feat(resolution): CFML receiver-type inference for locals, typed args, and component properties (#1155)

CFML joins the #1108 receiver-inference family: new/createObject/typed-arg/property(inject) declarations type the receiver, variables./this. fields scan whole-file, method QNs re-scoped to Class::member in all three extraction paths. 1,649 typed edges on fw1/ColdBox/CFWheels, 1,649/1,649 audit-consistent, inherited methods resolve via #1152 extends edges.

Co-authored-by: ghedwards <125586+ghedwards@users.noreply.github.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
Colby Mchenry
2026-07-02 19:02:16 -05:00
committed by GitHub
co-authored by ghedwards Claude Fable 5
parent 5f22da35f3
commit 7d624ecfac
5 changed files with 308 additions and 25 deletions
+1
View File
@@ -13,6 +13,7 @@ and adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
- CodeGraph now indexes **CFML** (`.cfc`, `.cfm`, `.cfs`) — both the classic tag-based style (`<cfcomponent>`/`<cffunction>`) and modern bare-script `component { ... }` syntax, including `extends`/`implements`, embedded `<cfscript>` blocks (at any nesting depth, including inside `<cfif>`/`<cfloop>`/`<cftry>`), call edges, and calls embedded in `#hash#` expressions inside `<cfquery>` SQL bodies. Files saved with a UTF-8 byte-order mark and tags with unquoted attribute values — both common in long-lived CFML codebases — are handled too. Thanks @ghedwards. (#1118)
- CFML inheritance written as a component path now links to the right component. `extends="coldbox.system.web.Controller"` names its supertype by dotted path and `extends="../base"` by relative path (the FW/1 style) — both previously produced no inheritance edge at all, which on framework-style CFML apps hid most of the type hierarchy from impact and blast-radius analysis (on ColdBox's own core, over 90% of inheritance was invisible). Resolution is deliberately conservative: the target's directory layout must corroborate the declared path — so a supertype that lives in an out-of-repo library (testbox, mxunit, an installed framework) correctly stays unlinked rather than being guessed at, and an ambiguous path produces no edge rather than a wrong one. (#1152)
- CFML method calls made through a local variable, typed argument, or component property now resolve to the right method — the same receiver-type inference the other object-oriented languages already had. `var svc = new UserService(); svc.save()`, `createObject("component", "path.UserService")`, a typed `<cfargument>` or cfscript parameter, and `variables.`/`this.`-scoped fields — including the pseudoconstructor pattern (`variables.svc = new UserService()` in `init()`) and WireBox-injected properties (`property name="svc" inject="UserService"`) — all now link the call to the declared component's method, with methods inherited from a supertype resolved through the inheritance links above. This makes callers, impact/blast-radius, and `codegraph_explore` flow traces follow CFML service calls instead of dropping them or guessing among same-named methods.
- The Claude Code context hook now recognizes prompts that describe code in plain words — in any language — by checking the prompt's words against the symbol names actually in your project's index. Asking about "the state machine des commandes" finds `OrderStateMachine` with no keyword involved. Confidence decides how much gets injected: structural questions and prompts naming a real symbol still get full context up front; a plain-words match gets a short pointer to the matching symbols so the agent queries them itself; everything else stays silent, exactly as before.
- Anonymous usage telemetry now counts how often the context hook injected context, offered a hint, or stayed silent — fixed counter names only; the prompt's content is never stored or sent. This makes the hook's accuracy measurable instead of guessed. The counters record what actually happened, not what was attempted: a lookup that errors or comes back empty counts as a distinct silent outcome, never as delivered context (#1143, thanks @inth3shadows).
- Metal shader files (`.metal`) are now indexed. Metal Shading Language is close enough to C++ that vertex/fragment/kernel functions, structs, type aliases, and the calls between them all land in the graph — so shader pipelines in Apple-platform projects show up in impact analysis and flow traces instead of being silently skipped. Metal's `[[buffer(0)]]`-style attribute annotations are handled so they can't corrupt what gets extracted. Thanks @FluxKo for the report. (#1121)
+185
View File
@@ -0,0 +1,185 @@
/**
* CFML local-variable / component-field receiver-type inference (#1108 family).
