codegraph/docs/design/value-reference-edges.md

Design + status: same-file value-reference edges

Status: SHIPPED (default-on for TS/JS/tsx + Go + Python + Rust + Ruby + C + Java + C# + PHP + Scala + Kotlin + Swift + Dart; CODEGRAPH_VALUE_REFS=0 disables). The emitter lives in TreeSitterExtractor.flushValueRefs (src/extraction/tree-sitter.ts). Motivation: close the impact-analysis hole for value consumers. Static extraction edges calls, imports, and inheritance, but never edges a constant to the symbols that read it — so changing a config object / lookup table / shared constant looked like "nothing depends on this." This is the "change this table, break its readers" class of change (the ReScript-PR false positive that motivated the work).

---

TL;DR for a new session

We emit a references edge (metadata: { valueRef: true }) from a reader symbol to the file/package-scope const/var it reads, same-file only, for TS/JS/tsx + Go + Python + Rust + Ruby + C + Java + C# + PHP + Scala + Kotlin + Swift + Dart. Those edges flow straight into getImpactRadius / codegraph impact and the impact trail in codegraph_explore / codegraph_node — no agent-behaviour change required.

The win is impact-radius correctness, not agent read-reduction (see "Agent A/B").

Edge semantics

• Target: a file-scope const/var whose name is "distinctive" (≥3 chars and contains an uppercase letter or _) — dodges the local-shadowing precision trap that single-letter / all-lowercase names invite.
• Reader (source): any function / method / const / var symbol whose body references the target name.
• Same-file only — resolution is unambiguous without import/scope analysis.
• Deduped per (reader, target). Additive — adds edges, never nodes.

Precision guards (in emission order)

1. isGeneratedFile(path) — skip suffix-recognised generated files (.pb.ts, .min.js, …). Path-only; it cannot catch content-minified bundles.
2. Shadow prune — drop a target when its declarator count exceeds its file-scope node count, i.e. it's also bound in an inner (local) scope. A bundled/Emscripten const Module re-declared as an inner var Module, a Go package const shadowed by a local :=, or a Python module const shadowed by a local = all resolve to the inner binding for nested readers — a file-scope edge would be a false positive. Inner re-bindings aren't graph nodes, so declarators are counted at the syntax level (per-grammar node types: variable_declarator for TS/JS, const_spec/var_spec/short_var_declaration for Go, assignment for Python, const_item/static_item/let_declaration for Rust). Comparing against file-scope node count (not a flat ">1") keeps conditional module defs (try: X=…; except: X=…), which legitimately bind a name twice at file scope. This catches the content-minified bundles guard #1 misses.
3. Distinctive-name + same-file as above.

Validation matrix — TS / JS / Go / Python / Rust / Ruby / C / Java / C# / PHP / Scala / Kotlin / Swift / Dart

Method per repo: index the same tree twice (value-refs on vs CODEGRAPH_VALUE_REFS=0), diff node/edge counts, spot-check precision, and measure codegraph impact on a few file-scope consts. Node count must be identical on/off (edges-only feature).

TypeScript

Repo	size	files	nodes (on=off)	+value-ref edges	precision	impact on→off example
sindresorhus/ky	small	54	562 (stable)	+29 (0.8%)	all sampled TP	—
excalidraw/excalidraw	medium	645	10,301 (stable)	+717 (1.6%)	TP after shadow prune (#895 removed 23 woff2-bundle FPs)	tablerIconProps 1→170
microsoft/vscode	large	11,548	333,999 (stable)	+10,605 (0.69%)	all sampled TP; no param-shadow / bundle FPs in top 200	LayoutStateKeys 1→85, CORE_WEIGHT 1→52

JavaScript (same extractor; CommonJS, var, IIFE/UMD)

