SCHEMA.md

The output bundle — schema

Generated from graph/bundle/schema.ts (schema version 1). Do not edit; run npm run schema-doc.

Every run, in every language, writes the same thing:

<out>/
  graph.sqlite      the contract — the tables below, the ext_* tables, and this document as tables
and only with --debug:
  csv/<table>.csv   the core tables as headered, tab-delimited text (RFC 4180 quoting)
  raw/              the per-language Soufflé relations, verbatim. Engine-internal; not a contract.

The same schema in every language — a table a front end does not derive is empty, not missing, and a note below says so. Where the front ends differ (which values a column can hold, what an id may point at) the difference is written here and in the schema_vocab / schema_notes tables inside the database, so a query can read it without leaving SQLite:

SELECT value, meaning FROM schema_vocab WHERE table_name='call_edges' AND column_name='tier'
  AND language = (SELECT value FROM run WHERE key='language');

Identifiers are the parser's hashes and are opaque; join them to methods / types / call_sites for names and positions. NULL in SQLite is the empty field in CSV.

How to use it (schema_guide)

1. This is a call graph of one codebase, derived by a type-directed Datalog engine. Start with SELECT value FROM run WHERE key='language' — every language-specific fact below is keyed on it.
2. The graph is call_edges: one row per (call site, possible target). Rows join to methods (names, files, lines) on caller_id / callee_method_id, and to call_sites on call_site_id for where the call is written. Identifiers are opaque hashes — never parse them, always join.
3. Trust is explicit. tier says what kind of claim a row is: known_edge (one resolved target), multi_inferred (a sound set — every row of the set is a real possibility), boundary_lib (leaves the client; not expanded further), ambiguous_* (a declared unknown: callee is NULL). Pick the tiers your question tolerates and filter on them; never treat an ambiguous_* row as an edge.
4. For a FIELD rather than a callable, the graph is field_access: one row per (site, resolved field), joined to fields on field_id and to methods on caller_id, with access saying read / write / readwrite. It carries the same four tiers and the same promise as call_edges, so "who writes Foo.bar" is answered with a confidence, not with a name match. Java only so far; the table is present and empty elsewhere.
5. For a TYPE, the graph is type_use: one row per place the type is named, with the context it was written in (FIELD_TYPE, METHOD_PARAM, METHOD_RETURN, OBJECT_CREATION_TYPE, CAST_EXPRESSION, SUPER_TYPE and the rest) and the depth that separates the type as written from its type arguments. That is what makes "what breaks if I change T" specific per construct rather than a count of mentions. Java only so far.
6. call_edges is what the engine CONCLUDED; dispatch_candidates is what the hierarchy ADMITTED. Read the second when you need an upper bound rather than a best answer — a candidate whose owner is absent from type_instantiated is admitted by the hierarchy but never constructed in this run, which is how you narrow it yourself. basis separates a declared relationship from a shape match.
7. Before answering "nothing calls X" or "X cannot reach Y", check unresolved_sites for the methods on the path: a caller listed there has a call the engine could not resolve, so the answer is a lower bound and should say so.
8. Library targets (callee_provenance = lib) are named in methods with provenance = lib but their bodies were not analysed; a Python builtin/external target has no methods row and lives in callee_label.
9. schema_vocab lists every value a column can hold FOR THIS LANGUAGE with its meaning — filter on language = (SELECT value FROM run WHERE key='language'). schema_notes lists the caveats for this language (empty tables, what an id may point at). Read both before interpreting kind, tier or an empty table.
10. schema_queries holds tested SQL for the common questions (callers, callees, blast radius, entry reachability, the method at a file:line, the blind spots of a method, the dispatch envelope of a method). Bind the named parameters and run.
11. Tables named ext_<relation> are the language's raw engine relations with positional columns c0…cN; schema_tables carries each one's description lifted from its rule. Use them only when a core table does not hold what you need.
12. When you report a result, carry the tier and the unresolved count with it. A consumer who cannot see the confidence of an edge cannot use it.

Canonical queries (schema_queries)

Each is verified to run against every language's bundle. Bind the named parameters.

skipped_files — Which files is this graph missing, and why? (Run this before reading any absence as an answer.) ((none))

SELECT k.file_path, k.reason, k.construct, k.start_line, k.detail,
       (SELECT count(*) FROM methods m WHERE m.file_path = k.file_path) AS methods_in_graph
FROM skipped k
ORDER BY k.reason, k.file_path

callers_of — Who calls this method, from where, and how sure is each edge? (:qualified_name)

SELECT caller.qualified_name AS caller, caller.file_path, s.start_line, e.tier, e.kind
FROM call_edges e
JOIN methods callee ON callee.id = e.callee_method_id
JOIN methods caller ON caller.id = e.caller_id
LEFT JOIN call_sites s ON s.id = e.call_site_id
WHERE callee.qualified_name = :qualified_name
ORDER BY caller.file_path, s.start_line

callees_of — What does this method call — resolved targets, library boundaries, and the sites it could not resolve? (:qualified_name)

SELECT s.start_line, s.callee_name AS written, e.tier, e.callee_provenance,
       COALESCE(t.qualified_name, e.callee_label) AS target
FROM call_edges e
JOIN methods caller ON caller.id = e.caller_id
LEFT JOIN methods t ON t.id = e.callee_method_id
LEFT JOIN call_sites s ON s.id = e.call_site_id
WHERE caller.qualified_name = :qualified_name
ORDER BY s.start_line, target

blast_radius — If this method changes, which methods are transitively affected, up to :depth hops, through RESOLVED client edges only (a declared unknown is not traversed, and the count of them is returned alongside)? (:qualified_name, :depth)

WITH RECURSIVE up(id, depth) AS (
  SELECT id, 0 FROM methods WHERE qualified_name = :qualified_name
  UNION
  SELECT e.caller_id, up.depth + 1
  FROM call_edges e JOIN up ON e.callee_method_id = up.id
  WHERE e.tier IN ('known_edge', 'multi_inferred', 'ambient_terminal', 'intrinsic_terminal')
    AND up.depth < :depth
)
SELECT MIN(up.depth) AS depth, m.qualified_name, m.file_path, m.start_line,
       (SELECT count(*) FROM unresolved_sites u WHERE u.caller_id = m.id) AS unresolved_calls_inside
FROM up JOIN methods m ON m.id = up.id
WHERE up.depth > 0
GROUP BY m.id ORDER BY depth, m.qualified_name

reachable_from_entries — Is this method reachable from any entry point (a main, a test, an HTTP handler, an unimported module)? (:qualified_name)

SELECT m.qualified_name,
       EXISTS (SELECT 1 FROM entry_reachable r WHERE r.method_id = m.id) AS reachable,
       (SELECT count(*) FROM entry_points) AS entry_points_known
FROM methods m WHERE m.qualified_name = :qualified_name

method_at — Which method contains this file:line? (:file_path, :line)

SELECT qualified_name, kind, start_line, end_line
FROM methods
WHERE file_path = :file_path AND start_line <= :line AND end_line >= :line
ORDER BY (end_line - start_line) LIMIT 1

blind_spots_of — Which calls inside this method could the engine not resolve — the caveat to attach to any answer about it? (:qualified_name)

SELECT s.start_line, s.callee_name AS written, s.kind
FROM unresolved_sites u
JOIN methods m ON m.id = u.caller_id
LEFT JOIN call_sites s ON s.id = u.call_site_id
WHERE m.qualified_name = :qualified_name
ORDER BY s.start_line

dispatch_envelope_of — What else might actually run at a call that resolves to this method — the set the graph narrowed from, and whether each candidate is a declaration or a shape match? (:qualified_name)

SELECT cand.qualified_name AS candidate, cand.file_path, cand.start_line, d.basis,
       EXISTS (SELECT 1 FROM type_instantiated i WHERE i.type_id = cand.owner_type_id) AS owner_instantiated
FROM dispatch_candidates d
JOIN methods base ON base.id = d.base_method_id
JOIN methods cand ON cand.id = d.candidate_method_id
WHERE base.qualified_name = :qualified_name
ORDER BY d.basis, cand.qualified_name

subtypes_of — Which types extend or implement this type (transitively)? (:qualified_name)

SELECT sub.qualified_name, sub.category, sub.file_path
FROM type_ancestors a
JOIN types anc ON anc.id = a.ancestor_type_id
JOIN types sub ON sub.id = a.type_id
WHERE anc.qualified_name = :qualified_name
ORDER BY sub.qualified_name

values_of — What can this column hold in THIS bundle's language, and what does each value mean? (:table_name, :column_name)

SELECT value, meaning FROM schema_vocab
WHERE table_name = :table_name AND column_name = :column_name AND language = (SELECT value FROM run WHERE key='language')
ORDER BY value

field_impact — Who reads or writes this field — and which of those answers are certain? (:owner_qualified_name, :field_name)

SELECT m.qualified_name AS accessor, fa.access, fa.tier, fa.file_path, fa.start_line
FROM field_access fa
JOIN fields f ON f.id = fa.field_id
LEFT JOIN methods m ON m.id = fa.caller_id
WHERE f.owner_qualified_name = :owner_qualified_name AND f.name = :field_name
ORDER BY fa.tier, m.qualified_name, fa.start_line

field_blind_spots — Which field accesses could the engine not resolve — the caveat to attach to any answer about a field?

SELECT m.qualified_name AS accessor, fa.access, fa.file_path, fa.start_line
FROM field_access fa
LEFT JOIN methods m ON m.id = fa.caller_id
WHERE fa.tier = 'ambiguous_unknown'
ORDER BY fa.file_path, fa.start_line

type_impact — Where is this type used, and in what construct — the answer to "what breaks if I change it"? (:qualified_name)

SELECT u.context, u.depth, u.tier,
       coalesce(m.qualified_name, ot.qualified_name) AS used_in,
       coalesce(m.file_path, ot.file_path) AS file_path
FROM type_use u
JOIN types t ON t.id = u.type_id
LEFT JOIN methods m ON m.id = u.owner_method_id
LEFT JOIN types ot ON ot.id = u.owner_type_id
WHERE t.qualified_name = :qualified_name
ORDER BY u.context, used_in

tier_summary — How much of this graph is certain, inferred, at a library boundary, or unknown?

