return hash
-}
-
-// hash_string_to_buffer will hash the given input and assign the
-// computed hash to the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_string_to_buffer :: proc(data: string, hash: []byte) {
-	hash_bytes_to_buffer(transmute([]byte)(data), hash)
-}
-
-// hash_bytes_to_buffer will hash the given input and write the
-// computed hash into the second parameter.
-// It requires that the destination buffer is at least as big as the digest size
-hash_bytes_to_buffer :: proc(data, hash: []byte) {
-	ctx: Context
-	init(&ctx)
-	update(&ctx, data)
-	final(&ctx, hash)
-}
-
-// hash_stream will read the stream in chunks and compute a
-// hash from its contents
-hash_stream :: proc(s: io.Stream) -> ([DIGEST_SIZE]byte, bool) {
-	hash: [DIGEST_SIZE]byte
-	ctx: Context
-	init(&ctx)
-
-	buf := make([]byte, 512)
-	defer delete(buf)
-
-	read := 1
-	for read > 0 {
-		read, _ = io.read(s, buf)
-		if read > 0 {
-			update(&ctx, buf[:read])
-		}
-	}
-	final(&ctx, hash[:])
-	return hash, true
-}
-
-// hash_file will read the file provided by the given handle
-// and compute a hash
-hash_file :: proc(hd: os.Handle, load_at_once := false) -> ([DIGEST_SIZE]byte, bool) {
-	if !load_at_once {
-		return hash_stream(os.stream_from_handle(hd))
-	} else {
-		if buf, ok := os.read_entire_file(hd); ok {
-			return hash_bytes(buf[:]), ok
-		}
-	}
-	return [DIGEST_SIZE]byte{}, false
-}
-
-hash :: proc {
-	hash_stream,
-	hash_file,
-	hash_bytes,
-	hash_string,
-	hash_bytes_to_buffer,
-	hash_string_to_buffer,
-}
-
-/*
-    Low level API
-*/
-
+// init initializes a Context.
 init :: proc(ctx: ^Context) {
 	ctx.state[0] = IV[0]
 	ctx.state[1] = IV[1]
@@ -118,6 +51,7 @@ init :: proc(ctx: ^Context) {
 	ctx.is_initialized = true
 }
 
+// update adds more data to the Context.
 update :: proc(ctx: ^Context, data: []byte) {
 	assert(ctx.is_initialized)
 
@@ -143,13 +77,26 @@ update :: proc(ctx: ^Context, data: []byte) {
 	}
 }
 
-final :: proc(ctx: ^Context, hash: []byte) {
+// final finalizes the Context, writes the digest to hash, and calls
+// reset on the Context.
+//
+// Iff finalize_clone is set, final will work on a copy of the Context,
+// which is useful for for calculating rolling digests.
+final :: proc(ctx: ^Context, hash: []byte, finalize_clone: bool = false) {
 	assert(ctx.is_initialized)
 
 	if len(hash) < DIGEST_SIZE {
 		panic("crypto/sm3: invalid destination digest size")
 	}
 
+	ctx := ctx
+	if finalize_clone {
+		tmp_ctx: Context
+		clone(&tmp_ctx, ctx)
+		ctx = &tmp_ctx
+	}
+	defer(reset(ctx))
+
 	length := ctx.length
 
 	pad: [BLOCK_SIZE]byte
@@ -168,25 +115,27 @@ final :: proc(ctx: ^Context, hash: []byte) {
 	for i := 0; i < DIGEST_SIZE / 4; i += 1 {
 		endian.unchecked_put_u32be(hash[i * 4:], ctx.state[i])
 	}
+}
 
-	ctx.is_initialized = false
+// clone clones the Context other into ctx.
+clone :: proc(ctx, other: ^Context) {
+	ctx^ = other^
+}
+
+// reset sanitizes the Context.  The Context must be re-initialized to
+// be used again.
+reset :: proc(ctx: ^Context) {
+	if !ctx.is_initialized {
+		return
+	}
+
+	mem.zero_explicit(ctx, size_of(ctx^))
 }
 
 /*
     SM3 implementation
 */
 
-BLOCK_SIZE :: 64
-
-Context :: struct {
-	state:     [8]u32,
-	x:         [BLOCK_SIZE]byte,
-	bitlength: u64,
-	length:    u64,
-
-	is_initialized: bool,
-}
-
 @(private)
 IV := [8]u32 {
 	0x7380166f, 0x4914b2b9, 0x172442d7, 0xda8a0600,

core/dynlib/doc.odin

@@ -1,7 +1,11 @@
+//+build ignore
 /*
 Package core:dynlib implements loading of shared libraries/DLLs and their symbols.
 
 The behaviour of dynamically loaded libraries is specific to the target platform of the program.
 For in depth detail on the underlying behaviour please refer to your target platform's documentation.
+
+See `example` directory for an example library exporting 3 symbols and a host program loading them automatically
+by defining a symbol table struct.
 */
-package dynlib
+package dynlib