Repo	size	files	nodes (on=off)	+value-ref edges	precision	impact on→off example
expressjs/express	small	147	1,082 (stable)	+27 (0.75%)	all sampled TP	—
eslint/eslint	medium	1,420	7,167 (stable)	+1,192 (4.2%)	all sampled TP; guard holds; no minified-file FPs	internalSlotsMap 1→32, INDEX_MAP 1→27
webpack/webpack	large	9,371	28,922 (stable)	+3,521 (4.8%)	all sampled TP; guard holds; no minified-file FPs	LogType 1→89, LOG_SYMBOL 1→90, UsageState 2→52

Go (package-level const/var; required extending the shadow prune — see below)

Repo	size	files	nodes (on=off)	+value-ref edges	precision	impact on→off example
gin-gonic/gin	small	110	2,599 (stable)	+166 (1.9%)	all sampled TP; guard holds	abortIndex 1→24, jsonContentType 1→8
gohugoio/hugo	medium	952	19,160 (stable)	+1,616 (2.5%)	all sampled TP; guard holds	filepathSeparator 2→26
prometheus/prometheus	large	1,329	23,322 (stable)	+3,466 (3.3%)	all sampled TP; guard holds	rdsLabelInstance 1→82, ec2Label 1→24
kubernetes/kubernetes	very large	19,160	251,086 (stable)	+20,574 (1.9%)	all sampled TP; guard holds on 250 targets	KubeletSubsystem 3→138, LEVEL_0 1→102

Python (module-level NAME = …; required extending the prune and refining its rule — see below)

Repo	size	files	nodes (on=off)	+value-ref edges	precision	impact on→off example
psf/requests	small	49	1,299 (stable)	+85 (2.9%)	all sampled TP; guard holds	ITER_CHUNK_SIZE 1→4, DEFAULT_POOLBLOCK 1→4
sqlalchemy/sqlalchemy	medium	679	59,963 (stable)	+1,929 (0.8%)	all sampled TP; guard holds	COMPARE_FAILED 1→26, DB_LINK_PLACEHOLDER 1→19
django/django	large	3,005	61,748 (stable)	+1,328 (0.7%)	all sampled TP; guard holds	_trans 1→138, SEARCH_VAR 4→8

Rust (module-level const/static; declarators added, no rule change needed)

Repo	size	files	nodes (on=off)	+value-ref edges	precision	impact on→off example
BurntSushi/ripgrep	small	107	3,731 (stable)	+144 (0.9%)	all sampled TP; guard holds	SHERLOCK 7→113
tokio-rs/tokio	medium	795	13,281 (stable)	+476 (1.1%)	all sampled TP; #[cfg]-conditional consts kept	PERMIT_SHIFT 1→97, LOCAL_QUEUE_CAPACITY 2→46
rust-lang/rust-analyzer	large	1,530	38,780 (stable)	+475 (0.25%)	all sampled TP; 0 real shadow leaks	INLINE_CAP 2→183, SPAN_PARTS_BIT 2→18

Ruby (CONST = …, almost always inside a class/module — needed the class-scope extension)

Repo	size	files	nodes (on=off)	+value-ref edges	precision	impact on→off example
sinatra/sinatra	small	96	1,800 (stable)	+73 (2.1%)	~100% TP (flags are valid nested reads)	HEADER_PARAM 1→5
jekyll/jekyll	medium	218	1,906 (stable)	+100 (2.4%)	~100% TP	DEFAULT_PRIORITY 1→3, LOG_LEVELS 4→5
rails/rails	large	1,452	61,911 (stable)	+2,255 (1.2%)	~98% TP (same-file ambiguity 21/1208 targets)	Post (Struct const) 75 readers

C (file-scope static const scalars + pointer/array lookup tables + mutable globals; required extracting the nodes first — see below)

Repo	size	files	nodes (on=off)	+value-ref edges	precision	impact on→off example
redis/hiredis	small	52	1,161 (stable)	+29 (2.5%)	all sampled TP; guard holds	hiredisAllocFns 1→71
curl/curl	large	994	16,124 (stable)	+597 (3.7%)	all sampled TP; guard holds; no minified FPs	Curl_ssl 3→57
redis/redis	medium	782	19,446 (stable)	+1,634 (8.4%)	all sampled TP after the macro-misparse fix; guard holds	asmManager 2→97, keyMetaClass 1→36, XXH3_kSecret 1→27, helpEntries 1→13