SELECT tier, count(*) AS edges, count(DISTINCT call_site_id) AS sites
FROM call_edges GROUP BY tier ORDER BY edges DESC

Core tables

run

What produced this bundle: one key/value row per fact about the run (language, engine commit, inputs, knobs, timestamps, schema version). Read language first — it selects which vocabulary rows apply.

#	column	type	key	null	idx	meaning
0	key	TEXT	yes			Fact name — see the run.key vocabulary.
1	value	TEXT				Fact value, as text.

run.key values

value	languages	meaning
schema_version	all	Version of this contract (SCHEMA.md).
language	all	Front end: java | typescript | python | javascript | csharp. Selects the applicable vocabulary rows.
engine_commit	all	Git commit of the rule set that produced the graph, when known.
client_ir	all	Path of the client IR directory the engine read.
library_roots	all	Comma-separated library IR roots staged as the type oracle; empty for a client-only run.
dispatch_cap	all	Fan-width cap on virtual dispatch in effect; off when uncapped.
jdk_depth	all	Platform-library hop cap (engine-ii).
lib_depth	all	External-library hop cap; uncapped when unset.
engine_ii	all	on when the library-frontier forward chain (engine-ii) was included; off for a client-only solve.
solve_iterations	all	Stage↔solve rounds until the library frontier converged.
solve_seconds	all	Wall-clock seconds of staging + solving, before the bundle stage.
created_at	all	ISO-8601 timestamp of the bundle.
raw_dir	all	Where the per-language Soufflé relations were read from (raw/ next to the bundle).
source_version	all	The version the IR was stamped with (bin/axiomcode): the git commit of the analysed source, or v1.0.0 when it was not a checkout. Present when the run went through bin/axiomcode all.
source_dir	all	The source directory that was parsed. Present when the run went through bin/axiomcode all.

methods

Every callable the graph refers to: all client methods/functions from the IR, plus every LIBRARY method some edge reaches (library methods nothing reaches are not listed — a library IR is GB-scale). A module-level function is a method whose owner columns are NULL; top-level code is the module initializer method (kind MODULE_INITIALIZER in TypeScript and Python, absent in Java).

#	column	type	key	null	idx	meaning
0	id	TEXT	yes			The parser's unique hash for the method (METHOD_REGISTRY_… / TS_METHOD_… / PY_METHOD_…). The value every other table uses to refer to a method.
1	name	TEXT			yes	Simple name as written (render, __init__). A constructor is named after its class in Java (Circle) and <constructor> in TypeScript, JavaScript and C#.
2	qualified_name	TEXT			yes	Parser-qualified name — package/module path plus owner plus name. Unique only together with the signature.
3	signature	TEXT				Parameter-type signature as the parser prints it, e.g. main(String[]); language-native formatting.
4	kind	TEXT				The parser's methodKind — see vocabulary; the sets differ per language.
5	owner_type_id	TEXT		yes	yes	FK → types.id of the declaring class/interface/enum; NULL for a free function or a module initializer.
6	owner_qualified_name	TEXT		yes		Qualified name of the owner, denormalised so a row prints without a join; NULL when owner_type_id is NULL.
7	file_path	TEXT			yes	Source file, as the parser recorded it (relative to the project root it was given).
8	start_line	INTEGER				1-based first line of the declaration.
9	end_line	INTEGER				1-based last line of the declaration.
10	provenance	TEXT				client — from the analysed project; lib — from a staged library IR; generated — declared by an annotation processor and synthesised here (Java). See vocabulary.
11	visibility	TEXT		yes	yes	Declared access as the parser recorded it, normalised (PUBLIC / PROTECTED / PACKAGE / PRIVATE). NULL where the language has no access modifiers. A public or protected declaration with no dependent in this graph is exported surface, not dead code.

methods.kind values

value	languages	meaning
METHOD	csharp	An ordinary method.
CONSTRUCTOR	csharp	A constructor, written as one. The parser names it <constructor>, while a new Foo(...) site carries Foo.
PRIMARY_CONSTRUCTOR	csharp	A constructor declared in the type header: class Svc(IDep d), and every positional record. It is the only constructor such a type has.
STATIC_CONSTRUCTOR	csharp	The type initializer. Static field initializers run in it; the compiler emits one only where it is needed, so it is often absent.
DESTRUCTOR	csharp	A finalizer. Called by the garbage collector, so it has no caller in the graph.
OPERATOR	csharp	A user-defined operator. a + b on such a type is a static call to it, with no call syntax at the site.
CONVERSION_OPERATOR	csharp	A user-defined conversion. An implicit one runs with no syntax at the call site at all.
LOCAL_FUNCTION	csharp	A function declared inside a method body. Visible only there, and it may capture the enclosing locals.
LAMBDA	csharp	A lambda body, which is its own method. Calls written in it are attributed to it, not to the method containing the lambda: the runtime reaches it through a delegate.
ANONYMOUS_METHOD	csharp	A delegate { ... } body. Same shape as LAMBDA.
TOP_LEVEL_ENTRY_POINT	csharp	The synthetic method holding a file of C# 9 top-level statements. An entry point, and the caller every top-level call site is attributed to.
PROPERTY_GET	csharp	A property getter. x.Name is a call to it. An auto-property getter has no body, which is the correct answer rather than a gap.
PROPERTY_SET	csharp	A property setter. x.Name = v is a call to it.
PROPERTY_INIT	csharp	An init accessor: settable only in an object initializer or a constructor.
INDEXER_GET	csharp	An indexer getter. a[i] on the read side is a call to it.
INDEXER_SET	csharp	An indexer setter. a[i] = v.
INDEXER_INIT	csharp	An indexer init accessor.
EVENT_ADD	csharp	An event add accessor. e += h is a call to it.
EVENT_REMOVE	csharp	An event remove accessor. e -= h.
INSTANCE_METHOD	java	Non-static method.
STATIC_METHOD	java	Static method.
ABSTRACT_METHOD	java	Abstract or interface method without a body.
DEFAULT_METHOD	java	Interface default method.
CONSTRUCTOR	java, typescript	A declared constructor. Java names it after the class (Circle, qualified shapes.Circle.Circle); TypeScript names it <constructor>.
STATIC_INITIALIZER	java	static { … } block.
GENERATED_METHOD	java, csharp	Synthesised for a member no IR declares, so it has no real methodKind: in Java one an annotation processor declares, in C# the Invoke of the delegate a field or property holds. Always paired with provenance generated.
ENUM_CONSTANT_METHOD	java	Method body declared on an enum constant.
RECORD_ACCESSOR	java	A record component accessor.
COMPACT_CONSTRUCTOR	java	A record's compact canonical constructor.
DEFAULT_CONSTRUCTOR	java, typescript	The implicit no-arg constructor the parser synthesises for a class that declares none (TypeScript: and extends nothing; a subclass runs the nearest declared base constructor). Named as a declared one is.
INSTANCE_INITIALIZER	java	{ … } instance initializer block.
ANNOTATION_ELEMENT	java	An element of an annotation interface.
RECORD_EQUALS	java	A record's implicit equals.
RECORD_HASH_CODE	java	A record's implicit hashCode.
RECORD_TO_STRING	java	A record's implicit toString.
ENUM_VALUES	java	An enum's implicit values().
ENUM_VALUE_OF	java	An enum's implicit valueOf(String).
FUNCTION_DECLARATION	typescript	function f() {}.
METHOD_DECLARATION	typescript	Class or interface method.
GETTER	typescript	get x().
SETTER	typescript	set x(v).
ARROW_FUNCTION	typescript	Arrow function value.
FUNCTION_EXPRESSION	typescript	function () {} value.
METHOD_SIGNATURE	typescript	Bodiless method in an interface/type.
CALL_SIGNATURE	typescript	Interface call signature (x): y.
CONSTRUCT_SIGNATURE	typescript	Interface construct signature new (x): y.
TYPE_LITERAL_METHOD_SIGNATURE	typescript	Method signature inside a type literal.
TYPE_LITERAL_CALL_SIGNATURE	typescript	Call signature inside a type literal.
TYPE_LITERAL_CONSTRUCT_SIGNATURE	typescript	Construct signature inside a type literal.
FUNCTION_TYPE_SIGNATURE	typescript	A function type (x) => y.
CONSTRUCTOR_TYPE_SIGNATURE	typescript	A constructor type new (x) => y.
OBJECT_LITERAL_METHOD	typescript	Method in an object literal.
CLASS_STATIC_BLOCK	typescript	static { … } in a class.
MODULE_INITIALIZER	typescript, python	Synthetic method holding a module's top-level code. Every module has one; top-level call sites belong to it.
FUNCTION	python	Module-level def.
INSTANCE_METHOD	python	Method taking self.
STATIC_METHOD	python	@staticmethod.
CLASS_METHOD	python	@classmethod.
PROPERTY_GETTER	python	@property.
PROPERTY_SETTER	python	@x.setter.
PROPERTY_DELETER	python	@x.deleter.
CONSTRUCTOR	python	__init__.
ALLOCATOR	python	__new__.
DUNDER_METHOD	python	Other __x__ method.
ABSTRACT_METHOD	python	@abstractmethod.
OVERLOAD_STUB	python	@overload signature.
LAMBDA	python	A lambda expression.
NESTED_FUNCTION	python	def inside a function.
GENERATOR	python	Function with yield.
ASYNC_FUNCTION	python	async def.
ASYNC_GENERATOR	python	async def with yield.
CLASS_INITIALIZER	python	Synthetic method holding a class body's top-level code.
FUNCTION_DECLARATION	javascript	function f() {} — hoisted.
FUNCTION_EXPRESSION	javascript	function () {} value, including an object literal's m() {}.
ARROW	javascript	Arrow function value; this is lexical.
CLASS_METHOD	javascript	A class member, syntactic or declared by assignment (F.prototype.m = …, F.s = …).
CONSTRUCTOR	javascript	constructor() of a class, named <constructor>. A pre-ES6 constructor function is not a CONSTRUCTOR: its row keeps the kind and name of its syntax (function F() {} is FUNCTION_DECLARATION F; var F = function () {} and exports.F = function () {} are FUNCTION_EXPRESSION <function-expression>, or the expression's own name if it has one), and the type it introduces, F, is CONSTRUCTOR_FUNCTION.
GETTER	javascript	get x().
SETTER	javascript	set x(v).
STATIC_BLOCK	javascript	static {} block of a class.
MODULE_INITIALIZER	javascript	Synthetic method holding a module's top-level code. Every module has one; top-level call sites belong to it.

methods.visibility values

value	languages	meaning
PUBLIC	java	Callable from anywhere; exported surface. No in-repo caller does not mean unused.
PROTECTED	java	Callable by subclasses outside the package; exported surface for extension.
PACKAGE	java	Default access; callable only within the declaring package.
PRIVATE	java	Callable only within the declaring type.

methods.provenance values

value	languages	meaning
client	all	Declared in the analysed project.
lib	all	Declared in a staged library IR; listed because an edge reaches it.
generated	java	Declared by a compile-time annotation processor: present in the compiled artefact and in every caller's source, and in no IR, so the bundle synthesises it to give the edge a target. id generated:<owner qualified name>#<name>/<arity>, no file and no line numbers.
generated	csharp	The Invoke of the delegate held in one field or property: id generated:<owner qualified name>#<member>.Invoke, no file and no line numbers. Every call through that member is an edge to it (tier boundary_lib, which is also where the compiler binds the call), and what the member holds is in dispatch_candidates with basis value.