*
* `var svc = new UserService(); svc.save()` the call's receiver type is
* recoverable from its declaration, and resolveMethodOnType validates the
* inferred type actually declares the method, so a mis-inference produces no
* edge. CFML brings four declaration idioms the shared inferrer must know:
* `new` (dotted component paths included), `createObject("component", "...")`,
* typed arguments (cfscript params and `<cfargument>` tags), and component
* properties including WireBox DI (`property name="svc" inject="..."`),
* whose receivers are `variables.`-scoped fields declared OUTSIDE the calling
* function (so the scan must widen to the whole file, in both directions).
*
* These tests also pin the extraction prerequisite: CFML method
* qualifiedNames carry the component scope (`UserService::save`) in all three
* extraction paths (bare-script, `<cffunction>`, component-level `<cfscript>`
* blocks) without that, type-validated resolution can never match.
*/
import { describe, it, expect, beforeEach, afterEach } from 'vitest';
import * as fs from 'node:fs';
import * as path from 'node:path';
import * as os from 'node:os';
import { CodeGraph } from '../src';
describe('CFML receiver-type inference', () => {
let dir: string;
beforeEach(() => { dir = fs.mkdtempSync(path.join(os.tmpdir(), 'cfml-recv-')); });
afterEach(() => { fs.rmSync(dir, { recursive: true, force: true }); });
const write = (rel: string, body: string) => {
const p = path.join(dir, rel);
fs.mkdirSync(path.dirname(p), { recursive: true });
fs.writeFileSync(p, body);
};
const load = async () => {
const cg = await CodeGraph.init(dir, { silent: true });
await cg.indexAll();
const db = (cg as any).db.db;
const calls: { src: string; tgt: string; tgtQn: string }[] = db
.prepare(
`SELECT s.name src, t.name tgt, t.qualified_name tgtQn
FROM edges e JOIN nodes s ON s.id = e.source JOIN nodes t ON t.id = e.target
WHERE e.kind = 'calls' AND t.kind = 'method'`
)
.all();
const methods: { name: string; qn: string }[] = db
.prepare(`SELECT name, qualified_name qn FROM nodes WHERE kind = 'method'`)
.all();
cg.close?.();
return { calls, methods };
};
const hasCall = (calls: any[], src: string, tgtQn: string) =>
calls.some((e) => e.src === src && e.tgtQn === tgtQn);
// Two same-named methods so resolution MUST disambiguate by receiver type —
// plain name-matching alone can't pick one.
const userService = `component {\n function save(any u) { return u; }\n}\n`;
const orderService = `component {\n function save(any o) { return o; }\n}\n`;
it('scopes method qualifiedNames under the component in all three extraction paths', async () => {
write('svc/UserService.cfc', userService);
write('tag/TagService.cfc', `<cfcomponent>\n<cffunction name="save"><cfreturn 1></cffunction>\n</cfcomponent>\n`);
write('mod/ModuleConfig.cfc', `<cfcomponent>\n<cfscript>\nfunction configure() { return 1; }\n</cfscript>\n</cfcomponent>\n`);
const { methods } = await load();
expect(methods.find((m) => m.name === 'save' && m.qn === 'UserService::save')).toBeDefined();
expect(methods.find((m) => m.name === 'save' && m.qn === 'TagService::save')).toBeDefined();
expect(methods.find((m) => m.name === 'configure' && m.qn === 'ModuleConfig::configure')).toBeDefined();
});
it('infers a local declared with new, including a dotted component path', async () => {
write('svc/UserService.cfc', userService);
write('svc/OrderService.cfc', orderService);
write('handlers/Main.cfc', `component {
function bare() {
var svc = new UserService();
return svc.save(1);
}
function dotted() {
var svc2 = new svc.UserService();
return svc2.save(2);
}
}
`);
const { calls } = await load();
expect(hasCall(calls, 'bare', 'UserService::save')).toBe(true);
expect(hasCall(calls, 'dotted', 'UserService::save')).toBe(true);
expect(hasCall(calls, 'bare', 'OrderService::save')).toBe(false);
});
it('infers a local declared with createObject (two-arg and single-arg forms)', async () => {
write('svc/UserService.cfc', userService);
write('svc/OrderService.cfc', orderService);
write('handlers/Legacy.cfc', `component {
function classic() {
var svc = createObject("component", "svc.UserService");
return svc.save(1);
}
function modern() {