Java (class-scope static final constants; required emitting them as constant kind — see below)

Repo	size	files	nodes (on=off)	+value-ref edges	precision	impact on→off example
google/gson	small	262	8,563 (stable)	+387	all sampled TP; guard holds	PEEKED_NONE 1→31
apache/commons-lang	medium	623	19,976 (stable)	+2,087	all sampled TP; guard holds; no minified FPs	INDEX_NOT_FOUND 4→165, EMPTY 5→161
google/guava	large	3,227	130,945 (stable)	+6,354	all sampled TP; guard holds; no minified FPs	APPLICATION_TYPE 2→126, ABSENT 4→66

C# (class-scope const / static readonly; same field→constant change as Java)

Repo	size	files	nodes (on=off)	+value-ref edges	precision	impact on→off example
AutoMapper/AutoMapper	small	511	19,254 (stable)	+133	all sampled TP; guard holds	ContextParameter 1→17, InstanceFlags 1→14
JamesNK/Newtonsoft.Json	medium	945	20,208 (stable)	+344	all sampled TP; guard holds	DefaultFlags 1→37, JsonNamespaceUri 1→15
dotnet/efcore	large	5,731	140,847 (stable)	+3,720	all sampled TP; guard holds; no minified FPs	_resourceManager 22→1664, Prefix 40→237, Guid77 2→191

PHP (top-level const + class const, both already constant; needed only a reader-scan tweak — see below)

Repo	size	files	nodes (on=off)	+value-ref edges	precision	impact on→off example
guzzle/guzzle	small	81	1,655 (stable)	+5 (sparse — see note)	all sampled TP; no collisions	CONNECTION_ERRORS 1→3
Seldaek/monolog	medium	217	3,047 (stable)	+79	all sampled TP; no class/const collisions	DEFAULT_JSON_FLAGS 1→18, RFC_5424_LEVELS 1→17
laravel/framework	large	2,956	57,519 (stable)	+86	all sampled TP; no minified/collision FPs	INVISIBLE_CHARACTERS 1→93, SESSION_ID_LENGTH 1→9

Scala (top-level val + object val — re-kinded from field; class instance vals stay field)

Repo	size	files	nodes (on=off)	+value-ref edges	precision	impact on→off example
com-lihaoyi/upickle	small	145	3,052 (stable)	+82	all sampled TP; no class/method collisions	IntegralPattern 1→9
typelevel/cats	medium	835	15,774 (stable)	+89	sampled TP; flagged val/def name-collisions were real object vals read by siblings	maxArity 3→17, fusionMaxStackDepth 1→13, minIntValue 1→7
apache/pekko	large	2,720	135,041 (stable)	+8,453 (2,065 Scala)	Scala object vals clean; the bulk are valid Java PARSER/DEFAULT_INSTANCE from generated protobuf .java	ErrorLevel 5→33, WarningLevel 5→29

Kotlin (top-level / object / companion object val → constant; class instance vals stay field)

Repo	size	files	nodes (on=off)	+value-ref edges	precision	impact on→off example
square/okio	small	307	8,540 (stable)	+157	all sampled TP; 0 collisions	STATE_IN_QUEUE 1→32, HMAC_KEY 1→9
Kotlin/kotlinx.coroutines	medium	1,039	17,058 (stable)	+210	all sampled TP; 1 cross-file collision	BLOCKING_SHIFT 1→24, TERMINATED 2→22 (companion bit-masks)
ktorio/ktor	large	2,302	43,272 (stable)	+849	object/companion consts (HTTP header names); flagged collisions are real consts; TYPE is a sibling-companion ambiguity	TYPE 8→109, FailedPath 1→22

Swift (top-level let + static let in struct/enum/class → constant; instance let stays field; computed properties skipped)