Notes

• all — Library rows are the subset an edge reaches. To see a library method nothing calls, query the library IR itself.
• javascript — signature is empty and owner_qualified_name is NULL: JavaScript declares neither. owner_type_id is set for class members, including members declared by assignment.
• javascript — A library row's qualified_name and file_path are prefixed with the package: its path under the client when installed there (node_modules/<pkg>/…, nested versions included), else its package name (<pkg>/…; a second root with the same name gets #2). The parser records both relative to the package root, where two packages with an index.js are indistinguishable. Same for types.

types

Every class-like declaration the graph refers to: all client types, plus every library type that owns a listed library method or appears in type_ancestors.

#	column	type	key	null	idx	meaning
0	id	TEXT	yes			The parser's unique hash (TYPE_REGISTRY_… / TS_TYPE_… / PY_TYPE_…).
1	name	TEXT			yes	Simple name.
2	qualified_name	TEXT			yes	Parser-qualified name.
3	category	TEXT				The parser's typeCategory — see vocabulary.
4	file_path	TEXT		yes	yes	Source file; NULL for an external type (no declaration was staged).
5	start_line	INTEGER		yes		1-based first line; NULL for an external type.
6	end_line	INTEGER		yes		1-based last line; NULL for an external type.
7	provenance	TEXT				client, lib, or external (Java: an unstaged ancestor, see vocabulary).
8	visibility	TEXT		yes	yes	Declared access as the parser recorded it, normalised (PUBLIC / PROTECTED / PACKAGE / PRIVATE). NULL where the language has no access modifiers.

types.category values

value	languages	meaning
CLASS	csharp	A class.
INTERFACE	csharp	An interface. Its members may have bodies (C# 8 default implementations), so an interface method is not always an abstract stub.
STRUCT	csharp	A struct. It cannot be derived from, so a receiver of this type is exact and needs no dispatch fan.
ENUM	csharp	An enum. Cannot be derived from.
DELEGATE	csharp	A delegate type. It names a signature; a call through a value of it is resolved by value flow rather than by member lookup.
RECORD	csharp	A record declared with the record keyword. Also a CLASS or STRUCT underneath; the row says which via its modifiers.
CLASS_TYPE	all	A class.
EXTERNAL_TYPE	java	An unstaged ancestor named by the client — see provenance external. Class or interface is not known.
INTERFACE_TYPE	java, typescript	An interface.
ENUM_TYPE	java, typescript	An enum.
RECORD_TYPE	java	A record.
ANNOTATION_TYPE	java	An annotation interface.
ANNOTATION_INTERFACE_TYPE	java	An annotation interface (@interface), as the parser categorises it in newer output.
CONST_ENUM_TYPE	typescript	const enum.
TYPE_ALIAS_TYPE	typescript	type X = ….
NAMESPACE_TYPE	typescript	namespace X {}.
CLASS_EXPRESSION_TYPE	typescript	A class expression value.
CLASS	javascript	An ES class declaration.
ANONYMOUS_CLASS	javascript	A class expression.
CONSTRUCTOR_FUNCTION	javascript	A function with prototype members — a pre-ES6 class.
JSDOC_TYPEDEF	javascript	A @typedef — comment-only, never constructed or dispatched into.
JSDOC_CALLBACK	javascript	A @callback — comment-only.
EXCEPTION_CLASS_TYPE	python	A class deriving from BaseException.
ENUM_CLASS_TYPE	python	An Enum subclass.
PROTOCOL_TYPE	python	A typing.Protocol.
ABC_TYPE	python	An abstract base class.
NAMEDTUPLE_TYPE	python	A NamedTuple class.
TYPEDDICT_TYPE	python	A TypedDict class.
DATACLASS_TYPE	python	A @dataclass.
METACLASS_TYPE	python	A metaclass (derives from type).
GENERIC_TYPE	python	A Generic[…] class.

types.provenance values

value	languages	meaning
client	all	Declared in the analysed project.
lib	all	Declared in a staged library IR.
external	java	Named by the client as an ancestor (extends/implements) but declared in no staged IR: id external:<qualified name>, category EXTERNAL_TYPE, no file, no members. Kept so the subtype edge survives; stage the library to replace it with the real declaration.

call_sites

One row per place a call is written (or, for a synthesised edge, the construct that implies the call). Every call_edges.call_site_id is here. Location columns come from the IR; they are NULL when the site is a construct the IR does not position (see notes).

#	column	type	key	null	idx	meaning
0	id	TEXT	yes			Site identifier — an expression hash in every language; in Python it may also be a decorator hash or, for METACLASS_CREATION, the class's type hash (see notes).
1	caller_id	TEXT			yes	The method whose body contains the site (FK → methods.id) — or, for code that runs outside any method, the enclosing TYPE (Java: TYPE_REGISTRY_… = the type's static/instance initializer) or MODULE (TypeScript: TS_MODULE_… as a fallback marker). Never NULL. See notes.
2	kind	TEXT				What syntactic form the call takes — the same value as call_edges.kind for this site; see vocabulary (language-specific sets).
3	callee_name	TEXT		yes	yes	The name written at the site (render in w.render(), the class name in new Widget()); NULL when the form has no written name (a constructor delegation, a record deconstruction).
4	file_path	TEXT		yes	yes	Source file.
5	start_line	INTEGER		yes		1-based line.
6	start_column	INTEGER		yes		Column, as the parser counts it.
7	end_line	INTEGER		yes		1-based last line; NULL where the IR records only the start.
8	end_column	INTEGER		yes		End column; NULL where the IR records only the start.

Notes

• java — caller_id is a TYPE_REGISTRY_ id (a types row, not a methods row) for a call written in a field initializer or a static/instance initializer block: the parser gives such code no enclosing method, and the rule set attributes it to the type — read it as "runs in this type's /".
• typescript — caller_id is normally the parser's caller method, or the module initializer for top-level code; when neither exists it is the TS_MODULE_ hash itself, kept as a greppable marker rather than a blank.
• java — callee_name for new X() and for new X() { … } (anon_new) is the class name written at the site; NULL for ctor_delegate (this(…)/super(…)) and record_accessor, which write no name. A by-name lookup must therefore exclude kind IN (new, anon_new) to avoid counting a construction as a call to a same-named method.
• java — A record_accessor site is the RECORD_PATTERN expression, positioned where the pattern is written.
• java — A resource_close site is the resource LOCAL of a try-with-resources, positioned at its declaration (line only, no column); callee_name is NULL.
• python — A DECORATOR_APPLICATION edge targets the callable the decorator factory RETURNS, not the name written at the @ — @deco(X) applies the inner callable that deco returned. The written name is carried by the separate DECORATOR_CALL row, so a by-name lookup must exclude DECORATOR_APPLICATION or it will read the wrapper as a mismatch.
• typescript — end_line / end_column come from the expression row; the call-site row itself records only the start.
• javascript — caller_id is the parser's enclosing method, or the module initializer for top-level code. end_line / end_column come from the expression row. require() is a module edge, not a call site.
• typescript — PROPERTY_READ and PROPERTY_WRITE rows are accessor invocations with no written call: the site is the property-access expression that runs the getter or setter, positioned from the expressions table, and callee_name is NULL because nothing was written; the accessor's name is on the callee's methods row. Filter them out with kind NOT IN (…) when counting calls.
• python — PROPERTY_READ, CONTEXT_MANAGER, ITERATION_PROTOCOL, METACLASS_CREATION and DYNAMIC_CALL rows are protocol or indirect edges with no written call: their site is the expression that triggers them, and callee_name is always NULL because nothing was written. SUBSCRIPT_CALL is NULL only when the subscript is not a written name (measured 206 of 337 rows on a Python subject). Filter them out with kind NOT IN (…) when counting calls.
• python — The id is an EXPRESSION hash for a written call; a DECORATOR hash (PY_DECORATOR_…) for DECORATOR_APPLICATION and DECORATOR_* sites, positioned at the decorator line; and the class's TYPE hash for METACLASS_CREATION, positioned at the class declaration.

call_edges

THE GRAPH. One row per (site, resolved target). A site with N possible targets has N rows, each carrying the same tier; a site the engine could not resolve has exactly one row with a NULL callee and an ambiguous_* tier — so every call site written in the client appears at least once, and the table alone shows where every chain ends and why. Filter on tier to choose your risk tolerance.