core/dynlib/example/example.odin

@@ -0,0 +1,45 @@
+package example
+
+import "core:dynlib"
+import "core:fmt"
+
+Symbols :: struct {
+	// `foo_` is prefixed, so we look for the symbol `foo_add`.
+	add: proc "c" (int, int) -> int,
+	// We use the tag here to override the symbol to look for, namely `bar_sub`.
+	sub: proc "c" (int, int) -> int `dynlib:"bar_sub"`,
+
+	// Exported global (if exporting an i32, the type must be ^i32 because the symbol is a pointer to the export.)
+	// If it's not a pointer or procedure type, we'll skip the struct field.
+	hellope: ^i32,
+
+	// Handle to free library.
+	// We can have more than one of these so we can match symbols for more than one DLL with one struct.
+	_my_lib_handle: dynlib.Library,
+}
+
+main :: proc() {
+	sym: Symbols
+
+	// Load symbols from `lib.dll` into Symbols struct.
+	// Each struct field is prefixed with `foo_` before lookup in the DLL's symbol table.
+	// The library's Handle (to unload) will be stored in `sym._my_lib_handle`. This way you can load multiple DLLs in one struct.
+	count, ok := dynlib.initialize_symbols(&sym, "lib.dll", "foo_", "_my_lib_handle")
+	defer dynlib.unload_library(sym._my_lib_handle)
+	fmt.printf("(Initial DLL Load) ok: %v. %v symbols loaded from lib.dll (%p).\n", ok, count, sym._my_lib_handle)
+
+	if count > 0 {
+		fmt.println("42 + 42 =", sym.add(42, 42))
+		fmt.println("84 - 13 =", sym.sub(84, 13))
+		fmt.println("hellope =", sym.hellope^)
+	}
+
+	count, ok = dynlib.initialize_symbols(&sym, "lib.dll", "foo_", "_my_lib_handle")
+	fmt.printf("(DLL Reload) ok: %v. %v symbols loaded from lib.dll (%p).\n", ok, count, sym._my_lib_handle)
+
+	if count > 0 {
+		fmt.println("42 + 42 =", sym.add(42, 42))
+		fmt.println("84 - 13 =", sym.sub(84, 13))
+		fmt.println("hellope =", sym.hellope^)
+	}
+}

core/dynlib/example/lib.odin

@@ -0,0 +1,14 @@
+package library
+
+@(export)
+foo_add :: proc "c" (a, b: int) -> (res: int) {
+	return a + b
+}
+
+@(export)
+bar_sub :: proc "c" (a, b: int) -> (res: int) {
+	return a - b
+}
+
+@(export)
+foo_hellope: i32 = 42

core/dynlib/lib.odin

@@ -1,5 +1,12 @@
 package dynlib
 
+import "base:intrinsics"
+import "core:reflect"
+import "base:runtime"
+_ :: intrinsics
+_ :: reflect
+_ :: runtime
+
 /*
 A handle to a dynamically loaded library.
 */
@@ -12,11 +19,11 @@ library available to resolve references in subsequently loaded libraries.
 The paramater `global_symbols` is only used for the platforms `linux`, `darwin`, `freebsd` and `openbsd`.
 On `windows` this paramater is ignored.
 
-The underlying behaviour is platform specific.  
-On `linux`, `darwin`, `freebsd` and `openbsd` refer to `dlopen`.  
+The underlying behaviour is platform specific.
+On `linux`, `darwin`, `freebsd` and `openbsd` refer to `dlopen`.
 On `windows` refer to `LoadLibraryW`.
 
-**Implicit Allocators**  
+**Implicit Allocators**
 `context.temp_allocator`
 
 Example:
@@ -27,6 +34,7 @@ Example:
 		LIBRARY_PATH :: "my_library.dll"
 		library, ok := dynlib.load_library(LIBRARY_PATH)
 		if ! ok {
+			fmt.eprintln(dynlib.last_error())
 			return
 		}
 		fmt.println("The library %q was successfully loaded", LIBRARY_PATH)
@@ -39,8 +47,8 @@ load_library :: proc(path: string, global_symbols := false) -> (library: Library
 /*
 Unloads a dynamic library.
 
-The underlying behaviour is platform specific.  
-On `linux`, `darwin`, `freebsd` and `openbsd` refer to `dlclose`.  
+The underlying behaviour is platform specific.
+On `linux`, `darwin`, `freebsd` and `openbsd` refer to `dlclose`.
 On `windows` refer to `FreeLibrary`.
 
 Example:
@@ -51,10 +59,12 @@ Example:
 		LIBRARY_PATH :: "my_library.dll"
 		library, ok := dynlib.load_library(LIBRARY_PATH)
 		if ! ok {
+			fmt.eprintln(dynlib.last_error())
 			return
 		}
 		did_unload := dynlib.unload_library(library)
 		if ! did_unload {
+			fmt.eprintln(dynlib.last_error())
 			return
 		}
 		fmt.println("The library %q was successfully unloaded", LIBRARY_PATH)
@@ -67,11 +77,11 @@ unload_library :: proc(library: Library) -> (did_unload: bool) {
 /*
 Loads the address of a procedure/variable from a dynamic library.
 
-The underlying behaviour is platform specific.  
-On `linux`, `darwin`, `freebsd` and `openbsd` refer to `dlsym`.  
+The underlying behaviour is platform specific.
+On `linux`, `darwin`, `freebsd` and `openbsd` refer to `dlsym`.
 On `windows` refer to `GetProcAddress`.
 