Repo	size	files	nodes (on=off)	+value-ref edges	precision	impact on→off example
Alamofire/Alamofire	small	98	4,192 (stable)	+108	all sampled TP; 0 collisions; computed properties skipped	defaultRetryLimit 1→3, defaultWait 1→4
apple/swift-argument-parser	medium	165	4,435 (stable)	+36	all sampled TP; 1 sibling-type collision (usageString)	usageString 8→18, labelColumnWidth 1→2
apple/swift-nio	large	554	20,136 (stable)	+589	all sampled TP; 0 collisions; eventLoop (static let) verified TP	CONNECT_DELAYER 1→15, SINGLE_IPv4_RESULT 1→12

Dart (top-level const/final + class static const/static final = the static_final_declaration node → constant)

Repo	size	files	nodes (on=off)	+value-ref edges	precision	impact on→off example
dart-lang/http	small	324	4,860 (stable)	+668	real source TP; numbers skewed by a JNIGEN _bindings.dart (sibling-class collapse)	Finishing 1→10, CONNECTION_PREFACE 5→7
flame-engine/flame	medium	1,655	19,608 (stable)	+465	all sampled TP; bounded const-vs-getter collisions	cardWidth 4→15, tileSize 3→12
flutter/packages	large	3,452	116,075 (stable)	+10,015	real Flutter consts; some .gen.dart (pigeon) generated noise	iconFont 1→1790, _channel 6→72, kMaxId 1→23

Across S/M/L in all fourteen languages: node count never moved, the precision guards held, and the impact OFF column is the bug — a const that 80–140 symbols read reports "1 affected" without value-refs.

Go required a code change (unlike JS/tsx, which the existing guards covered unchanged). Go puts its constants at package = file scope (good — the target gate fits), but its declarators are const_spec/var_spec/short_var_declaration, not variable_declarator, so the shadow prune was a no-op for Go and a package const Timeout shadowed by a local Timeout := … produced a false positive. Extending the prune's declarator switch to Go's node types fixed it (one synthetic repro, then clean across gin/hugo/prometheus). This is the template for the next language: the shadow prune is per-grammar and must be wired per language (see the playbook).

Python forced a refinement of the prune rule — a general improvement. Python's declarator is assignment (added to the switch). But Python also conditionally defines module constants (try: HAS_SSL = True; except: HAS_SSL = False) — a very common idiom that binds the name twice at module scope. The old "bound more than once → drop" rule over-pruned these (dropping a real const and its readers). The fix distinguishes a conditional module def from a real shadow by comparing declarator count against the number of file-scope nodes the name has: a conditional def makes them equal (both bindings are file-scope), a local shadow makes declarators exceed file-scope nodes (the excess is the local). This is strictly more correct for all languages. (It also made the two halves of a conditional def cross-reference via their own names, so same-name value-ref edges are now suppressed.)

Rust needed only declarators — the rule was already right. Rust's are const_item / static_item (module consts) and let_declaration (the local that shadows). Adding them to the switch fixed the expected shadow FP (a const TIMEOUT shadowed by a local let TIMEOUT). Rust also has the conditional-def pattern — #[cfg(unix)] const SEP = …; #[cfg(windows)] const SEP = … — and the Python-era file-scope-count rule already keeps those correctly (validated on tokio's io/interest.rs cfg-gated flags). One nice property fell out: consts written inside a config macro (cfg_aio! { … }) live in an unparsed token tree, so the prune's syntax walk doesn't even see them.

Ruby is the class-scope case — and required three changes. Ruby keeps almost all constants inside a class/module (jekyll's lib/: 0 top-level vs 58 class-internal), so the original file-scope-only target gate covered ~nothing. Three Ruby-specific fixes: (1) the extractor now creates nodes for constant assignments (CONST = … has a constant-typed LHS, not identifier, so they were never extracted at all) — including class-internal ones; (2) the value-ref target gate accepts class:/module: parents, not just file:; (3) the reader-scan matches constant nodes, since in Ruby both a constant's definition and its references are constant-typed. Effectively Ruby-only: Rust impl consts are parented to file: already (so the gate change doesn't touch them — ripgrep stayed at 144 edges), and TS/Python class members aren't constant/variable kind.