#	column	type	key	null	idx	meaning
0	call_site_id	TEXT			yes	FK → call_sites.id.
1	caller_id	TEXT			yes	Same value as call_sites.caller_id for this site: the containing method, or the enclosing type/module id when there is none (see call_sites).
2	callee_method_id	TEXT		yes	yes	FK → methods.id of the resolved target, when the target is a method the bundle knows (callee_provenance client or lib). NULL otherwise.
3	callee_label	TEXT		yes		The target when it is NOT a method row: a builtin (builtin:len) or an import path outside every staged IR (requests.get) — Python only today. NULL when callee_method_id is set or the site is unresolved.
4	callee_provenance	TEXT		yes		Where the target lives — see vocabulary. NULL for an unresolved site.
5	tier	TEXT			yes	Confidence class of this edge — see vocabulary. known_edge and multi_inferred are assertions about client code; boundary_lib leaves the client; the ambiguous_* tiers are declared blind spots, not edges.
6	kind	TEXT				Syntactic form of the site — see vocabulary; language-specific sets, kept native.

call_edges.kind values

value	languages	meaning
method	csharp	A call written with or without a receiver: a.M(), M(), a?.M().
new	csharp	An object creation. The target is the constructed type's constructor, and it is never dispatched.
ctor_delegate	csharp	: this(...) or : base(...). No name is written, so the target is structural.
primary_ctor_base	csharp	SYNTHESISED. A primary constructor's base invocation, written in the heritage clause: class D(int a) : B(a). There is no call syntax anywhere in the body. FromExpr is the heritage type reference.
delegate	csharp	A call through a delegate value: handler(x) or handler.Invoke(x).
operator	csharp	A user-defined operator invoked by operator syntax.
conversion	csharp	A user-defined conversion. An implicit one has no syntax at the call site.
indexer	csharp	A user-defined indexer accessor, invoked by a[i].
dynamic	csharp	A call through a dynamic value. Dispatched at runtime by the DLR, so the target is undecidable from source; the tier is ambiguous_dynamic and that is final, not a gap.
property_read	csharp	Reading a property runs its get accessor. No written call; the site is the member access.
property_write	csharp	Assigning a property runs its set (or init) accessor.
event_subscribe	csharp	e += h / e -= h on an event with custom add / remove accessors runs that accessor.
runtime	csharp	The edge of a runtime_observed row: no written call behind it.
other	csharp	A call kind this engine has no rule for. Present so a kind added to the schema later is unresolved rather than invisible.
method	java	obj.m(), Class.m(), super.m(), or an unqualified m().
new	java	new X(…).
ref	java	A method reference X::m, obj::m, X::new.
lambda_body	java	An edge from the method that INVOKES a lambda to something the lambda body calls. The body is not a method of its own, so its calls are attributed to the invoker rather than lost (call-edge-generation/lambda_dispatch.dl).
ctor_delegate	java	this(…) / super(…) inside a constructor.
anon_new	java	new X() { … } — an anonymous class creation.
event	java	Synthesised: the publishEvent call site to a listener it runs (tier event_dispatch). Not a written call to that method.
entity_callback	java	Synthesised: a call that writes an entity (repository.save(e), em.persist(e), delete(e)) to a JPA lifecycle callback it runs (tier event_dispatch). Not a written call to that method.
record_accessor	java	Synthesised: a record pattern case Pair(var l, var r) calls each accessor. Not a written call; the site is the pattern expression.
resource_close	java	Synthesised: a try-with-resources resource try (var w = open()) is closed when the block exits, so its close() is called. Not a written call; the site is the resource local.
FUNCTION_CALL	typescript	f(…) — a bare callee.
METHOD_CALL	typescript	obj.m(…).
CONSTRUCTOR_CALL	typescript	new X(…).
SUPER_CALL	typescript	super(…) or super.m(…).
TAGGED_TEMPLATE_CALL	typescript	tag….
INDEX_CALL	typescript	obj[k](…) through an index signature.
DYNAMIC_IMPORT_CALL	typescript	import(…).
DECORATOR_CALL	typescript	A decorator application @d / @d(…).
OPTIONAL_CALL	typescript	f?.(…).
JSX_COMPONENT_CALL	typescript	<Component …/> — a JSX element naming a component; <div> emits no site.
PROPERTY_READ	typescript, javascript	Reading obj.x where x is a get accessor runs the getter (engine-authored). No written call; the site is the property-access expression. A compound assignment or ++ reads before it writes, so it carries this and PROPERTY_WRITE.
PROPERTY_WRITE	typescript, javascript	Assigning obj.x = v where x is a set accessor runs the setter (engine-authored). No written call; the site is the property-access expression on the left.
FUNCTION_CALL	javascript	f(…) — a bare callee, resolved by the binder.
METHOD_CALL	javascript	obj.m(…).
CONSTRUCTOR_CALL	javascript	new X(…).
SUPER_CALL	javascript	super(…).
COMPUTED_CALL	javascript	obj[expr](…) — the name is not fixed by syntax.
FUNCTION_CALL_CALL	javascript	f.call(o, …) — the target is f; the receiver moved into argument position.
FUNCTION_CALL_APPLY	javascript	f.apply(o, args) — the target is f.
FUNCTION_CALL_BIND	javascript	f.bind(o) — produces a function that runs f; the edge names f.
IIFE_CALL	javascript	(function () {…})().
OPTIONAL_CALL	javascript	obj?.m(…).
TAGGED_TEMPLATE_CALL	javascript	tag….
DYNAMIC_CODE_CALL	javascript	eval(…) / new Function(…) — unknowable by construction.
DYNAMIC_IMPORT_CALL	javascript	import(…) — a module load that is also a site.
JSX_ELEMENT	javascript	<Component …/> — the renderer runs the component (a function component, or a class component's constructor and render) with the element's attributes as its props. Engine-authored: no written call; the site is the JSX element expression. An intrinsic tag (<div/>) is no site.
JSX_ATTRIBUTE	javascript	<form onSubmit={submit}>, <UserForm action={create}/> — a function handed over as a JSX attribute, on any tag; always tier callback_registered. Engine-authored: the site is the attribute.
SIMPLE_CALL	python	f(…) — a bare name.
METHOD_CALL	python	obj.m(…).
CHAINED_CALL	python	a.b().c(…) — the receiver is itself a call.
SUPER_CALL	python	super().m(…).
SELF_CALL	python	self.m(…).
CLS_CALL	python	cls.m(…).
MODULE_CALL	python	module.f(…) on an imported module.
SUBSCRIPT_CALL	python	d[k](…).
DYNAMIC_CALL	python	Callee computed at runtime (getattr(...)() and the like).
DECORATOR_CALL	python	The factory call of a parenthesised decorator @d(…).
INSTANCE_CALL	python	Calling an instance — dispatches to __call__.
BUILTIN_CALL	python	The parser recognised a builtin (len, print, …).
UNKNOWN_CALLEE_CALL	python	The parser could not classify the callee.
DECORATOR_APPLICATION	python	Applying a parenthesised decorator's RESULT to the decorated definition. The site is the decorator hash.
DECORATOR_*	python	Applying an unparenthesised decorator; the suffix is the parser's decorator kind: BARE, ATTRIBUTE, SUBSCRIPT, EXPRESSION (and CALL/ATTRIBUTE_CALL when the factory expression is not itself a call site). The site is the decorator hash.
METACLASS_CREATION	python	A class statement invokes its metaclass's __new__ / __init__ at import time, whether the metaclass is written on the statement (class X(metaclass=M)) or inherited from a base, and the nearest base's __init_subclass__. No written call; the site is the class's type hash.
PROPERTY_READ	python	Reading obj.attr where attr is a @property runs the getter; reading Cls.attr where the METACLASS defines attr as a property runs that getter. No written call; the site is the attribute-access expression.
CONTEXT_MANAGER	python	with expr: runs __enter__ / __exit__ (or the async pair). No written call; the site is the context-manager expression.
ITERATION_PROTOCOL	python	for x in expr: (and comprehensions) runs __iter__ / __next__ (or the async pair). No written call; the site is the iterated expression.
SUBSCRIPT_PROTOCOL	python	x[k] runs __getitem__ (and x[k] = v / del x[k] the setter and deleter) of the receiver's class. No written call; the site is the subscript expression. Its own kind so it is never counted as a written call.

call_edges.tier values

value	languages	meaning
ambiguous_dynamic	csharp	A call through dynamic. Unresolvable BY DESIGN rather than by omission, and separated from ambiguous_unknown so a known-undecidable site is not counted as an engine failure.
known_builtin_operator	csharp	An operator or conversion with no user-defined declaration anywhere (a + b on built-in types). No user code runs, so no target is the right answer; callee is NULL. An operator on an unstaged type IS a library member and is boundary_lib instead.
boundary_generated	csharp	A read of a property the compiler generated (a positional record's property). There is no written accessor to point at, so callee_method_id is NULL, callee_label names the member and callee_provenance is generated. A correct end, not a blind spot.
runtime_observed	csharp	Only when a runtime trace is supplied: an edge the trace saw and the static pass did not name. Real by construction, and added only, never used to remove an edge. The site is the CALLER METHOD, not an expression (the tracer sees method entry, not the call site), and kind is runtime.
known_implicit_ctor	csharp	new Foo() where Foo declares no constructor. The compiler supplies a parameterless one, so there is no user code to call and nothing resolving is the right answer.
event_dispatch	csharp	A mediator's Send(request) reaching the Handle of the handler registered for the request's type (IRequestHandler<TRequest, TResponse>, with a generic request's type arguments), or Publish(notification) reaching every INotificationHandler<T> for its type and each of its bases. Not the site's own callee: the site also keeps its row for the library method. The container picks the handler in process, before the call returns, so this is a call edge and not a remote_edge. A request declared as a base (not an interface) fans to the handler of each type in its family.
known_edge	all	Exactly one target resolved. The strongest claim.
multi_inferred	all	A sound SET of possible targets; each member is one row. The set over-approximates — every member is a real possibility, but not every member runs. HOW WIDE the set is differs by language: see the per-language notes on this table for whether the fan is narrowed by the instantiation set.
boundary_lib	all	The target is outside the client (library, builtin, or unstaged external). The chain is not expanded past it here.
ambiguous_unknown	all	Declared blind spot: the engine could not resolve the site (unresolved receiver, missing type, reflection…). callee is NULL. Never dropped.
ambiguous_anon	java	Known structural gap: an anonymous-class creation has no candidate rule yet. callee is NULL.
ambient_terminal	typescript	The target is an ambient declaration (a .d.ts signature with no body anywhere) — resolved, but there is nothing to expand into.
ambient_terminal	javascript	The callee or receiver VALUE is the platform (console.log, path.join, arr.forEach) — a correct end, not a blind spot; callee is NULL. Beside a project edge it is the platform ALTERNATIVE of a multi_inferred site.
implicit_constructor	javascript	new C() / super() where no constructor exists up the chain: the synthesized default runs. A correct end; callee is NULL.
dynamic_terminal	javascript	obj[expr](), eval, import(): no static target by construction; callee is NULL.
fan_capped	javascript, java, csharp	More targets than --dispatch-cap: the set was refused rather than emitted. JavaScript: callee is NULL. Java and C#: callee is the declared base method the fan would have started from; dispatch-capped-sites.csv carries the refused count.
callback_registered	javascript, typescript	The site HANDS the callee this function (xs.forEach(f), p.then(f), emitter.on('x', h), setTimeout(f)), which may invoke it. Not the site's own callee; a reachability edge, labelled so it is never read as a resolved call.
event_dispatch	javascript	x.emit('name') reaching a handler registered by x.on('name', h) on a value x may hold — name-sensitive for literal names, every handler on that value for a computed one.
event_dispatch	java	A Spring application event: publishEvent(e) reaching each listener (@EventListener, @TransactionalEventListener, ApplicationListener<E>.onApplicationEvent) whose declared event type e's static type is, or is a subtype of. Added beside the publishEvent boundary row, never in place of it (call-edge-generation/event_dispatch.dl). Also a JPA entity write (save, persist, merge, delete on a repository or EntityManager) reaching the @PrePersist/@PreUpdate/@PreRemove-style callbacks of the written entity's type and of the listeners @EntityListeners names on it or a superclass (kind entity_callback; call-edge-generation/entity_lifecycle.dl).
intrinsic_terminal	typescript	The site is a JSX intrinsic element or a dynamic import() — a runtime intrinsic, not a function the graph can name.