var svc2 = CreateObject("svc.OrderService");
return svc2.save(2);
}
}
`);
const { calls } = await load();
expect(hasCall(calls, 'classic', 'UserService::save')).toBe(true);
expect(hasCall(calls, 'modern', 'OrderService::save')).toBe(true);
});
it('infers a typed cfscript parameter', async () => {
write('svc/UserService.cfc', userService);
write('svc/OrderService.cfc', orderService);
write('handlers/Typed.cfc', `component {
function process(required UserService svc) {
return svc.save(1);
}
}
`);
const { calls } = await load();
expect(hasCall(calls, 'process', 'UserService::save')).toBe(true);
expect(hasCall(calls, 'process', 'OrderService::save')).toBe(false);
});
it('infers a <cfargument> typed argument used inside a <cfscript> body', async () => {
write('svc/UserService.cfc', userService);
write('svc/OrderService.cfc', orderService);
write('handlers/TagTyped.cfc', `<cfcomponent>
<cffunction name="process">
<cfargument name="svc" type="svc.UserService">
<cfscript>
return svc.save(1);
</cfscript>
</cffunction>
</cfcomponent>
`);
const { calls } = await load();
expect(hasCall(calls, 'process', 'UserService::save')).toBe(true);
});
it('infers a variables-scoped field from its pseudoconstructor assignment, even when init sits below the call', async () => {
write('svc/UserService.cfc', userService);
write('svc/OrderService.cfc', orderService);
write('handlers/Fielded.cfc', `component {
function handle() {
return variables.svc.save(1);
}
function init() {
variables.svc = new UserService();
return this;
}
}
`);
const { calls } = await load();
expect(hasCall(calls, 'handle', 'UserService::save')).toBe(true);
expect(hasCall(calls, 'handle', 'OrderService::save')).toBe(false);
});
it('infers a WireBox-injected property (the ColdBox DI shape)', async () => {
write('svc/UserService.cfc', userService);
write('svc/OrderService.cfc', orderService);
write('handlers/Injected.cfc', `component {
property name="svc" inject="UserService";
function handle() {
return variables.svc.save(1);
}
}
`);
const { calls } = await load();
expect(hasCall(calls, 'handle', 'UserService::save')).toBe(true);
});
it('creates no method edge when the inferred type does not declare the method', async () => {
write('svc/UserService.cfc', userService);
write('handlers/Wrong.cfc', `component {
function go() {
var svc = new UserService();
return svc.destroyEverything();
}
}
`);
const { calls } = await load();
expect(calls.filter((e) => e.src === 'go')).toHaveLength(0);
});
});
+34 -19
View File
@@ -73,6 +73,12 @@ export class CfmlExtractor {
if (node.name === '<anonymous>' && (node.kind === 'class' || node.kind === 'interface')) {
node.name = componentName;
node.qualifiedName = `${this.filePath}::${componentName}`;
} else if (node.qualifiedName === '<anonymous>' || node.qualifiedName.startsWith('<anonymous>::')) {
// Members were scoped under the anonymous component (`<anonymous>::save`)
// — carry the rename into their scope chains so type-validated method
// resolution (which wants `UserService::save`, see resolveMethodOnType)
// can match them. Inner genuinely-anonymous segments are untouched.
node.qualifiedName = componentName + node.qualifiedName.slice('<anonymous>'.length);
}
this.nodes.push(node);
}
@@ -225,13 +231,13 @@ export class CfmlExtractor {
break;
}
if (sibling.type === 'cf_function_tag') {
this.extractFunctionTag(sibling, classNode.id, classNode.id);
this.extractFunctionTag(sibling, classNode.id, classNode.id, classNode.name);
} else if (sibling.type === 'cf_script_tag') {
this.delegateScriptTag(sibling, classNode.id, true);
this.delegateScriptTag(sibling, classNode.id, classNode.name);
} else if (sibling.type === 'cf_query_tag') {
this.delegateQueryTag(sibling, classNode.id);
} else {
this.delegateNestedTags(sibling, classNode.id, true);
this.delegateNestedTags(sibling, classNode.id, classNode.name);
}
lastNode = sibling;
sibling = sibling.nextSibling;
@@ -246,9 +252,11 @@ export class CfmlExtractor {
* the `contains`-edge target (the class when inside one, otherwise the file
* node for a bare top-level cffunction) kept separate so a top-level
* function still gets a containment edge without being misclassified as a
* method of the file.