The interesting precision question — which class does a class-scope target belong to — turns out to favor a file-wide target map (a name maps to one target per file), because Ruby's constant lookup is lexical + ancestor: a method in a nested class legitimately reads an enclosing class's constant (verified on jekyll's ERBRenderer→ThemeBuilder::SCAFFOLD_DIRECTORIES and sinatra's AcceptEntry→Request::HEADER_PARAM). Strict same-class matching would wrongly drop those. The only real false positive is the same constant name defined in sibling (un-nested) classes in one file — 21 of 1,208 targets (1.7%) on rails, and most of those resolve fine too; referencing a sibling class's bare constant is a NameError in real Ruby, so valid code rarely hits it. Net precision ~98–100%.

C was NOT the "easy path" the language tracker first assumed — it needed the extractor to emit the nodes first. C keeps shareable values at file scope (static const scalars, and very commonly pointer/array lookup tables + mutable global state), which fits the file-scope target gate. But unlike Go/Rust (whose const nodes already existed), C's file-scope const/var were never extracted as nodes at all: a C declaration nests its name inside an init_declarator (through pointer_declarator/array_declarator), and the generic variable-extraction fallback only finds a direct identifier child — so it produced nothing. Three changes (the same shape as Ruby's): (1) a C branch in extractVariable that resolves the name through the declarator chain and emits file-scope declarations as constant/variable (skipping function-body locals via an ancestor check, and function_declarator prototypes); (2) an isConst on the C extractor (a const type_qualifier → constant kind); (3) the shadow prune's declarator switch extended with init_declarator. Scoped to C only — C++ stays on the generic fallback (its class-scope members are the harder bucket).

The one false-positive cluster the sweep surfaced was a macro-prefixed-prototype misparse, and the fix is the load-bearing C detail: an unknown leading macro (CURL_EXTERN, XXH_PUBLIC_API) makes tree-sitter-c misparse a prototype MACRO RetType fn(args); as a declaration whose declared "variable" is the bare return-type identifier (XXH_errorcode/CURLcode), splitting fn(args) off as a bogus expression — minting one spurious type-named global per prototype, then edged by every function returning that type (redis's XXH_errorcode 1→18 before the fix). These misparses always yield a bare identifier declarator (verified across pointer/array/sized return variants); real consts/tables always carry an initializer (init_declarator) and real pointer/array globals carry their own declarator. So the C branch skips bare-identifier declarators entirely — killing the whole FP class at the cost of only uninitialized scalar globals (static int g;), which are rare and low-value. After the fix: every sampled edge on hiredis/redis/curl was a true positive, the guard-invariant leak check found 0 shadows across all three, and impact deltas confirm the blind→real radius (asmManager 2→97, Curl_ssl 3→57, hiredisAllocFns 1→71).

Java + C# were the cleanest class-scope languages — one kind switch, no new guards. Both keep constants inside a class (Java static final fields; C# const / static readonly), so unlike C the nodes already existed — but as field kind, which the value-ref gate (constant/ variable only) rejects. The whole change was emitting the constant subset as constant: an isConst predicate on each extractor (Java = a static final field; C# = a const, or a static readonly) plus a kind switch in extractField. Everything else was already in place — the class-scope target gate (from Ruby), the identifier reader-scan, and crucially the shadow prune: a method-local that shadows a class const is a variable_declarator in both grammars, already in the prune switch, so a class const shadowed by a local is dropped with no new wiring (validated by the Java/C# shadow tests). Instance fields stay field — a Java instance final or a C# instance readonly is per-object state, not a shared constant, so it's never a target. The distinctive-name gate fits both conventions cleanly (Java UPPER_SNAKE, C# PascalCase), so no FP class emerged: across S/M/L (gson/commons-lang/guava, automapper/newtonsoft/efcore) every sampled edge was a true positive, 0 shadow leaks, no minified-file FPs, node count identical on/off. The impact wins are the headline — Java's canonical public static final constants (INDEX_NOT_FOUND 4→165, EMPTY 5→161) and C#'s const/static readonly (Prefix 40→237, a generated _resourceManager 22→1664) all went from a blind "1 affected" to their real radius. The known sibling-class limitation (the same const name in two classes in one file resolves to the file-wide target) is shared with Ruby and stayed negligible.