-**Implicit Allocators**  
+**Implicit Allocators**
 `context.temp_allocator`
 
 Example:
@@ -82,13 +92,94 @@ Example:
 		LIBRARY_PATH :: "my_library.dll"
 		library, ok := dynlib.load_library(LIBRARY_PATH)
 		if ! ok {
+			fmt.eprintln(dynlib.last_error())
 			return
 		}
 
 		a, found_a := dynlib.symbol_address(library, "a")
-		if found_a do fmt.printf("The symbol %q was found at the address %v", "a", a)
+		if found_a {
+			fmt.printf("The symbol %q was found at the address %v", "a", a)
+		} else {
+			fmt.eprintln(dynlib.last_error())
+		}
 	}
 */
 symbol_address :: proc(library: Library, symbol: string) -> (ptr: rawptr, found: bool) #optional_ok {
 	return _symbol_address(library, symbol)
 }
+
+/*
+Scans a dynamic library for symbols matching a struct's members, assigning found procedure pointers to the corresponding entry.
+Optionally takes a symbol prefix added to the struct's member name to construct the symbol looked up in the library.
+Optionally also takes the struct member to assign the library handle to, `__handle` by default.
+
+This allows using one struct to hold library handles and symbol pointers for more than 1 dynamic library.
+
+Loading the same library twice unloads the previous incarnation, allowing for straightforward hot reload support.
+
+Returns:
+* `-1, false` if the library could not be loaded.
+* The number of symbols assigned on success. `ok` = true if `count` > 0
+
+See doc.odin for an example.
+*/
+initialize_symbols :: proc(
+	symbol_table: ^$T, library_path: string,
+	symbol_prefix := "", handle_field_name := "__handle",
+) -> (count: int = -1, ok: bool = false) where intrinsics.type_is_struct(T) {
+	assert(symbol_table != nil)
+
+	handle := load_library(library_path) or_return
+
+	// Buffer to concatenate the prefix + symbol name.
+	prefixed_symbol_buf: [2048]u8 = ---
+
+	count = 0
+	for field, i in reflect.struct_fields_zipped(T) {
+		// Calculate address of struct member
+		field_ptr := rawptr(uintptr(symbol_table) + field.offset)
+
+		// If we've come across the struct member for the handle, store it and continue scanning for other symbols.
+		if field.name == handle_field_name {
+			// We appear to be hot reloading. Unload previous incarnation of the library.
+			if old_handle := (^Library)(field_ptr)^; old_handle != nil {
+				unload_library(old_handle) or_return
+			}
+			(^Library)(field_ptr)^ = handle
+			continue
+		}
+
+		// We're not the library handle, so the field needs to be a pointer type, be it a procedure pointer or an exported global.
+		if !(reflect.is_procedure(field.type) || reflect.is_pointer(field.type)) {
+			continue
+		}
+
+		// Let's look up or construct the symbol name to find in the library
+		prefixed_name: string
+
+		// Do we have a symbol override tag?
+		if override, tag_ok := reflect.struct_tag_lookup(field.tag, "dynlib"); tag_ok {
+			prefixed_name = override
+		}
+
+		// No valid symbol override tag found, fall back to `<symbol_prefix>name`.
+		if len(prefixed_name) == 0 {
+			offset := copy(prefixed_symbol_buf[:], symbol_prefix)
+			copy(prefixed_symbol_buf[offset:], field.name)
+			prefixed_name = string(prefixed_symbol_buf[:len(symbol_prefix) + len(field.name)])
+		}
+
+		// Assign procedure (or global) pointer if found.
+		sym_ptr := symbol_address(handle, prefixed_name) or_continue
+		(^rawptr)(field_ptr)^ = sym_ptr
+		count += 1
+	}
+	return count, count > 0
+}
+
+/*
+Returns an error message for the last failed procedure call.
+*/
+last_error :: proc() -> string {
+	return _last_error()
+}

core/dynlib/lib_js.odin

@@ -13,3 +13,7 @@ _unload_library :: proc(library: Library) -> bool {
 _symbol_address :: proc(library: Library, symbol: string) -> (ptr: rawptr, found: bool) {
 	return nil, false
 }
+
+_last_error :: proc() -> string {
+	return ""
+}

core/dynlib/lib_unix.odin

@@ -22,3 +22,8 @@ _symbol_address :: proc(library: Library, symbol: string) -> (ptr: rawptr, found
 	found = ptr != nil
 	return
 }
+
+_last_error :: proc() -> string {
+	err := os.dlerror()
+	return "unknown" if err == "" else err
+}

core/dynlib/lib_windows.odin

@@ -4,7 +4,8 @@ package dynlib
 
 import win32 "core:sys/windows"
 import "core:strings"
-import "core:runtime"
+import "base:runtime"
+import "core:reflect"
 
 _load_library :: proc(path: string, global_symbols := false) -> (Library, bool) {
 	// NOTE(bill): 'global_symbols' is here only for consistency with POSIX which has RTLD_GLOBAL
@@ -27,3 +28,9 @@ _symbol_address :: proc(library: Library, symbol: string) -> (ptr: rawptr, found
 	found = ptr != nil
 	return
 }
+
+_last_error :: proc() -> string {
+	err := win32.System_Error(win32.GetLastError())
+	err_msg := reflect.enum_string(err)
+	return "unknown" if err_msg == "" else err_msg
+}

core/encoding/endian/endian.odin

@@ -1,6 +1,6 @@
 package encoding_endian
 
-import "core:intrinsics"
+import "base:intrinsics"
 import "core:math/bits"
 
 Byte_Order :: enum u8 {

core/encoding/json/marshal.odin

@@ -2,11 +2,12 @@ package json
 
 import "core:mem"
 import "core:math/bits"
-import "core:runtime"
+import "base:runtime"
 import "core:strconv"
 import "core:strings"
 import "core:reflect"
 import "core:io"
+import "core:slice"
 
 Marshal_Data_Error :: enum {
 	None,
@@ -18,29 +19,40 @@ Marshal_Error :: union #shared_nil {
 	io.Error,
 }
 
-// careful with MJSON maps & non quotes usage as keys without whitespace will lead to bad results
+// careful with MJSON maps & non quotes usage as keys with whitespace will lead to bad results
 Marshal_Options :: struct {
 	// output based on spec
 	spec: Specification,
 
-	// use line breaks & tab|spaces
+	// Use line breaks & tabs/spaces
 	pretty: bool, 
 
