When target_projects = ["*"], scan the cache directory for all .db files. */

Gather every stub in the allowlist plus any re-export targets they pull in transitively (e.g. os.path -> posixpath -> genericpath). Returns (m

Phase 1: Walk AST iteratively and collect normalised LEAF token types. * Leaf-only counting is language-agnostic: leaf nodes correspond to actual *

Walk a field node and collect type identifier names into out[]. Handles: direct type_identifier/generic_type/qualified_name, type_list children (Java

Collect caller edges (HTTP_CALLS + ASYNC_CALLS) targeting a route. */

Collect direct-child decorator texts into result[] starting at idx.

Collect config Variable nodes with ≥2 tokens, each ≥3 chars. */

Collect bases from a single base_list node into an arena-allocated array. */

Scan cache dir for .db files, writing comma-separated quoted names into out. * Returns the number of projects found. */

Phase 1: Collect decorated nodes and count word frequency. */

Collect all Function/Method nodes with fingerprints from graph buffer. */

Collect extra handler IDs reachable via INFRA_MAPS chain from a route. */

Collect decorator texts from a modifiers node into result array starting at idx.

Project RETURN * — all bound variable properties */ Collect all variable names from query patterns */

Collect package names from nodes matching a SQL query (must use ?1 = project). */

TypeScript/TSX: bases live in a `class_heritage` (class) or directly in an * `extends_type_clause` (interface). The extractor previously captured th

* collect_usage_sources: index the fixture once, find the IRepository node, * walk all USAGE edges that target it, and write the source file_paths *

Collect a single namespace_use_clause's local→target mapping and add. * * tree-sitter-php emits `use Vendor\X as Y;` either as a use_clause with *

Count common dot-separated prefix segments. */

Compile a path filter regex. Returns true if compiled successfully. */

Resolve a CFML tag-function's QN for scope tracking. A <cffunction name="foo"> * is a `cf_function_tag`; the name lives in a `cf_attribute` child (na

Compute class QN for scope tracking.

Resolve an Elixir def/defp/defmacro's QN for scope tracking. The def is a * `call` node whose target (first child) is the def macro and whose first

Compute function QN for scope tracking (mirrors cbm_enclosing_func_qn logic).

Resolve a Go-template named-template's QN for scope tracking. A * {{ define "greeting" }} ... {{ end }} is a `define_action` whose name is a * quote

Resolve a Lisp (Clojure/Scheme/Racket) def-form's QN for scope tracking. * The def node is a list/list_lit whose head names the def kind and whose *

Try to resolve Wolfram function QN from set_delayed_top/set_top/set_delayed/set LHS.

── Structural-base battery (dims 1-4) ────────────────────────────────────── * * Runs the four core invariants on valid input. No defs-present asser

── Callable battery with calls-extracted (dims 1-6) ─────────────────────── * * Adds dims 5 (optional) and 6 (calls-extracted) to the base invariant

── Full-pipeline battery (dims 7-8) ─────────────────────────────────────── * * Indexes the single-file fixture through the production pipeline and

── Robustness helper: assert call RETURNS on malformed input ─────────────── * * A truncated version of the fixture is passed through cbm_extract_fi

── Structural battery with defs-present (dims 1-5) ──────────────────────── * * Adds the defs-present dimension for languages with class_types, func

Phase 2b: Connect prefix Routes to decorator handler Functions. */

Walk a type tree to detect any CBM_TYPE_INFER nodes.

RRI pass 2 accumulator for a single target token, reading int8-quantized * pass1 vectors as the source. */

Accumulate co-occurrence context for a single target token into `target`. * Reads neighbors within ±CBM_SEM_WINDOW positions across all documents tha

Copy a trimmed header value into out. */

Copy a node from an array into a heap-allocated standalone node. */

Count consecutive alphanumeric characters */

Count consecutive alphanumeric or dash characters */

Count resolved (confidence > 0) vs unresolved (confidence == 0) and total. */

Count calls whose callee_name is exactly `name` (has_call is substring). */

Count occurrences of a character in a string */

Count direct children of `node` that are decorator/annotation nodes (used by languages like Scala where the annotation is a direct child of the def no

───────────────────────────────────────────────────────────────────────── * INVARIANT 3: No dangling edges (graph integrity guard). * * For every e

Count how many defs in r have exactly this label AND name. */

Count definitions with a given label. */

* count_defs_with_label — count all defs carrying the given label. * Mirrors the helper in test_extraction.c. */

── Helper: count distinct qualified_names among nodes by name ─────────── */ Returns the number of DISTINCT qualified_name strings found. */ This catc

Count nodes carrying `label`. Returns -1 on query error. */

* count_method_defs_named — count defs with label "Method" matching name. * Mirrors the `has_def` helper in test_extraction.c but counts all matches.

Extract decorator names from preceding decorator/annotation nodes Count annotations inside a Java/Kotlin/C# "modifiers" node.

── Helper: count distinct nodes by name for a project ─────────────────── */

Count nodes with the given exact name in the project (e.g. a Route path). */

Count parameters from a signature string like "(int a, Foo* b, cb (*)(int,int))". Fallback for languages where param_names isn't populated (e.g. C kee

Count resolved calls matching pattern */

Count SIMILAR_TO edges in graph buffer. */

Count the number of '/' in a string. */

Sum callable nodes (Function + Method). */

Count edges of `edge_type`; -1 on DB error. */

Count CROSS_HTTP_CALLS rows in a project DB (re-opened from the cache dir, * because cbm_cross_repo_match wrote through its own store handles). */

Index a one-file client and a one-file server project through the production * pipeline. Closes both stores again — cbm_cross_repo_match re-opens the

Segment-wise match of a concrete path against a route template path, where a * "{...}" segment in the template matches any single non-empty concrete