PHP was a near-pure "easy path" — one reader-scan line, no extractor change, no prune wiring. PHP already extracts both top-level const X = … and class const X = … as constant kind (a dedicated const_declaration handler), inside the right scope (file: / class:, both gated). The only change was the reader-scan: PHP represents a constant reference — bare X, or the const half of self::X / Foo::X / static::X — as a name node, which the scan (matching identifier / constant) missed, so it found nothing until name was added. That's safe across languages: flushValueRefs only runs for the value-ref set, and name is PHP-only among them. No shadow prune was needed at all — a PHP local is a $var (variable_name), a different namespace from a bare constant, so a local can never shadow a constant; there is nothing to prune (the cleanest case yet). Precision was excellent: UPPER_SNAKE constants fit the distinctive-name gate, and a dedicated check for a target whose name collides with a same-file class (PHP's one realistic FP — name nodes also name classes in new Foo() / Foo::) found zero collisions across guzzle/monolog/laravel; every sampled edge was a true positive, node count identical on/off.

The honest caveat: PHP is lower-yield than the class-scope languages, by design. PHP idiom reads constants across files far more than within one (a Logger::DEBUG or a config constant consumed everywhere), and value-refs is same-file only — so laravel (2,956 files) produced only 86 edges vs. Ruby rails's 2,255 (1,452 files). This is not a miss: the cross-file reads are out of scope for every language (resolution would need import/scope analysis), and PHP simply leans on them more. The same-file reads it does capture are clean and the transitive impact wins are real (INVISIBLE_CHARACTERS 1→93 from 3 direct readers). Net: correct and additive, just a smaller absolute contribution than Java/C#/Go.

Scala — the object is the constant scope. Scala has no static; the idiom for a shared constant is a val inside a singleton object (object Config { val Timeout = 30 }). A top-level val already extracted as constant, but object and class vals both came out as field (the gate rejects field). The fix is a kind refinement in the Scala val_definition handler: walk to the enclosing definition and treat an object_definition (or top level) val as constant/variable — while a class/trait/enum val stays field, because it is per-instance immutable state, the exact analogue of the Java instance final we also keep as field. (object and class both extract as class kind, so the distinction is the enclosing AST node type, not the node kind.) The shadow prune gained val_definition/var_definition (a method-local val can shadow an object val); the reader-scan needed nothing, since a Scala val reference is a plain identifier. Method-local vals are not extracted at all, so they're not a target source. The one known limitation is Scala's interchangeable val/def for members: a camelCase val can share a name with a method in the same file, and same-file name matching can't distinguish them — but it's bounded (like Ruby's sibling-class case), and on the sweep every flagged val/def collision turned out to be a real object val read by sibling vals (cats' typeclass instances: val flatMap = monad, read by invariantSemigroupal). Validated S/M/L (upickle/cats/pekko): node count identical on/off, top targets genuine object vals (maxArity val = 22, DigitTens lookup table), impact wins real (maxArity 3→17). The distinctive-name gate fits Scala's camelCase/PascalCase constants (maxArity, IntegralPattern) via their internal uppercase letter.