-	// spacing
+	// Use spaces for indentation instead of tabs
 	use_spaces: bool,
+
+	// Given use_spaces true, use this many spaces per indent level. 0 means 4 spaces.
 	spaces: int,
 
-	// state
-	indentation: int,
-
-	// option to output uint in JSON5 & MJSON
+	// Output uint as hex in JSON5 & MJSON
 	write_uint_as_hex: bool, 
 
-	// mjson output options
+	// If spec is MJSON and this is true, then keys will be quoted.
+	//
+	// WARNING: If your keys contain whitespace and this is false, then the
+	// output will be bad.
 	mjson_keys_use_quotes: bool,
+
+	// If spec is MJSON and this is true, then use '=' as delimiter between
+	// keys and values, otherwise ':' is used.
 	mjson_keys_use_equal_sign: bool,
 
-	// mjson state
+	// When outputting a map, sort the output by key.
+	//
+	// NOTE: This will temp allocate and sort a list for each map.
+	sort_maps_by_key: bool,
+
+	// Internal state
+	indentation: int,
 	mjson_skipped_first_braces_start: bool,
 	mjson_skipped_first_braces_end: bool,
 }
@@ -50,6 +62,9 @@ marshal :: proc(v: any, opt: Marshal_Options = {}, allocator := context.allocato
 	defer if err != nil {
 		strings.builder_destroy(&b)
 	}
+	
+	// temp guard in case we are sorting map keys, which will use temp allocations
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = allocator == context.temp_allocator)
 
 	opt := opt
 	marshal_to_builder(&b, v, &opt) or_return
@@ -213,6 +228,9 @@ marshal_to_writer :: proc(w: io.Writer, v: any, opt: ^Marshal_Options) -> (err:
 	case runtime.Type_Info_Matrix:
 		return .Unsupported_Type
 
+	case runtime.Type_Info_Bit_Field:
+		return .Unsupported_Type
+
 	case runtime.Type_Info_Array:
 		opt_write_start(w, opt, '[') or_return
 		for i in 0..<info.count {
@@ -263,36 +281,81 @@ marshal_to_writer :: proc(w: io.Writer, v: any, opt: ^Marshal_Options) -> (err:
 			map_cap := uintptr(runtime.map_cap(m^))
 			ks, vs, hs, _, _ := runtime.map_kvh_data_dynamic(m^, info.map_info)
 
-			i := 0
-			for bucket_index in 0..<map_cap {
-				runtime.map_hash_is_valid(hs[bucket_index]) or_continue
+			if !opt.sort_maps_by_key {
+				i := 0
+				for bucket_index in 0..<map_cap {
+					runtime.map_hash_is_valid(hs[bucket_index]) or_continue
 
-				opt_write_iteration(w, opt, i) or_return
-				i += 1
+					opt_write_iteration(w, opt, i) or_return
+					i += 1
 
-				key   := rawptr(runtime.map_cell_index_dynamic(ks, info.map_info.ks, bucket_index))
-				value := rawptr(runtime.map_cell_index_dynamic(vs, info.map_info.vs, bucket_index))
+					key   := rawptr(runtime.map_cell_index_dynamic(ks, info.map_info.ks, bucket_index))
+					value := rawptr(runtime.map_cell_index_dynamic(vs, info.map_info.vs, bucket_index))
 
-				// check for string type
-				{
-					v := any{key, info.key.id}
-					ti := runtime.type_info_base(type_info_of(v.id))
-					a := any{v.data, ti.id}
-					name: string
+					// check for string type
+					{
+						v := any{key, info.key.id}
+						ti := runtime.type_info_base(type_info_of(v.id))
+						a := any{v.data, ti.id}
+						name: string
 
-					#partial switch info in ti.variant {
-					case runtime.Type_Info_String:
-						switch s in a {
-						case string: name = s
-						case cstring: name = string(s)
+						#partial switch info in ti.variant {
+						case runtime.Type_Info_String:
+							switch s in a {
+							case string: name = s
+							case cstring: name = string(s)
+							}
+							opt_write_key(w, opt, name) or_return
+
+						case: return .Unsupported_Type
 						}
-						opt_write_key(w, opt, name) or_return
-
-					case: return .Unsupported_Type
 					}
+
+					marshal_to_writer(w, any{value, info.value.id}, opt) or_return
+				}
+			} else {
+				Entry :: struct {
+					key: string,
+					value: any,
 				}
 
-				marshal_to_writer(w, any{value, info.value.id}, opt) or_return
+				// If we are sorting the map by key, then we temp alloc an array
+				// and sort it, then output the result.
+				sorted := make([dynamic]Entry, 0, map_cap, context.temp_allocator)
+				for bucket_index in 0..<map_cap {
+					runtime.map_hash_is_valid(hs[bucket_index]) or_continue
+
+					key   := rawptr(runtime.map_cell_index_dynamic(ks, info.map_info.ks, bucket_index))
+					value := rawptr(runtime.map_cell_index_dynamic(vs, info.map_info.vs, bucket_index))
+					name: string
+
+					// check for string type
+					{
+						v := any{key, info.key.id}
+						ti := runtime.type_info_base(type_info_of(v.id))
+						a := any{v.data, ti.id}
+
+						#partial switch info in ti.variant {
+						case runtime.Type_Info_String:
+							switch s in a {
+							case string: name = s
+							case cstring: name = string(s)
+							}
+
+						case: return .Unsupported_Type
+						}
+					}
+
+					append(&sorted, Entry { key = name, value = any{value, info.value.id}})
+				}
+
+				slice.sort_by(sorted[:], proc(i, j: Entry) -> bool { return i.key < j.key })
+
+				for s, i in sorted {
+					opt_write_iteration(w, opt, i) or_return
+					opt_write_key(w, opt, s.key) or_return
+					marshal_to_writer(w, s.value, opt) or_return
+				}
 			}
 		}
 