call_edges.callee_provenance values

value	languages	meaning
external	all	The target is outside every staged IR. callee_method_id is NULL and callee_label names it, as external:<Type>.<Member>. A client-only run of a C# project reaches the BCL this way for most of its calls.
client	all	Target is a client method (callee_method_id set).
lib	all	Target is a method of a staged library IR (callee_method_id set, methods.provenance = lib).
generated	csharp	The target is a member the compiler generated and no source declares (a boundary_generated row); callee_label names it.
builtin	python	Target is a CPython builtin with no Python source (callee_label = builtin:NAME).
external	python, java	Target is outside every staged IR and has no methods row. Python: an import path (callee_label = the written path). Java: a method of an unstaged ancestor type (callee_label = external:<type>.<name>), reached through a receiver declared as that type or inherited by a client subclass; see types.provenance external.

Notes

• javascript — Targets are VALUES the receiver may hold, not declared types: a multi_inferred set is the union of what flowed into the receiver. An untyped receiver is ambiguous_unknown, never a name match.
• python — A boundary_lib edge may point at a builtin (callee_provenance builtin, callee_label builtin:NAME) or at an unstaged import path (callee_provenance external) — neither has a methods row.
• java — A boundary_lib edge with callee_provenance external names a method of an ancestor type no staged IR declares (callee_label external:<type>.<name>, no methods row). A site whose receiver is declared as such a type is multi_inferred even with one client override: the platform method itself, and the platform's own subclasses, are the other possible targets. Stage the library to replace the label with the real method.
• python — The reason a site is ambiguous_unknown is exported per site in ext_call_site_unresolved (site, caller, reason, detail).
• all — THE TRUST LINE, and it is not the same set of tiers in every language. RESOLVED (callee_method_id is set): known_edge and multi_inferred in every language, and boundary_lib where the library is staged (--library) — without it boundary_lib names the target in callee_label and leaves callee_method_id NULL; ALSO ambient_terminal in TypeScript, fan_capped in Java and C# (the declared base, the fan refused), and runtime_observed in C#. HANDED OVER (callee set, but the site passes the function rather than calling it): callback_registered and event_dispatch in JavaScript, callback_registered in TypeScript; event_dispatch in C# too, where a mediator Send or Publish runs the handler for the request type, beside the row for the site itself. CORRECT END (callee NULL, and nothing is missing): intrinsic_terminal in TypeScript; ambient_terminal, implicit_constructor and dynamic_terminal in JavaScript; known_implicit_ctor, known_builtin_operator and boundary_generated (callee_label set) in C#. BLIND SPOT (callee NULL; exactly the tiers named ambiguous_*, which are what unresolved_sites holds): ambiguous_unknown everywhere, ALSO ambiguous_anon in Java and ambiguous_dynamic in C#. CAPPED (callee NULL, not in unresolved_sites): fan_capped in JavaScript. Python emits only the four shared tiers. A filter written as tier IN (known_edge, multi_inferred) therefore drops resolved edges in every language but Python — derive the set from this note or from unresolved_sites, never from a hardcoded list.
• java — A multi_inferred fan is CHA-wide: it is every override the hierarchy admits, bounded only by the dispatch cap. type_instantiated is computed and exported but NOT read by any rule, so the fan is not narrowed to types the program constructs. Narrow it yourself by joining dispatch_candidates to type_instantiated — see the dispatch_envelope_of query. A receiver is ALSO typed by what flows into it (a local's initializer, the arguments callers pass to a parameter, the receivers callers invoke a method on for its this), and each flow-in type resolves its member directly, outside the fan: that is why a fan_capped site still carries edges, and why they are the types the program was seen to hand over, not the whole hierarchy.
• typescript — A multi_inferred fan is CHA-wide, as in Java: type_instantiated is computed and exported but NOT read by any rule. The fan also has sources that are not virtual dispatch at all — an overload set or a union-typed receiver produces one too.
• python — A multi_inferred fan IS narrowed by the instantiation set: type_instantiated_reachable (the constructed classes and their bases) bounds dispatch in resolution/dispatch.dl. Python is the only front end where that narrowing is applied, so a fan here is tighter than the same shape would be in Java or TypeScript.
• all — The raw relation has a seventh column, ToExpr, that is always - (reserved). It is dropped here.
• all — An unresolved site (tier ambiguous_*) has NULL callee_method_id, callee_label and callee_provenance. The raw relation writes - in those slots.

type_ancestors

Transitive supertype closure: (type, ancestor) for every ancestor reachable through extends/implements/bases, client and library alike. Not a member-inheritance claim — in TypeScript an implements edge inherits nothing (the rule set keeps two closures; this is the conformance one).

#	column	type	key	null	idx	meaning
0	type_id	TEXT	yes		yes	FK → types.id.
1	ancestor_type_id	TEXT	yes		yes	FK → types.id.

dispatch_candidates

THE DISPATCH ENVELOPE: (base method, method that may run instead) for every call that statically resolves to the base. This is the set call_edges narrowed FROM — the difference between "these are the targets" and "these are the targets, out of these possibilities". Populated in every language but JavaScript (see its note); basis says what admitted the pair, because the front ends admit by different means.

#	column	type	key	null	idx	meaning
0	base_method_id	TEXT	yes		yes	FK → methods.id — the method a call resolves to statically.
1	candidate_method_id	TEXT	yes		yes	FK → methods.id — a method that may run instead at such a call.
2	basis	TEXT	yes			What admitted the pair — see vocabulary. Filter on it to trust only declarations.

dispatch_candidates.basis values

value	languages	meaning
nominal	java, typescript, csharp	A written extends/implements reaches the candidate's owner from the base's owner. The strongest evidence there is: the author declared the relationship.
structural	typescript	No declaration; the candidate's owner satisfies the base's owner by SHAPE. Emitted only for supertypes with no nominal implementor at all, so it never competes with a declared answer — but it is a heuristic, and a consumer that wants declarations only filters it out.
value	java, typescript, csharp, python	A function stored in a field, variable or parameter whose type is the base's callable type; a call through that holder may run it. In TypeScript also a function written as an implementation of an interface's member signature: an object-literal member of a literal typed by the interface (#1208), or a function assigned to the member through a property chain, inst.i.run = (x) => … (#1283). In C# also what one delegate member is assigned from another (a.Run = a.Parse). In Python, a function assigned onto an instance's member from outside its class, which shadows the class's method of that name. The base is the signature the call resolves to (bodiless), so the pair is how a walk over callers reaches the function that runs. Flow-derived rather than declared: a consumer that wants declarations only filters it out.
mro	python	The subtype's C3 linearisation picks the candidate for that attribute name. Not merely "the subtype declares this name" — a name a sibling base wins is attributed to that sibling.

Notes

• javascript — EMPTY — JavaScript has no declared dispatch to widen from: a multi_inferred set is already the values the receiver may hold (see call_edges), so there is no envelope to record.

overrides

Virtual-dispatch pairs: (base method, overriding method) wherever a call to the base may run the override. Java and C# only, and kept for compatibility — it is exactly dispatch_candidates filtered to basis = nominal. Prefer dispatch_candidates, which is populated in every language.

#	column	type	key	null	idx	meaning
0	method_id	TEXT	yes		yes	FK → methods.id — the base (declared) method.
1	overriding_method_id	TEXT	yes		yes	FK → methods.id — the override in a subtype.

Notes

• typescript — EMPTY — this table is Java-shaped. The TypeScript dispatch envelope is in dispatch_candidates, with basis nominal or structural.
• python — EMPTY — this table is Java-shaped. The Python dispatch envelope is in dispatch_candidates with basis mro, and value for a function assigned onto an instance's member; the raw linearisation is in ext_mro_position.

entry_points

Methods the runtime invokes without a client call site — process roots, test methods, HTTP handlers, framework hooks. The seeds of entry_reachable.