Kotlin combined three already-built techniques. Kotlin has no static: shared constants live at top level, in an object (singleton), or in a class's companion object — all val/const val. A class instance val is per-object state. Nothing extracted before because a Kotlin property name nests (property_declaration → variable_declaration → simple_identifier) and the generic path reads only a direct child — the C problem. The fix handles property_declaration in the Kotlin visitNode hook (where the existing one already manages fun interface misparses): pull the nested name, then walk to the enclosing definition to set the kind — object_declaration/companion_object (or top level) → constant/variable (the Scala object-vs-class rule), class_declaration → field, and a property under a function_body/init/lambda is a local and skipped. The reader-scan gained simple_identifier (Kotlin's reference node — the PHP name move; simple_identifier is Kotlin-only among the value-ref set), and the shadow prune gained property_declaration (a method-local val can shadow an object const). Kotlin's parse fidelity is clean (its one known misparse, fun interface, is already handled), so unlike C++ no precision tail emerged. It validated as one of the cleanest languages: companion-object bit-masks and state constants are a heavy, same-file-read idiom (coroutines' BLOCKING_SHIFT 1→24, TERMINATED 2→22 in the scheduler; okio's STATE_IN_QUEUE 1→32; ktor's content-type TYPE 8→109). okio had 0 collisions, coroutines 1 (cross-file). The same val/def-or-class name-overlap limitation as Scala applies (ktor's HTTP DSL names a header const and a class the same), plus the sibling-companion case (several companion object { const val TYPE } in one file collapse to the file-wide target, like Ruby's sibling-class) — both bounded, and every flagged collision investigated was a real object/companion const.

Swift reused the Kotlin techniques and added two Swift-specific touches. Swift has no static keyword for globals; its shared-constant idiom is a top-level let or a static let inside a type — and Swift idiomatically namespaces constants in enum/struct (enum Constants { static let X }). A property name nests (property_declaration → <name> pattern → simple_identifier), the C-style problem; the reader-scan already matched simple_identifier (added for Kotlin — Swift shares it). The kind rule: top-level let and static let (in any type) → constant (var → variable); an instance let/var stays field (Swift instance stored properties otherwise aren't own nodes — unchanged). The two Swift-specific touches: (1) the value-ref target gate was widened to struct:/ enum: parents, because Swift namespaces constants in those (every other language's targets sit at file:/class:/module:); without it, the heavily-used enum/struct static consts would all be missed. (2) Computed properties are skipped — a var x: Int { … } has a getter block, no stored value, and isn't a constant; the extractor detects the computed_property child and emits no node (verified: no computed-property leaks across the sweep). The node creation slots into the existing Swift property_declaration handler (which already extracts property-wrapper / type-annotation dependencies like @Published/@State), so that behavior is untouched. Validated S/M/L (Alamofire/swift-argument-parser/swift-nio): node count identical on/off, genuine static-let constants (defaultRetryLimit, swift-nio's CONNECT_DELAYER/SINGLE_IPv4_RESULT test constants, a shared static let eventLoop read by 37 methods), computed properties skipped, 0–1 collisions per repo (the same sibling-type name-overlap bound as Kotlin/Ruby).

Dart — the grammar did the scope separation; the catch was a sibling body. Dart's tree-sitter grammar is unusually helpful here: a static_final_declaration node is exactly a top-level or class-static const/final — the shared-constant idiom — while instance fields and var use initialized_identifier and method-locals use initialized_variable_definition. So a single visitNode rule (static_final_declaration → constant, named by its identifier child) captures all and only the constants, with no instance/local leaks to guard and no scope-walk needed (the node stack gives file: for top-level, class: for a static member). The reader-scan was already covered (Dart references are plain identifier). The non-obvious bug: Dart attaches a method/function body as a next sibling of the signature node — and the signature is what gets stored as the reader scope — so the scan walked only the signature and produced zero edges until it was taught to also pull in a function_body next-sibling (Dart is the only value-ref language that structures bodies this way, so the check is inert elsewhere). The shadow prune counts all three Dart declarator nodes so a method-local const X correctly drops a file-scope const X. Validated S/M/L (http / flame-engine/flame / flutter/packages): node count identical on/off, genuine static consts on real source (flame's cardWidth 4→15, tileSize 3→12; HTTP/2's Finishing 1→10), the same bounded const-vs-getter name overlap as Kotlin/Scala. The one caveat is generated code: the common Dart codegen suffixes (.g.dart / .freezed.dart / .pb.dart) are already skipped by isGeneratedFile, but a header-only-marked generator (a JNIGEN _bindings.dart with hundreds of static final _class) isn't suffix-detected, so it collapses to the file-wide target and dominates a small repo's numbers (http) — real source stays clean.