@@ -424,8 +487,9 @@ opt_write_key :: proc(w: io.Writer, opt: ^Marshal_Options, name: string) -> (err
 
 // insert start byte and increase indentation on pretty
 opt_write_start :: proc(w: io.Writer, opt: ^Marshal_Options, c: byte) -> (err: io.Error)  {
-	// skip mjson starting braces
-	if opt.spec == .MJSON && !opt.mjson_skipped_first_braces_start {
+	// Skip MJSON starting braces. We make sure to only do this for c == '{',
+	// skipping a starting '[' is not allowed.
+	if opt.spec == .MJSON && !opt.mjson_skipped_first_braces_start && opt.indentation == 0 && c == '{' {
 		opt.mjson_skipped_first_braces_start = true
 		return
 	}
@@ -473,11 +537,9 @@ opt_write_iteration :: proc(w: io.Writer, opt: ^Marshal_Options, iteration: int)
 
 // decrease indent, write spacing and insert end byte
 opt_write_end :: proc(w: io.Writer, opt: ^Marshal_Options, c: byte) -> (err: io.Error)  {
-	if opt.spec == .MJSON && opt.mjson_skipped_first_braces_start && !opt.mjson_skipped_first_braces_end {
-		if opt.indentation == 0 {
-			opt.mjson_skipped_first_braces_end = true
-			return
-		}
+	if opt.spec == .MJSON && opt.mjson_skipped_first_braces_start && !opt.mjson_skipped_first_braces_end && opt.indentation == 0 && c == '}' {
+		opt.mjson_skipped_first_braces_end = true
+		return
 	}
 
 	opt.indentation -= 1

core/encoding/json/types.odin

@@ -1,5 +1,7 @@
 package json
 
+import "core:strings"
+
 /*
 	JSON 
 		strict JSON
@@ -104,4 +106,27 @@ destroy_value :: proc(value: Value, allocator := context.allocator) {
 	case String:
 		delete(v)
 	}
+}
+
+clone_value :: proc(value: Value, allocator := context.allocator) -> Value {
+	context.allocator = allocator
+
+	#partial switch &v in value {
+	case Object:
+		new_o := make(Object, len(v))
+		for key, elem in v {
+			new_o[strings.clone(key)] = clone_value(elem)
+		}
+		return new_o
+	case Array:
+		new_a := make(Array, len(v))
+		for elem, idx in v {
+			new_a[idx] = clone_value(elem)
+		}
+		return new_a
+	case String:
+		return strings.clone(v)
+	}
+
+	return value
 }

core/encoding/json/unmarshal.odin

@@ -5,7 +5,7 @@ import "core:math"
 import "core:reflect"
 import "core:strconv"
 import "core:strings"
-import "core:runtime"
+import "base:runtime"
 
 Unmarshal_Data_Error :: enum {
 	Invalid_Data,
@@ -137,9 +137,9 @@ assign_float :: proc(val: any, f: $T) -> bool {
 	case complex64:  dst = complex(f32(f), 0)
 	case complex128: dst = complex(f64(f), 0)
 	
-	case quaternion64:  dst = quaternion(f16(f), 0, 0, 0)
-	case quaternion128: dst = quaternion(f32(f), 0, 0, 0)
-	case quaternion256: dst = quaternion(f64(f), 0, 0, 0)
+	case quaternion64:  dst = quaternion(w=f16(f), x=0, y=0, z=0)
+	case quaternion128: dst = quaternion(w=f32(f), x=0, y=0, z=0)
+	case quaternion256: dst = quaternion(w=f64(f), x=0, y=0, z=0)
 	
 	case: return false
 	}
@@ -492,7 +492,6 @@ unmarshal_object :: proc(p: ^Parser, v: any, end_token: Token_Kind) -> (err: Unm
 			}
 		}
 
-		return nil
 	case:
 		return UNSUPPORTED_TYPE
 	}

core/encoding/xml/debug_print.odin

@@ -1,3 +1,5 @@
+package xml
+
 /*
 	An XML 1.0 / 1.1 parser
 
@@ -9,7 +11,7 @@
 	List of contributors:
 		Jeroen van Rijn: Initial implementation.
 */
-package xml
+
 
 import "core:io"
 import "core:fmt"
@@ -81,4 +83,4 @@ print_element :: proc(writer: io.Writer, doc: ^Document, element_id: Element_ID,
 	}
 
 	return written, .None
-}
+}

core/encoding/xml/example/xml_example.odin

@@ -20,7 +20,7 @@ example :: proc() {
 		xml.destroy(docs[round])
 	}
 
-	DOC :: #load("../../../../tests/core/assets/XML/unicode.xml")
+	DOC :: #load("../../../../tests/core/assets/XML/utf8.xml")
 	input := DOC
 
 	for round in 0..<N {
@@ -109,4 +109,4 @@ main :: proc() {
 		}
 	}
 	println("Done and cleaned up!")
-}
+}

core/encoding/xml/helpers.odin

@@ -1,3 +1,5 @@
+package xml
+
 /*
 	An XML 1.0 / 1.1 parser
 
@@ -6,7 +8,7 @@
 
 	This file contains helper functions.
 */
-package xml
+
 
 // Find parent's nth child with a given ident.
 find_child_by_ident :: proc(doc: ^Document, parent_id: Element_ID, ident: string, nth := 0) -> (res: Element_ID, found: bool) {
@@ -47,4 +49,4 @@ find_attribute_val_by_key :: proc(doc: ^Document, parent_id: Element_ID, key: st
 		if attr.key == key { return attr.val, true }
 	}
 	return "", false
-}
+}