* method of the file. A method's qualifiedName is scoped under
* `parentClassName` (`TagService::save`, the same `Class::member` shape the
* generic extractor produces) so type-validated method resolution can match.
*/
private extractFunctionTag(tag: SyntaxNode, parentClassId: string | undefined, containerId: string | undefined): void {
private extractFunctionTag(tag: SyntaxNode, parentClassId: string | undefined, containerId: string | undefined, parentClassName?: string): void {
const name = this.tagAttr(tag, 'name');
if (!name) return;
@@ -264,7 +272,7 @@ export class CfmlExtractor {
id,
kind,
name,
qualifiedName: `${this.filePath}::${name}`,
qualifiedName: parentClassName ? `${parentClassName}::${name}` : `${this.filePath}::${name}`,
filePath: this.filePath,
language: this.language,
startLine: tag.startPosition.row + 1,
@@ -293,34 +301,38 @@ export class CfmlExtractor {
* `<cfcomponent>`'s body see the implicit-end-tag note on `extractComponent`)
* ARE normal children, just possibly several levels deep, so a direct-children
* check misses them. Does not descend into a nested `cf_function_tag` that
* has its own scope and is walked separately. `parentIsClass` rides along so
* a `<cfscript>` at component scope classifies its functions as methods.
* has its own scope and is walked separately. `parentClassName` rides along
* so a `<cfscript>` at component scope classifies its functions as methods
* scoped under the component.
*/
private delegateNestedTags(node: SyntaxNode, containerId: string | undefined, parentIsClass = false): void {
private delegateNestedTags(node: SyntaxNode, containerId: string | undefined, parentClassName?: string): void {
for (let i = 0; i < node.namedChildCount; i++) {
const child = node.namedChild(i);
if (!child) continue;
if (child.type === 'cf_script_tag') {
this.delegateScriptTag(child, containerId, parentIsClass);
this.delegateScriptTag(child, containerId, parentClassName);
} else if (child.type === 'cf_query_tag') {
this.delegateQueryTag(child, containerId);
} else if (child.type === 'cf_function_tag') {
continue;
} else {
this.delegateNestedTags(child, containerId, parentIsClass);
this.delegateNestedTags(child, containerId, parentClassName);
}
}
}
/**
* Delegate a `<cfscript>...</cfscript>` tag body to the cfscript grammar.
* With `parentIsClass`, functions declared at the script's top level are the
* component's methods (`<cfcomponent><cfscript>function configure(){}` the
* standard ColdBox ModuleConfig shape), so they're re-kinded `function`
* `method` to match how the same function classifies in a script-style CFC.
* Functions nested inside another function (closures) keep kind `function`.
* With `parentClassName` set (the block sits at component scope), functions
* declared at the script's top level are the component's methods
* (`<cfcomponent><cfscript>function configure(){}` the standard ColdBox
* ModuleConfig shape): they're re-kinded `function` `method`, and every
* merged symbol's qualifiedName is prefixed with the component scope
* (`configure` `ModuleConfig::configure`) so type-validated method
* resolution can match them. Functions nested inside another function
* (closures) keep kind `function`.
*/
private delegateScriptTag(scriptTag: SyntaxNode, parentId: string | undefined, parentIsClass = false): void {
private delegateScriptTag(scriptTag: SyntaxNode, parentId: string | undefined, parentClassName?: string): void {
const content = scriptTag.namedChildren.find((c: SyntaxNode) => c.type === 'cf_script_content');
if (!content) return;
@@ -349,8 +361,11 @@ export class CfmlExtractor {
node.startLine += startLine;
node.endLine += startLine;
node.language = this.language;
if (parentIsClass && node.kind === 'function' && topLevelIds.has(node.id)) {
node.kind = 'method';
if (parentClassName) {
if (node.kind === 'function' && topLevelIds.has(node.id)) {
node.kind = 'method';
}
node.qualifiedName = `${parentClassName}::${node.qualifiedName}`;
}
this.nodes.push(node);
if (parentId) {
+15
View File
@@ -3830,6 +3830,21 @@ export class TreeSitterExtractor {
else reencode = !!innerCallee;
}
calleeName = reencode ? `${innerCallee}().${methodName}` : methodName;
} else if (
this.language === 'cfscript' &&
receiver &&
receiver.type === 'member_expression' &&
/^(variables|this|local|arguments)\.[A-Za-z_][\w]*$/i.test(getNodeText(receiver, this.source))
) {
// CFML scope-prefixed member call — `variables.svc.save()` /
// `arguments.svc.save()`: the receiver is a component field,
// injected property, or typed argument reached through one of
// CFML's file-local scopes. Keep the full receiver chain so
// resolution can strip the scope prefix and infer the field's
// component type from its declaration (#1108). Gated to these
// scope keywords: such calls previously emitted a bare method
// name, which either failed to resolve or resolved ambiguously.