#	column	type	key	null	idx	meaning
0	method_id	TEXT	yes		yes	FK → methods.id.
1	reason	TEXT	yes			Why it is an entry — see vocabulary.

entry_points.reason values

value	languages	meaning
main	java, csharp	A static main. C#: a static Main, or the method top-level statements compile to.
test	java, typescript, csharp	Java: a JUnit test or lifecycle method. TypeScript: a function body handed to a test registrar (it, describe), inline or named, which the runner invokes. C#: an xUnit, NUnit or MSTest test method, or a set-up or tear-down hook of one.
http	java, python, typescript, csharp	A route handler a web framework invokes on a request. Java: a JAX-RS / Spring MVC handler. Python: a function registered with a decorator naming an HTTP verb and a URL path. TypeScript: a handler passed to a route registration (app.get('/x', h)), inline or named, or a method carrying a route decorator inside a container-owned class (@Controller + @Get). C#: a routed controller action, or a Razor Pages page model's On<Verb>[Handler][Async] method.
cli	java	A CLI command method (picocli etc.).
bean_ctor	java, typescript	Constructor of a container-managed class. TypeScript: the class carries a framework decorator (@Injectable, @Component, @Module), so the container constructs it and nothing in the repository does.
factory	java	A @Bean factory method.
lifecycle	java, typescript, csharp	Java: @PostConstruct / @PreDestroy and similar hooks. TypeScript: a hook the container calls by name on a decorated class (ngOnInit, onModuleInit), which has no call site anywhere. C#: a method the host calls on a hosted service (ExecuteAsync, StartAsync, StopAsync, and the IHostedLifecycleService hooks), including one that derives from the host's base through the project's own base class.
queue	java, csharp	A message-listener method. Java: also a Spring application event listener (@EventListener, @TransactionalEventListener, an ApplicationListener implementation). C#: a broker consumer or a bus message handler.
scheduled	java	A @Scheduled method.
grpc_service	java, python, csharp	A gRPC service implementation the server invokes on a request, with no call site reaching it: a generated ImplBase override (Java); a class deriving from a generated *Servicer base in a _pb2_grpc module, overriding a method that base declares (Python).
web_servlet	java	A servlet class named in web.xml (<servlet-class>): its container callbacks (doGet, service, …) and the library methods it overrides are invoked by the container.
web_filter	java	A filter class named in web.xml (<filter-class>); the container calls its callbacks.
web_listener	java	A listener class named in web.xml (<listener-class>); the container calls its callbacks (contextInitialized, sessionCreated, …).
lifecycle_init	java	The method an XML bean definition names as init-method; the container calls it after construction.
lifecycle_destroy	java	The method an XML bean definition names as destroy-method; the container calls it at shutdown.
lifecycle_factory	java	The method an XML bean definition names as factory-method; the container calls it to build the bean.
config_handler	java	A callback of a class a configuration file names under a key that expects a class (not an annotation); the container instantiates it and calls it.
service_loader	java	A callback of a provider listed in META-INF/services; ServiceLoader instantiates it and the caller reaches it through the service interface.
auto_configuration	java	The constructor or a container callback of a configuration class named in META-INF/spring/*.imports or under a configuration key of META-INF/spring.factories (EnableAutoConfiguration); Spring Boot registers it as a bean.
spring_factories	java	The constructor or callback of a class named under any other key of META-INF/spring.factories (ApplicationContextInitializer, EnvironmentPostProcessor, …); SpringFactoriesLoader instantiates it and calls it. Not a bean.
unimported_module	typescript, javascript	The initializer of a module nothing imports — a script or a bundle root.
exported_from_entry_module	typescript	A named function exported from a module nothing in the project imports. Its caller is the package's consumer, which is not in the repository. Placeholder names (<arrow>) are not roots.
package_export	typescript	A named function exported from a module the project's package.json publishes (main, module, exports, types, source), mapped from build output back to its source. Its caller is the package's consumer. Covers the entry a project's own tests import, which exported_from_entry_module cannot see.
task	python	A function registered as a queue task. A worker process runs the body; the producer only enqueues, so nothing in the client calls it.
fixture	python	A declared fixture that some collected test requests by parameter name. The runner calls it to build the argument.
url	python	A view named as a value in a module-level route table. The framework calls it on a request.
signal_receiver	python	A handler attached to a signal, by decorator or by connect(). It runs when the signal fires, whether or not this tree contains the send.
di_provider	python	A provider named in a dependency-injection marker in a parameter default. The framework calls it and passes the result in.
orm_hook	python, csharp	A hook the data layer calls; nothing in the client does. Python: a lifecycle or validation hook registered by decoration. C#: an Entity Framework Core override or implementation: OnModelCreating, OnConfiguring, IEntityTypeConfiguration<T>.Configure, a migration's Up/Down, a save-changes interceptor, IDesignTimeDbContextFactory<T>.CreateDbContext.
framework_hook	csharp	A method a framework calls on a class because the class derives from one of its base types or implements one of its interfaces, directly or through the project's own bases: an options setup class, a view component, an authorization handler, a model binder, a gRPC interceptor, a FluentValidation validator's constructor and overrides, a MediatR request or notification handler, Dispose/DisposeAsync on an IDisposable/IAsyncDisposable.
hub	csharp	A public instance method of a SignalR hub. A connected client invokes it by name.

Notes

• python — Framework entry points only: url, http, orm_hook, task, signal_receiver, fixture, di_provider and grpc_service. There is no test and no main reason: a pytest test is recognised by the query layer from its file and name, not here.

entry_reachable

Methods reachable from some entry point through call_edges (client edges only). A method absent here is dead from every known entry — or reachable only through a declared unknown.

#	column	type	key	null	idx	meaning
0	method_id	TEXT	yes			FK → methods.id.

unresolved_sites

The blind spots, attributed to the code that contains them: (caller, site) for every call site whose tier is ambiguous_* — a declared unknown, not an edge. A change-impact answer computed from a caller listed here is a lower bound. Derived from call_edges, so it is present in every language.

#	column	type	key	null	idx	meaning
0	caller_id	TEXT	yes		yes	Same domain as call_sites.caller_id: usually FK → methods.id.
1	call_site_id	TEXT	yes			FK → call_sites.id.

fields

Every field-like storage location the graph refers to: all client fields and enum constants from the IR, plus every LIBRARY field some field_access edge reaches OR some config_binding row names. A field is not a callable, so it has no row in methods; this is where field_access.field_id resolves to a name, an owner and a position.

#	column	type	key	null	idx	meaning
0	id	TEXT	yes			The parser's unique hash for the field or enum constant (FIELD_REGISTRY_… / ENUM_CONSTANT_…). The value field_access.field_id refers to.
1	name	TEXT			yes	Simple name as written (count, RED).
2	kind	TEXT				field or enum_constant — see vocabulary. An enum constant is a static final field of its enum and is recorded here so Colour.RED resolves like any other read.
3	owner_type_id	TEXT		yes	yes	FK → types.id of the declaring class/interface/enum/record.
4	owner_qualified_name	TEXT		yes		Qualified name of the owner, denormalised so a row prints without a join.
5	type_name	TEXT		yes		The declared type as the parser wrote it; NULL for an enum constant, whose type is its own enum.
6	modifiers	TEXT		yes		The parser's comma-separated modifier set (STATIC,FINAL); NULL where the IR records none.
7	file_path	TEXT		yes	yes	Source file.
8	start_line	INTEGER		yes		1-based first line of the declaration.
9	end_line	INTEGER		yes		1-based last line.
10	provenance	TEXT				client — from the analysed project; lib — from a staged library IR; generated — declared by an annotation processor and synthesised here (Java). See vocabulary.

fields.kind values

value	languages	meaning
field	csharp	A field declaration, including a const.
field	java, typescript	An ordinary field declaration.
enum_constant	java	An enum constant. It is a static final field of its enum, and is listed here so Colour.RED resolves like any other read; the parser gives it its own table and its own hash prefix.

fields.provenance values

value	languages	meaning
client	csharp	Declared in the analysed project.
lib	csharp	Declared in a staged library IR.
client	java, typescript	Declared in the analysed project.
lib	java, typescript	Declared in a staged library IR.
generated	java	Declared by a compile-time annotation processor and synthesised by the bundle, same shape and same reason as methods.provenance generated.

Notes

• all — JAVA, TYPESCRIPT AND C#. EMPTY for Python and JavaScript: a Python attribute is a symbols row of kind field. In C# it holds true fields only; a property is a symbols row of kind field with a property id, and its accessors are methods rows (PROPERTY_GET, PROPERTY_SET, PROPERTY_INIT).
• typescript — An enum member is absent: the parser gives it its own table with no declared type, and the property-access relation resolves through the field table. Colour.Red is therefore an unresolved field access, unlike Java where an enum constant is a fields row.
• java — A library field is listed when some field_access edge reaches it, exactly as methods lists only the library methods an edge reaches, OR when a config_binding row names it as the field a configuration key binds to. The second was added in #890: a @Value field on a library type that nothing reads has no access edge, so config_binding named a field the table did not list and the join lost the row silently.

field_access

THE DATA GRAPH. One row per (site, resolved field), and the answer to "who reads or writes this field" — the question call_edges cannot answer, because a field access is not a call. A site with N possible fields has N rows carrying the same tier; a site the engine could not resolve has exactly one row with a NULL field and tier ambiguous_unknown, so every field access written in the client appears at least once and the table alone shows where the resolution stopped. A field is NOT virtually dispatched: a known_edge row names the storage location, not a best estimate of one.

#	column	type	key	null	idx	meaning
0	site_id	TEXT			yes	The expression where the access is written. Not a call_sites id: a field access is not a call site. The location columns on this row are the site's own.
1	caller_id	TEXT			yes	The method whose body contains the access (FK → methods.id) — or, for an access written in a field initializer or an initializer block, the enclosing TYPE, exactly as call_sites.caller_id does. Never NULL.
2	field_id	TEXT		yes	yes	FK → fields.id of the resolved field or enum constant. NULL when the site is unresolved.
3	field_provenance	TEXT		yes		Where the field is declared — client or lib. NULL for an unresolved site.
4	access	TEXT			yes	Which way the data moves — see vocabulary. Present on an unresolved row too: the direction is decided by how the access is WRITTEN, which does not need the receiver to resolve.
5	tier	TEXT			yes	Confidence class of this edge — see vocabulary. Same four values and same promises as call_edges.tier.
6	file_path	TEXT		yes	yes	Source file of the site.
7	start_line	INTEGER		yes		1-based line of the site.
8	start_column	INTEGER		yes		Column, as the parser counts it.
9	end_line	INTEGER		yes		1-based last line.
10	end_column	INTEGER		yes		End column.

field_access.access values

value	languages	meaning
read	java, typescript	The value is used and not replaced.
write	java, typescript	The value is replaced without being read: a plain assignment f = v.
readwrite	java, typescript	The value is read and replaced at the one site: a compound assignment f += v, or f++ / --f. One row, not two — a consumer asking "who writes f" and one asking "who reads f" must both match it.

field_access.tier values

value	languages	meaning
known_edge	java, typescript	Exactly one field resolved. Stronger than the call_edges tier of the same name: a field is not virtually dispatched, so this IS the storage location the access binds to.
multi_inferred	java, typescript	A sound SET: the receiver has more than one possible type, or two unrelated ancestors declare the name (which Java itself treats as ambiguous). Each member is one row.
boundary_lib	java, typescript	The field is declared in a staged library type. field_id is set and resolves in fields with provenance lib.
ambiguous_unknown	java, typescript	Declared blind spot: the receiver could not be typed, or the name is not a member of the type it was typed to. field_id is NULL. Never dropped, and never replaced by a match on simple name.

field_access.field_provenance values

value	languages	meaning
client	java, typescript	The field is declared in the analysed project.
lib	java, typescript	The field is declared in a staged library IR.