core/encoding/xml/tokenizer.odin

@@ -1,3 +1,5 @@
+package xml
+
 /*
 	An XML 1.0 / 1.1 parser
 
@@ -9,7 +11,7 @@
 	List of contributors:
 		Jeroen van Rijn: Initial implementation.
 */
-package xml
+
 
 import "core:fmt"
 import "core:unicode"
@@ -433,4 +435,4 @@ scan :: proc(t: ^Tokenizer) -> Token {
 		lit = string(t.src[offset : t.offset])
 	}
 	return Token{kind, lit, pos}
-}
+}

core/encoding/xml/xml_reader.odin

@@ -1,39 +1,39 @@
 /*
-	An XML 1.0 / 1.1 parser
+ XML 1.0 / 1.1 parser
 
-	Copyright 2021-2022 Jeroen van Rijn <nom@duclavier.com>.
-	Made available under Odin's BSD-3 license.
+ 2021-2022 Jeroen van Rijn <nom@duclavier.com>.
+ available under Odin's BSD-3 license.
 
-	A from-scratch XML implementation, loosely modelled on the [spec](https://www.w3.org/TR/2006/REC-xml11-20060816).
+ from-scratch XML implementation, loosely modelled on the [spec](https://www.w3.org/TR/2006/REC-xml11-20060816).
 
-	Features:
-		- Supports enough of the XML 1.0/1.1 spec to handle the 99.9% of XML documents in common current usage.
-		- Simple to understand and use. Small.
+Features:
+- Supports enough of the XML 1.0/1.1 spec to handle the 99.9% of XML documents in common current usage.
+- Simple to understand and use. Small.
 
-	Caveats:
-		- We do NOT support HTML in this package, as that may or may not be valid XML.
-		  If it works, great. If it doesn't, that's not considered a bug.
+Caveats:
+- We do NOT support HTML in this package, as that may or may not be valid XML.
+  If it works, great. If it doesn't, that's not considered a bug.
 
-		- We do NOT support UTF-16. If you have a UTF-16 XML file, please convert it to UTF-8 first. Also, our condolences.
-		- <[!ELEMENT and <[!ATTLIST are not supported, and will be either ignored or return an error depending on the parser options.
+- We do NOT support UTF-16. If you have a UTF-16 XML file, please convert it to UTF-8 first. Also, our condolences.
+- <[!ELEMENT and <[!ATTLIST are not supported, and will be either ignored or return an error depending on the parser options.
 
-	MAYBE:
-	- XML writer?
-	- Serialize/deserialize Odin types?
+MAYBE:
+- XML writer?
+- Serialize/deserialize Odin types?
 
-	List of contributors:
-		Jeroen van Rijn: Initial implementation.
+List of contributors:
+- Jeroen van Rijn: Initial implementation.
 */
 package xml
 // An XML 1.0 / 1.1 parser
 
 import "core:bytes"
 import "core:encoding/entity"
-import "core:intrinsics"
+import "base:intrinsics"
 import "core:mem"
 import "core:os"
 import "core:strings"
-import "core:runtime"
+import "base:runtime"
 
 likely :: intrinsics.expect
 
@@ -43,48 +43,32 @@ DEFAULT_OPTIONS :: Options{
 }
 
 Option_Flag :: enum {
-	/*
-		If the caller says that input may be modified, we can perform in-situ parsing.
-		If this flag isn't provided, the XML parser first duplicates the input so that it can.
-	*/
+	// If the caller says that input may be modified, we can perform in-situ parsing.
+	// If this flag isn't provided, the XML parser first duplicates the input so that it can.
 	Input_May_Be_Modified,
 
-	/*
-		Document MUST start with `<?xml` prologue.
-	*/
+	// Document MUST start with `<?xml` prologue.
 	Must_Have_Prolog,
 
-	/*
-		Document MUST have a `<!DOCTYPE`.
-	*/
+	// Document MUST have a `<!DOCTYPE`.
 	Must_Have_DocType,
 
-	/*
-		By default we skip comments. Use this option to intern a comment on a parented Element.
-	*/
+	// By default we skip comments. Use this option to intern a comment on a parented Element.
 	Intern_Comments,
 
-	/*
-		How to handle unsupported parts of the specification, like <! other than <!DOCTYPE and <![CDATA[
-	*/
+	// How to handle unsupported parts of the specification, like <! other than <!DOCTYPE and <![CDATA[
 	Error_on_Unsupported,
 	Ignore_Unsupported,
 
-	/*
-		By default CDATA tags are passed-through as-is.
-		This option unwraps them when encountered.
-	*/
+	// By default CDATA tags are passed-through as-is.
+	// This option unwraps them when encountered.
 	Unbox_CDATA,
 
-	/*
-		By default SGML entities like `&gt;`, `&#32;` and `&#x20;` are passed-through as-is.
-		This option decodes them when encountered.
-	*/
+	// By default SGML entities like `&gt;`, `&#32;` and `&#x20;` are passed-through as-is.
+	// This option decodes them when encountered.
 	Decode_SGML_Entities,
 
-	/*
-		If a tag body has a comment, it will be stripped unless this option is given.
-	*/
+	// If a tag body has a comment, it will be stripped unless this option is given.
 	Keep_Tag_Body_Comments,
 }
 Option_Flags :: bit_set[Option_Flag; u16]
@@ -97,28 +81,20 @@ Document :: struct {
 	encoding: Encoding,
 
 	doctype: struct {
-		/*
-			We only scan the <!DOCTYPE IDENT part and skip the rest.
-		*/
+		// We only scan the <!DOCTYPE IDENT part and skip the rest.
 		ident:   string,
 		rest:    string,
 	},
 
