Local code intelligence for developers and AI agents. Connect code, docs and infrastructure in a persistent graph, with CBM Lens, MCP tools, agent integrations and change-aware test selection.
C aiderastclaude-codecode-analysiscode-intelligencecodexcursorcypherdeveloper-toolsgemini-cligraph-visualizationkilocodeknowledge-graphmcpmcp-servermodel-context-protocolopencodesqlitetree-sitterwindsurf
/*
* Herkunft: portiert am 2026-08-28 aus CodeAtlasIDE,
* /Users/bernhard/Desktop/CodeAtlasIDE/theia-extensions/codeatlas-core/src/common/trace-protocol.ts
* Gleicher Urheber (Bernhard Jackiewicz). Semantisch unveraendert uebernommen
* wurden genau die zwei Funktionen, die `twin-view-model.ts` braucht:
* `runtimeCitationsFor` und der Namensvergleich `sameSymbol`, den sie benutzt.
*
* Weggelassen: der ganze Rest der Datei (Ereignis-Schema, Speicherpfad,
* Import-Bericht, die Bug-Pfad-DTOs, die Deckel). Der Grund ist nicht
* Bequemlichkeit: dieses Projekt hat keinen Trace-Speicher und importiert
* nichts, also gaebe es zu jenen Typen hier nichts, was sie beschreiben. Was
* bleiben MUSS, ist die Zuordnung von Beobachtung zu Aufruf, weil der Twin sie
* beim Zitieren einer Schritt-Zeile aufruft: liesse man sie weg und gaebe
* stattdessen eine leere Liste zurueck, wuerde eine IR mit `runtime`-Fakt hier
* still weniger Belege zeigen als im Referenzprojekt, und genau das ist die Art
* Abweichung, die niemand mehr findet.
*
* Aenderungen gegenueber dem Original: keine, ausser den Importpfaden auf die
* ebenfalls portierte semantic-ir.ts im selben Verzeichnis.
*/
import type { Evidence, SemanticIR } from './semantic-ir';
/**
* The runtime citations that back one row of a call fact.
*
* One function, imported by both sides, because the backend's evidence lookup
* and the twin's have to agree about which arrow a citation belongs to. They
* already duplicate the fact-path grammar; duplicating the matching rule as well
* would let the popover and the panel disagree about the same call.
*
* The match is by name: a stored event carries whatever the recorder wrote, and
* a recorder that emits qualified names and one that emits bare names describe
* the same call. So a bare `createUser` matches a qualified
* `src.services.userService.createUser` and the other way round, and nothing
* else does: a name that merely contains another is not a match, because
* `createUserGroup` is a different function.
*/
export function runtimeCitationsFor(ir: SemanticIR, factPath: string): Evidence[] {
const match = /^(steps|calls)\[(\d+)\]$/.exec(factPath);
const runtime = ir.runtime;
if (!match || runtime === undefined) {
return [];
}
const calls = match[1] === 'steps' ? ir.steps.value : ir.calls.value;
const call = calls[Number(match[2])];
if (call === undefined) {
return [];
}
// Only trust the row index when the two arrays line up, exactly as the
// evidence lookups do: a mislabelled citation is worse than no citation.
if (runtime.value.length !== runtime.evidence.length) {
return [];
}
const out: Evidence[] = [];
runtime.value.forEach((observed, index) => {
if (sameSymbol(observed.targetName, call.targetName)
|| sameSymbol(observed.targetQualifiedName, call.targetQualifiedName)
|| sameSymbol(observed.targetQualifiedName, call.targetName)
|| sameSymbol(observed.targetName, call.targetQualifiedName)) {
out.push(runtime.evidence[index]);
}
});
return out;
}
/**
* Whether two written names denote the same symbol.
*
* Equal, or one is the last dotted segment of the other. Nothing looser: a
* substring rule would make `validateUser` match `validateUserInput`, and an
* arrow attributed to the wrong function is the failure this whole feature
* exists to avoid.
*/
export function sameSymbol(left: string | undefined, right: string | undefined): boolean {
if (left === undefined || right === undefined || left.length === 0 || right.length === 0) {
return false;
}
return left === right || left.endsWith(`.${right}`) || right.endsWith(`.${left}`);
}