Notes

• all — JAVA AND TYPESCRIPT. The table is declared in every bundle and is EMPTY for Python, JavaScript and C# (C# fills fields, not field_access), so the schema does not churn as the remaining front ends land (#663). Check SELECT count(*) FROM field_access before reading an empty result as "nothing reads this field".
• typescript — AN ACCESSOR IS NOT HERE. get url() read as c.url is a CALL, and call_edges already carries it with kind PROPERTY_READ or PROPERTY_WRITE (#703). The two tables are disjoint by construction: this one holds properties, call_edges holds accessors. Ask both when you want every read of a member.
• typescript — AN ELEMENT ACCESS IS NOT HERE either: obj["x"] with a literal key is a different node kind and is not yet a site. A known gap, not a silent one.
• typescript — A METHOD IS NOT A SITE. The callee of obj.m() is a PROPERTY_ACCESS node (37% of them, measured on one TypeScript library), and const f = obj.m reads a method as a value; neither is a data edge, and admitting them would fill the ambiguous_unknown tier with sites the engine HAS resolved elsewhere. Both are excluded and counted in ext_field_site_excluded with reasons method_callee and method_value.
• java — A field access written in a SWITCH CASE LABEL is deliberately absent. An enum constant in a case label is recorded TYPE for some arms and FIELD for others (#760), and javac compiles the switch through a $SwitchMap array rather than through a read of the constant, so there is no field access in the bytecode either.
• java — A field read that PRECEDES a same-named local declared later in the same method is missing: the parser classifies such a name LOCAL_VARIABLE against the whole body rather than against the scope at the use site (#725), so the site never reaches the engine and is absent rather than ambiguous. Rare (1 in 5,647 local references measured) but it is an absence, not a declared unknown.
• java — ARRAY ELEMENTS are not tracked: a[i] = v where a is a field is recorded as a READ of a (the array reference is read; the element write is not a field write). This matches the bytecode, where the instruction is getfield a followed by aastore.

type_use

THE OTHER HALF OF CHANGE IMPACT: one row per place a type is NAMED, with the context it was written in. call_edges says who calls a method and field_access who touches a field; this says what breaks if the TYPE changes — every signature, field, local, new, cast, instanceof, throws clause and generic argument that mentions it. A name that resolved to nothing is one row with a NULL type and tier ambiguous_unknown, so an unstaged third party is a declared unknown rather than an absence.

#	column	type	key	null	idx	meaning
0	reference_id	TEXT			yes	The type-reference node. Not an expression id: a type reference is its own IR entity.
1	type_id	TEXT		yes	yes	FK → types.id of the type the name denotes. NULL when it resolved to nothing.
2	context	TEXT			yes	Where the reference is written — see vocabulary. This is the column that makes an impact answer specific: seventeen METHOD_PARAM and four FIELD_TYPE, not twenty-one mentions.
3	depth	INTEGER				0 for the type as written, 1 or more for a type argument of the one above it. A field of type Map<String, Widget> yields Map at depth 0 and String and Widget at depth 1; all three are uses of the type named.
4	owner_kind	TEXT				What kind of declaration carries the reference — see vocabulary. It says what owner_id points at.
5	owner_id	TEXT				The declaration that carries the reference. FK → methods.id when owner_kind is METHOD, → types.id for TYPE, → fields.id for FIELD; for METHOD_PARAM, LOCAL_VARIABLE, EXPRESSION and the annotation kinds it is the parser hash of an entity the bundle does not table, so join on owner_method_id / owner_type_id instead.
6	owner_method_id	TEXT		yes	yes	FK → methods.id of the method whose declaration or body contains the reference; NULL where there is none (a field type, a supertype clause).
7	owner_type_id	TEXT		yes	yes	FK → types.id of the type whose source contains the reference. Set for every reference written inside a type declaration.
8	type_provenance	TEXT		yes		Where the referenced type is declared — client or lib. NULL when unresolved.
9	tier	TEXT			yes	Confidence class — see vocabulary. Same four values and same promises as call_edges.tier.

type_use.tier values

value	languages	meaning
known_edge	java, typescript	Exactly one type. A type reference is not dispatched, so this IS the declaration the name denotes.
multi_inferred	java, typescript	A sound SET: two resolution paths both answer a simple name. Each member is one row.
boundary_lib	java, typescript	The type is declared in a staged library IR. type_id resolves in types with provenance lib.
ambiguous_unknown	java, typescript	Declared blind spot: the name resolved to nothing — an unstaged third party, or a type variable with no bound in view. type_id is NULL. Never dropped.

type_use.type_provenance values

value	languages	meaning
client	java, typescript	The referenced type is declared in the analysed project.
lib	java, typescript	The referenced type is declared in a staged library IR.

type_use.owner_kind values

value	languages	meaning
TYPE	java, typescript	The reference is on the type declaration itself: an extends or implements clause, or a type parameter bound. owner_id is a types.id.
METHOD	java, typescript	A return type, a throws clause, or a method type-parameter bound. owner_id is a methods.id.
METHOD_PARAM	java, typescript	A formal parameter's declared type. owner_id is the parameter's parser hash; join on owner_method_id.
FIELD	java, typescript	A field's declared type. owner_id is a fields.id.
LOCAL_VARIABLE	java	A local, a catch parameter or a resource's declared type. owner_id is the local's parser hash; join on owner_method_id.
EXPRESSION	java, typescript	A type written inside an expression: new T(), a cast, an instanceof, a pattern, a method-reference qualifier. owner_id is the expression hash; join on owner_method_id.
ANNOTATION	java	The annotation type itself, on whatever it annotates.
ANNOTATION_ARGUMENT	java	A type named as an annotation argument, e.g. a Class<?> value.
VARIABLE	typescript	A const / let declaration's written type. A module-scope variable is a first-class declaration in TypeScript, so this covers what Java splits between FIELD and LOCAL_VARIABLE.
HERITAGE	typescript	An extends or implements clause, which the parser gives its own entity rather than hanging off the type.
TYPE_PARAMETER	typescript	A type parameter's bound or default.
TYPE_REFERENCE	typescript	Another type reference: the row is a type ARGUMENT or an element of the reference named in owner_id. depth says how deep.
DECORATOR	typescript	A decorator application.
ENUM_MEMBER	typescript	An enum member's written type.
EXPORT	typescript	An export type clause.
MODULE	typescript	A module-level position with no finer owner.

type_use.context values

value	languages	meaning
FIELD_TYPE	java, typescript	The declared type of a field.
METHOD_PARAM	java, typescript	The declared type of a formal parameter.
METHOD_RETURN	java, typescript	The declared return type.
LAMBDA_PARAMETER_TYPE	java	The declared type of an explicitly typed lambda parameter, (Foo f) -> f.bar(). It is a receiver-typing source, so it is what types f at the call inside the body.
LOCAL_VARIABLE	java	The declared type of a local, a catch parameter or a try-with-resources resource.
OBJECT_CREATION_TYPE	java, typescript	The type of a new T(...).
ARRAY_CREATION_TYPE	java	The element type of a new T[n].
CAST_EXPRESSION	java	The type of a (T) x.
INSTANCEOF_TYPE	java, typescript	The type tested by an x instanceof T.
PERMITS	java	A permitted subtype named in a sealed type's permits clause.
SUPER_TYPE	java, typescript	An extends clause.
IMPLEMENTS_INTERFACE	java, typescript	An implements clause.
THROWS_CLAUSE	java	A declared thrown type.
ANNOTATION_TYPE	java	The annotation type applied to a declaration.
ANNOTATION_PARAM	java	A type named as an annotation argument.
TYPE_PARAM_BOUND	java, typescript	The bound of a type parameter declared on a TYPE.
METHOD_TYPE_PARAM_BOUND	java, typescript	The bound of a type parameter declared on a METHOD.
METHOD_TYPE_ARGUMENT	java, typescript	An explicit type argument at a call site, x.<T>m().
METHOD_REFERENCE_QUALIFIER	java	The qualifier of a method reference, T::m.
PATTERN_BINDING_TYPE	java	The type of a record-pattern component.
SWITCH_TYPE_PATTERN	java	The type of a switch type pattern, case T t ->.
RECORD_PATTERN_TYPE	java	The record type a deconstruction pattern matches.
VARIABLE_TYPE	typescript	The written type of a const / let / var.
TYPE_ELEMENT	typescript	A member's type inside an interface or a type literal.
HERITAGE_TWIN	typescript	The second half of a heritage clause a declaration-merged type carries.
AS_TARGET	typescript	The target of an x as T.
SATISFIES_TARGET	typescript	The target of an x satisfies T.
TYPE_ASSERTION	typescript	The target of a <T>x assertion.
TYPE_ARGUMENT	typescript	A type argument of the reference in owner_id — Widget in Map<string, Widget>.
TYPE_PARAM_DEFAULT	typescript	A type parameter's default, the = T in <K = string>.
TYPE_ALIAS_RHS	typescript	The right-hand side of a type X = ….
INDEX_SIGNATURE_KEY	typescript	The key type of an index signature.
INDEX_SIGNATURE_VALUE	typescript	The value type of an index signature.
MAPPED_CONSTRAINT	typescript	The constraint of a mapped type.
MAPPED_TEMPLATE	typescript	The template of a mapped type.
CONDITIONAL_*	typescript	A prefix: the check, extends, true and false branches of a conditional type.
TEMPLATE_SPAN	typescript	A span of a template-literal type.
IMPORT_TYPE_QUALIFIER	typescript	The qualifier of an import("m").T.
TYPE_PREDICATE_TARGET	typescript	The target of an x is T predicate.
ENUM_MEMBER_TYPE	typescript	An enum member's written type.
DECORATOR_TYPE	typescript	The decorator itself.
DECORATOR_ARGUMENT_TYPE	typescript	A type named in a decorator argument.