-	/*
-		If we encounter comments before the root node, and the option to intern comments is given, this is where they'll live.
-		Otherwise they'll be in the element tree.
-	*/
+	// If we encounter comments before the root node, and the option to intern comments is given, this is where they'll live.
+	// Otherwise they'll be in the element tree.
 	comments: [dynamic]string,
 
-	/*
-		Internal
-	*/
+	// Internal
 	tokenizer: ^Tokenizer,
 	allocator: mem.Allocator,
 
-	/*
-		Input. Either the original buffer, or a copy if `.Input_May_Be_Modified` isn't specified.
-	*/
+	// Input. Either the original buffer, or a copy if `.Input_May_Be_Modified` isn't specified.
 	input:           []u8,
 	strings_to_free: [dynamic]string,
 }
@@ -158,34 +134,24 @@ Encoding :: enum {
 	UTF_8,
 	ISO_8859_1,
 
-	/*
-		Aliases
-	*/
+	// Aliases
 	LATIN_1 = ISO_8859_1,
 }
 
 Error :: enum {
-	/*
-		General return values.
-	*/
+	// General return values.
 	None = 0,
 	General_Error,
 	Unexpected_Token,
 	Invalid_Token,
 
-	/*
-		Couldn't find, open or read file.
-	*/
+	// Couldn't find, open or read file.
 	File_Error,
 
-	/*
-		File too short.
-	*/
+	// File too short.
 	Premature_EOF,
 
-	/*
-		XML-specific errors.
-	*/
+	// XML-specific errors.
 	No_Prolog,
 	Invalid_Prolog,
 	Too_Many_Prologs,
@@ -194,11 +160,9 @@ Error :: enum {
 	Too_Many_DocTypes,
 	DocType_Must_Preceed_Elements,
 
-	/*
-		If a DOCTYPE is present _or_ the caller
-		asked for a specific DOCTYPE and the DOCTYPE
-		and root tag don't match, we return `.Invalid_DocType`.
-	*/
+	// If a DOCTYPE is present _or_ the caller
+	// asked for a specific DOCTYPE and the DOCTYPE
+	// and root tag don't match, we return `.Invalid_DocType`.
 	Invalid_DocType,
 
 	Invalid_Tag_Value,
@@ -211,27 +175,20 @@ Error :: enum {
 	Unsupported_Version,
 	Unsupported_Encoding,
 
-	/*
-		<!FOO are usually skipped.
-	*/
+	// <!FOO are usually skipped.
 	Unhandled_Bang,
 
 	Duplicate_Attribute,
 	Conflicting_Options,
 }
 
-/*
-	Implementation starts here.
-*/
 parse_bytes :: proc(data: []u8, options := DEFAULT_OPTIONS, path := "", error_handler := default_error_handler, allocator := context.allocator) -> (doc: ^Document, err: Error) {
 	data := data
 	context.allocator = allocator
 
 	opts := validate_options(options) or_return
 
-	/*
-		If `.Input_May_Be_Modified` is not specified, we duplicate the input so that we can modify it in-place.
-	*/
+	// If `.Input_May_Be_Modified` is not specified, we duplicate the input so that we can modify it in-place.
 	if .Input_May_Be_Modified not_in opts.flags {
 		data = bytes.clone(data)
 	}
@@ -252,10 +209,8 @@ parse_bytes :: proc(data: []u8, options := DEFAULT_OPTIONS, path := "", error_ha
 	element, parent: Element_ID
 	open: Token
 
-	/*
-		If a DOCTYPE is present, the root tag has to match.
-		If an expected DOCTYPE is given in options (i.e. it's non-empty), the DOCTYPE (if present) and root tag have to match.
-	*/
+	// If a DOCTYPE is present, the root tag has to match.
+	// If an expected DOCTYPE is given in options (i.e. it's non-empty), the DOCTYPE (if present) and root tag have to match.
 	expected_doctype := options.expected_doctype
 
 	loop: for {
@@ -263,17 +218,13 @@ parse_bytes :: proc(data: []u8, options := DEFAULT_OPTIONS, path := "", error_ha
 		// NOTE(Jeroen): This is faster as a switch.
 		switch t.ch {
 		case '<':
-			/*
-				Consume peeked `<`
-			*/
+			// Consume peeked `<`
 			advance_rune(t)
 
 			open = scan(t)
 			// NOTE(Jeroen): We're not using a switch because this if-else chain ordered by likelihood is 2.5% faster at -o:size and -o:speed.
 			if likely(open.kind, Token_Kind.Ident) == .Ident {
-				/*
-					e.g. <odin - Start of new element.
-				*/
+				// e.g. <odin - Start of new element.
 				element = new_element(doc)
 				if element == 0 { // First Element
 					parent = element
@@ -286,11 +237,9 @@ parse_bytes :: proc(data: []u8, options := DEFAULT_OPTIONS, path := "", error_ha
 
 				parse_attributes(doc, &doc.elements[element].attribs) or_return
 
-				/*
-					If a DOCTYPE is present _or_ the caller
-					asked for a specific DOCTYPE and the DOCTYPE
-					and root tag don't match, we return .Invalid_Root_Tag.
-				*/
+				// If a DOCTYPE is present _or_ the caller
+				// asked for a specific DOCTYPE and the DOCTYPE
+				// and root tag don't match, we return .Invalid_Root_Tag.
 				if element == 0 { // Root tag?
 					if len(expected_doctype) > 0 && expected_doctype != open.text {
 						error(t, t.offset, "Root Tag doesn't match DOCTYPE. Expected: %v, got: %v\n", expected_doctype, open.text)
@@ -298,23 +247,17 @@ parse_bytes :: proc(data: []u8, options := DEFAULT_OPTIONS, path := "", error_ha
 					}
 				}
 