C++ was attempted and reverted — the machinery (file/namespace-scope + class field_declaration extraction) is correct on clean C++, but tree-sitter-cpp's parse fidelity on real template/macro-heavy code (and the .h→C-grammar routing) leaks class members and parameters to file scope as bogus constants. Two guards (skip declarations under an ERROR or compound_statement ancestor) removed ~83% of the gross leaks, but the residual pervaded even well-structured library source (template-class member leaks, amalgamated mega-headers, .h-as-C++). It did not reach the precision bar the other languages hold, so it was reverted. Reviving C++ needs prior work on C++ parse handling (template-class member scoping, .h-as-C++ detection, amalgamated-header exclusion), not a value-refs wiring pass. See the playbook's §2b C++ note.

tsx is covered by the TS rows — excalidraw is a React/.tsx codebase, so the headline tablerIconProps (1→170) and most of its targets live in .tsx files. The one tsx-specific path — a const read only inside JSX (<Foo x={CONST}/>) — relies on the reader-scan descending into the JSX subtree; it's locked by a unit test (value-reference-edges.test.ts), so no separate tsx repo sweep is needed.

Svelte / Vue / Astro are covered for free — their extractors re-parse the <script> / frontmatter block as typescript / javascript, which are in VALUE_REF_LANGS, so a const in a .svelte/.vue/.astro script edges its readers without any extra work (verified on a synthetic .svelte). No separate matrix row. See the playbook's coverage tracker (§2b) for the full status against the README's language list.

JavaScript note — CommonJS require bindings are targets, and that's correct. JS edge growth (~4–5%) runs higher than TS (~0.7–1.6%) because var x = require('…') bindings and module-level var state pass the distinctive-name gate and are read by same-file functions. These are not noise: changing such a binding (swap the dependency, reassign the state) genuinely affects its readers, so it's a legitimate impact target. Where it overlaps an existing calls edge, getImpactRadius dedups by node — no double-counting. (TS imports dodge this entirely: they're import-kind nodes, not const/var, so never targets.)

Agent A/B — what it does and doesn't buy (excalidraw, sonnet/high, 12 runs)

• Impact API (the win): impact ON vs OFF — tablerIconProps 1→170, COLOR_PALETTE 15→26, CaptureUpdateAction 61→86. This is what codegraph impact and CodeGraph Pro's verdict engine consume via getImpactRadius.
• Agent read-displacement: none — and that's expected. On an indexed repo the agent answers impact questions in one codegraph call (0 Read / 0 Grep in both arms), and it reaches for codegraph_search / callers, not impact/explore, so it often doesn't query the value-ref edges at all. ON was never worse than OFF. Do not claim value-refs reduces agent reads — the win is blast-radius correctness, not fewer turns. (This is the "adapt the tool to the agent" wall: edges only help if the agent calls the edge-traversing tool.)

Known limitations (intentional)

• Parameter-only shadowing is not guarded. The shadow prune counts variable_declarators, so a file-scope const shadowed only by a function parameter of the same name would slip through. Not observed in S/M/L TS validation, and guarding it would over-prune legitimate consts whose name coincides with a parameter elsewhere in the file — so it's left unguarded until a real repo surfaces it.
• Same-file only. Cross-file value consumers (a const imported and read elsewhere) are not edged; that needs import/scope resolution and is out of scope.
• Reactive/computed reads (a value read only through a framework getter) have no static identifier to match and aren't covered.

Extending to another language

The step-by-step runbook — wiring checklist, validation scripts, FP hunts, per-language declarator types, and traps — is in value-reference-edges-playbook.md. Point a fresh session at it and say "Start on language X." In short: decide whether the language's constants are file/module-scope (fits) or class-scope (bigger change); confirm the declarator node type for the shadow prune; sweep small/medium/large public OSS repos; fix FP clusters; add a matrix row here + a test.