Notes

• all — JAVA AND TYPESCRIPT. Declared in every bundle and EMPTY for Python, JavaScript and C#, so the schema does not churn as the remaining front ends land (#663).
• typescript — The context set is TypeScript's own and is wider than Java's: AS_TARGET, SATISFIES_TARGET, TYPE_ALIAS_RHS, the CONDITIONAL_* family, MAPPED_, INDEX_SIGNATURE_ and TEMPLATE_SPAN have no Java counterpart. A use inside a conditional type IS a use of that type and is recorded as one.
• typescript — Only a reference whose KIND can name a declaration is a row: TYPE_REFERENCE and IMPORT_TYPE. ARRAY, UNION, TUPLE and PARENTHESIZED are structure whose CHILDREN are the named references; PRIMITIVE, LITERAL, TYPE_VARIABLE, MAPPED, CONDITIONAL, INDEXED_ACCESS and INTRINSIC name nothing declared.
• java — EVERY DEPTH is here, unlike the receiver-typing relations the engine uses internally, which filter to depth 0. A field of type Map<String, Widget> produces three rows. Filter on depth = 0 when you want the type an expression has rather than every type its declaration mentions.
• java — A TYPE_VARIABLE reference (T, E) is not a row: it names the declaration's own parameter, not a type. Where the parameter has a written bound the USE resolves to that bound and IS a row, so <T extends Node> void f(T t) records a use of Node.
• java — The reference rows carry no line in the Java IR (every all-type-references row has an empty startLine), so this table has no position columns. Use owner_method_id, or owner_type_id plus the types row, to locate a use.
• csharp — EMPTY — the C# front end does not export type uses yet.

skipped

WHAT THE PARSER DID NOT READ — one row per source file it declined, with the reason. Every other table describes code that WAS read, so without this one a file the parser skipped is indistinguishable from a file with nothing in it: its declarations are absent, and every call that targets them reads as an engine miss rather than as a target that was never indexed. Join file_path against the other tables' file_path to separate the two. Empty is the normal case and means the parser read everything it was given.

#	column	type	key	null	idx	meaning
0	file_path	TEXT			yes	The file that was not read, as the parser recorded it — the same spelling the other tables' file_path uses, so the two join. For a DIRECTORY_EXCLUDED row it is a DIRECTORY, not a file (see notes).
1	reason	TEXT			yes	Why it was declined — see vocabulary. The sets differ per language because the front ends decline for different things.
2	construct	TEXT		yes		The syntactic form that caused the rejection (except_clause_comma_target), where the front end names one. Python only; NULL everywhere else.
3	start_line	INTEGER		yes		1-based line the offending construct is written at; NULL where the reason is a property of the whole file rather than of one place in it.
4	start_column	INTEGER		yes		Column, as the parser counts it; NULL with start_line.
5	detail	TEXT		yes		Free text from the front end: the error message for a READ_ERROR or EXTRACTION_ERROR, the count of files behind a DIRECTORY_EXCLUDED row. Not a vocabulary — do not match on it.

skipped.reason values

value	languages	meaning
READ_ERROR	java, typescript, python, javascript	The file could not be read — an I/O or encoding failure. The ENVIRONMENT failed, which is nobody's bug; contrast EXTRACTION_ERROR.
EMPTY_CONTENT	java	The file is empty or is only whitespace. Nothing was lost.
FILE_TOO_LARGE	java	The file is longer than the parser's line threshold and was declined whole. Everything it declares is missing from every other table.
EXTRACTION_ERROR	typescript, python, javascript	The file read and parsed, and the extractor then threw. The PARSER failed, which is always a bug — detail carries the message.
PY2_CONSTRUCT_DETECTED	python	Python 2 source, rejected whole rather than misread under Python 3 scoping (tree-sitter parses print "x" without erroring). construct, start_line and start_column name the form that gave it away.
NO_PROGRAM_CLAIMS_FILE	typescript	A file under a root that declares programs (a tsconfig) which no program claims and no claimed file imports. It is out of every program, not unparseable.
DIRECTORY_EXCLUDED	javascript	A directory the walker pruned by name (node_modules, dist, …). ONE row per DIRECTORY, with the file count in detail — the files were never enumerated, and naming them individually would invent paths.

Notes

• all — THE ONLY TABLE ABOUT CODE THAT IS NOT IN THE GRAPH. Read it before reading any absence as an engine result: a call into a skipped file is unresolved because the target was never indexed, not because the rules could not resolve it. Rows describe the CLIENT only — a library file the parser skipped is not reported here.
• csharp — EMPTY — the C# front end writes no skipped-files report. It takes the opposite line: a construct its grammar does not cover fails the run rather than skipping the file, so there is no per-file decision to record. An empty table here is not evidence that every file was read.
• javascript — A DIRECTORY_EXCLUDED row's file_path is a PRUNED DIRECTORY, not a file, and detail carries how many files are behind it; those files have no rows of their own. So SELECT count(*) FROM skipped is not the number of files missing, and a join on file_path will not match them. Filter the reason out when you want per-file rows.
• python — The only front end that positions a skip: a PY2_CONSTRUCT_DETECTED row carries the construct and its line and column, so the file can be triaged without re-running the parser.

type_instantiated

Types this run creates an instance of — the rapid-type-analysis set that bounds virtual dispatch. (A subtype nothing instantiates cannot receive a dispatched call.) Deliberately an over-approximation: narrowing it on evidence the run does not have would lose real edges. Populated in every language.

#	column	type	key	null	idx	meaning
0	type_id	TEXT	yes		yes	FK → types.id.
1	how	TEXT	yes			What creates the instance — see vocabulary.

type_instantiated.how values

value	languages	meaning
new	all	A constructor call — new C() / C().
anonymous	java	An anonymous class exists only by being instantiated.
enum_constant	java	An enum's constants are its instances.
service_loader	java	A provider listed in META-INF/services: ServiceLoader constructs it reflectively, with no new in the source.
auto_configuration	java	A configuration class named in META-INF/spring/*.imports or spring.factories (EnableAutoConfiguration): Spring Boot constructs it reflectively.
spring_factories	java	A class named under another key of META-INF/spring.factories: SpringFactoriesLoader constructs it reflectively.

Notes

• typescript — Every row has how = new. Not restricted to client provenance: a type the library constructs is still a type that exists at run time, and dropping it would narrow the envelope unsoundly.
• python — Every row has how = new: the rule set records that some client call constructs the class, not which form.
• csharp — EMPTY — the C# engine derives the instantiation set but does not export it. Dispatch in C# is not narrowed by it either (see the README of the C# engine).

Extended tables — ext_<relation>

Every relation in the language's graph/<lang>/souffle/export_manifest.tsv, loaded as ext_<relation> with positional columns c0…cN (the raw relation is declared positionally; nothing here invents a name). schema_tables lists each one with its arity and the comment lifted from the rule that derives it — read that before querying. They are language-specific by construction: a bundle holds only the ext tables of its own language.

Catalog tables

schema_tables

Every table in this bundle with its scope and what it holds.

#	column	type	meaning
0	name	TEXT	Table name.
1	scope	TEXT	core — same schema in every language; ext — a language-specific relation, columns c0…cN; catalog — this documentation.
2	language	TEXT	For ext: the front end that defines it. NULL for core and catalog.
3	description	TEXT	What a row means. For ext this is the comment lifted from the rule that derives the relation, verbatim.

schema_columns

Every column of every core and catalog table. (ext tables are positional: c0…cN, arity in schema_tables.description.)

#	column	type	meaning
0	table_name	TEXT	FK → schema_tables.name.
1	ordinal	INTEGER	0-based position.
2	name	TEXT	Column name.
3	type	TEXT	TEXT or INTEGER.
4	nullable	INTEGER	1 if the column may be NULL (empty in CSV).
5	description	TEXT	Meaning.

schema_vocab

Every enumerated value a core column may hold, and WHICH LANGUAGES emit it. Filter on the language in run to see the values that can occur in this bundle. A value observed in this run but not in the authored list is inserted with meaning undocumented — observed in this run, so the table is complete for the data it sits next to.

#	column	type	meaning
0	table_name	TEXT	Core table.
1	column_name	TEXT	Column.
2	language	TEXT	java, typescript, javascript, python, csharp — the front end that emits this value (one row per language; a value shared by all has one row for each).
3	value	TEXT	The value as it appears in the column. A trailing * marks a prefix (e.g. DECORATOR_*).
4	meaning	TEXT	What it means.

schema_guide

READ THIS FIRST. An ordered walkthrough of how to use this database: which tables answer which questions, what to check before trusting an answer, and where the language-specific details are.

#	column	type	meaning
0	step	INTEGER	Reading order.
1	text	TEXT	The instruction.

schema_queries

Canonical questions and the SQL that answers each, parameterised with named :params. Every query is verified to run against every language's bundle. Copy, bind, run.

#	column	type	meaning
0	name	TEXT	Short identifier.
1	question	TEXT	The question in words.
2	params	TEXT	Comma-separated named parameters the SQL expects, e.g. :qualified_name, :depth.
3	sql	TEXT	The SQL.

schema_notes

Per-language caveats that are not a vocabulary: what a table lacks in one front end, where an id may point, what a NULL means here.

#	column	type	meaning
0	language	TEXT	java, typescript, javascript, python, csharp, or all.
1	table_name	TEXT	The table the note is about.
2	note	TEXT	The caveat.