-				/*
-					One of these should follow:
-					- `>`,  which means we've just opened this tag and expect a later element to close it.
-					- `/>`, which means this is an 'empty' or self-closing tag.
-				*/
+				// One of these should follow:
+				// - `>`,  which means we've just opened this tag and expect a later element to close it.
+				// - `/>`, which means this is an 'empty' or self-closing tag.
 				end_token := scan(t)
 				#partial switch end_token.kind {
 				case .Gt:
-					/*
-						We're now the new parent.
-					*/
+					// We're now the new parent.
 					parent = element
 
 				case .Slash:
-					/*
-						Empty tag. Close it.
-					*/
+					// Empty tag. Close it.
 					expect(t, .Gt) or_return
 					parent      = doc.elements[element].parent
 					element     = parent
@@ -325,9 +268,7 @@ parse_bytes :: proc(data: []u8, options := DEFAULT_OPTIONS, path := "", error_ha
 				}
 
 			} else if open.kind == .Slash {
-				/*
-					Close tag.
-				*/
+				// Close tag.
 				ident := expect(t, .Ident) or_return
 				_      = expect(t, .Gt)    or_return
 
@@ -339,9 +280,7 @@ parse_bytes :: proc(data: []u8, options := DEFAULT_OPTIONS, path := "", error_ha
 				element     = parent
 
 			} else if open.kind == .Exclaim {
-				/*
-					<!
-				*/
+				// <!
 				next := scan(t)
 				#partial switch next.kind {
 				case .Ident:
@@ -370,10 +309,8 @@ parse_bytes :: proc(data: []u8, options := DEFAULT_OPTIONS, path := "", error_ha
 					}
 
 				case .Dash:
-					/*
-						Comment: <!-- -->.
-						The grammar does not allow a comment to end in --->
-					*/
+					// Comment: <!-- -->.
+					// The grammar does not allow a comment to end in --->
 					expect(t, .Dash)
 					comment := scan_comment(t) or_return
 
@@ -395,23 +332,17 @@ parse_bytes :: proc(data: []u8, options := DEFAULT_OPTIONS, path := "", error_ha
 				}
 
 			} else if open.kind == .Question {
-				/*
-					<?xml
-				*/
+				// <?xml
 				next := scan(t)
 				#partial switch next.kind {
 				case .Ident:
 					if len(next.text) == 3 && strings.equal_fold(next.text, "xml") {
 						parse_prologue(doc) or_return
 					} else if len(doc.prologue) > 0 {
-						/*
-							We've already seen a prologue.
-						*/
+						// We've already seen a prologue.
 						return doc, .Too_Many_Prologs
 					} else {
-						/*
-							Could be `<?xml-stylesheet`, etc. Ignore it.
-						*/
+						// Could be `<?xml-stylesheet`, etc. Ignore it.
 						skip_element(t) or_return
 					}
 				case:
@@ -425,15 +356,11 @@ parse_bytes :: proc(data: []u8, options := DEFAULT_OPTIONS, path := "", error_ha
 			}
 
 		case -1:
-			/*
-				End of file.
-			*/
+			// End of file.
 			break loop
 
 		case:
-			/*
-				This should be a tag's body text.
-			*/
+			// This should be a tag's body text.
 			body_text        := scan_string(t, t.offset) or_return
 			needs_processing := .Unbox_CDATA          in opts.flags
 			needs_processing |= .Decode_SGML_Entities in opts.flags
@@ -613,9 +540,7 @@ parse_prologue :: proc(doc: ^Document) -> (err: Error) {
 				doc.encoding = .LATIN_1
 
 			case:
-				/*
-					Unrecognized encoding, assume UTF-8.
-				*/
+				// Unrecognized encoding, assume UTF-8.
 				error(t, offset, "[parse_prologue] Warning: Unrecognized encoding: %v\n", attr.val)
 			}
 
@@ -658,11 +583,11 @@ skip_element :: proc(t: ^Tokenizer) -> (err: Error) {
 
 parse_doctype :: proc(doc: ^Document) -> (err: Error) {
 	/*
-		<!DOCTYPE greeting SYSTEM "hello.dtd">
+	<!DOCTYPE greeting SYSTEM "hello.dtd">
 
-		<!DOCTYPE greeting [
-			<!ELEMENT greeting (#PCDATA)>
-		]>
+	<!DOCTYPE greeting [
+		<!ELEMENT greeting (#PCDATA)>
+	]>
 	*/
 	assert(doc != nil)
 	context.allocator = doc.allocator
@@ -675,9 +600,7 @@ parse_doctype :: proc(doc: ^Document) -> (err: Error) {
 	offset := t.offset
 	skip_element(t) or_return
 
-	/*
-		-1 because the current offset is that of the closing tag, so the rest of the DOCTYPE tag ends just before it.
-	*/
+	// 	-1 because the current offset is that of the closing tag, so the rest of the DOCTYPE tag ends just before it.
 	doc.doctype.rest = string(t.src[offset : t.offset - 1])
 	return .None
 }
@@ -700,4 +623,4 @@ new_element :: proc(doc: ^Document) -> (id: Element_ID) {
 	cur := doc.element_count
 	doc.element_count += 1
 	return cur
-}
+}