new view rule table definition in df, merging graphical view rules & views totally

This commit is contained in:
Ryan Fleury
2024-09-13 13:39:40 -07:00
parent 7cd1d175de
commit 793ad8fe1e
10 changed files with 296 additions and 159 deletions
@@ -1,6 +1,93 @@
// Copyright (c) 2024 Epic Games Tools
// Licensed under the MIT license (https://opensource.org/license/mit/)
////////////////////////////////
//~ rjf: Built-In View Rules
//
// @view_rule_info
//
// NOTE(rjf): View rules are subtle in that they may impact any subset of the
// eval visualization pipeline. The "array" view rule, for example, functions
// by tweaking the type of an eval from `X *` to `X (*)[N]` (where N is
// computed from whatever expression is specified by the view rule). The "list"
// view rule, on the other hand, does not require any changes to the actual
// eval nor its type - instead, it follows an alternative path in constructing
// "viz blocks", and then constructing "viz rows" from those blocks. Compare
// these to the simpler 'dec', 'bin', or 'oct' rules, which simply tweak the
// radix used when stringizing numbers, which is something that only occurs in
// single-line eval stringization building.
//
// As such, each view rule specification has a mask, which determines which
// stages it may be used for. For a given view rule specification, if the bit
// corresponding to a particular eval stage is set, then that view rule spec-
// -ification also includes a hook which can be called from that stage.
//
// Below is a list of the stages in the eval visualization pipeline, as well as
// abbreviations which are used in the tables.
//
// expr resolution, "xp" -> provides a chance for a view rule to make
// modifications to expression trees that it is
// applied to
//
// viz block prod, "vb" -> given a resolved eval, produce a list of non-
// windowed "viz blocks", which correspond to one or
// many contiguous rows in a watch-window-style UI.
// one level of expanded struct members, with no sub-
// expansions, would be one viz block. if one of those
// members - in the middle - were expanded too, then
// it would require three viz blocks - one for the
// members before the sub-expansion, one for the
// sub expansion members, and one for the members
// after, and so on. this is done recursively.
//
// viz row prod, "vr" -> given a list of viz blocks, a windowed list of viz
// rows may be produced. each of these rows has info
// for building actual UI in e.g. a watch window -
// whether or not the row can be expanded, whether or
// not the row's value can be edited, what the edit-
// able string is for a row, what the display string
// is for a row, what the expression string is for a
// row, what the type is for a row, and so on.
//
// line stringize, "ls" -> this is the stage used to produce display strings
// in the "viz row prod" stage, as well as basically
// any time UI needs to display the result of an eval
// in a single line. this also occurs recursively,
// descending into members & elements as needed,
// constrained by # of available pixels and font size
// and so on.
//
// row ui build, "ru" -> finally, after the previous stages are completed,
// ui can finally be built according to all of the
// per-row information produced. this is the stage
// where view rules can insert their own arbitrary ui
// on a per-row basis.
//
// view ui build, "vu" -> view rules which want to supply more sophisticated
// visualizers have the ability to provide full
// arbitrary UI hooks, which can either be produced
// in a minified form via watch views, or via a
// standalone tab.
//
// A few other bits are included for various ways in which a view rule may be
// applied throughout the eval visualization pipeline. A list follows:
//
// inherited, "ih" -> is this view rule included, or not included, in
// child expansions?
//
// expandable, "ex" -> does this view rule force the ability to expand
// an expression, even if traditional analysis of type
// info would not allow expansion?
//
// Not all of these stages are specified at this layer, however, since the
// "df_core" layer is for the non-graphical core debugger features. So the
// information pertaining to the eval visualization pipeline stages which
// do require graphical subsystems (e.g. UI, fonts, rendering) are specified
// in the "df_gfx" layer.
//
// For any view rules in this layer which also have graphical features, they
// are specified in both tables under the same name.
@table(coverage_check name name_lower string ih ex xp vb display_name docs schema description)
EV_ViewRuleTable:
{
@@ -178,7 +178,7 @@ EV_Key key, \
EV_ExpandNode *expand_node, \
String8 string, \
E_Expr *expr, \
EV_ViewRuleList *view_rules, \
EV_ViewRuleList *view_rules, \
MD_Node *view_params, \
S32 depth, \
struct EV_BlockList *out)