calleeName = `${getNodeText(receiver, this.source)}.${methodName}`;
} else {
calleeName = methodName;
}
+73 -6
View File
@@ -1185,6 +1185,36 @@ function localReceiverTypePatterns(language: Language, r: string): RegExp[] {
new RegExp(`\\b${r}\\b\\s*:\\s*([A-Z][\\w]*)`), // var lg: TLogger / param lg: TLogger
new RegExp(`\\b${r}\\b\\s*:=\\s*([A-Z][\\w.]*)\\.Create\\b`), // lg := TLogger.Create
];
case 'cfml':
case 'cfscript':
return [
// svc = new UserService() / new path.to.UserService() — dotted component
// paths reduce to their final segment via normalizeInferredTypeName.
// Also matches inside tag markup (`<cfset svc = new UserService()>`)
// since the scan reads raw source lines.
new RegExp(`\\b${r}\\b\\s*=\\s*new\\s+([A-Za-z_][\\w.]*)`),
// The classic form: svc = createObject("component", "path.to.UserService")
// (casing of createObject varies in the wild), plus the modern
// single-argument form createObject("path.to.UserService").
new RegExp(`\\b${r}\\b\\s*=\\s*[Cc]reate[Oo]bject\\s*\\(\\s*["']component["']\\s*,\\s*["']([\\w.]+)["']`),
new RegExp(`\\b${r}\\b\\s*=\\s*[Cc]reate[Oo]bject\\s*\\(\\s*["']([\\w.]+)["']\\s*\\)`),
// Typed cfscript parameter: `function save(UserService svc)` /
// `required UserService svc` — CFML's built-in types (string, numeric,
// any, struct…) are lowercase by convention, so the PascalCase guard
// excludes them.
new RegExp(`\\b([A-Z][\\w.]*)\\s+${r}\\b\\s*[=;,)]`),
// Tag-form typed argument, either attribute order:
// <cfargument name="svc" type="path.to.UserService">
new RegExp(`\\bcfargument[^>\\n]*\\bname\\s*=\\s*["']${r}["'][^>\\n]*\\btype\\s*=\\s*["']([\\w.]+)["']`, 'i'),
new RegExp(`\\bcfargument[^>\\n]*\\btype\\s*=\\s*["']([\\w.]+)["'][^>\\n]*\\bname\\s*=\\s*["']${r}["']`, 'i'),
// Component property (incl. WireBox DI): `property name="svc"
// inject="UserService";` / `<cfproperty name="svc" type="UserService">`,
// either attribute order. An inject DSL value with a namespace
// (`inject="svc@core"`) captures only the leading name and simply
// fails type-validation — no edge, never a wrong one.
new RegExp(`\\b(?:cf)?property\\b[^;\\n]*\\bname\\s*=\\s*["']${r}["'][^;\\n]*\\b(?:type|inject)\\s*=\\s*["']([\\w.]+)["']`, 'i'),
new RegExp(`\\b(?:cf)?property\\b[^;\\n]*\\b(?:type|inject)\\s*=\\s*["']([\\w.]+)["'][^;\\n]*\\bname\\s*=\\s*["']${r}["']`, 'i'),
];
default:
return [];
}
@@ -1215,9 +1245,28 @@ function inferLocalReceiverType(
ref: UnresolvedRef,
context: ResolutionContext,
): string | null {
// CFML scope prefixes: `variables.svc` / `this.svc` name a COMPONENT-scoped
// field whose assignment or `property` declaration usually lives outside the
// calling function (the init-pseudoconstructor / WireBox-injection pattern),
// and `local.svc` is an explicit function-local. Strip the prefix so the
// declaration patterns match (`variables.svc = new X()`, `property
// name="svc" …`, `var svc = …` all bind the bare name), and widen the scan
// to the whole file for the component-scoped forms — nearest-declaration-
// backward still wins, so a function-local shadowing the field is preferred.
let scanReceiver = receiverName;
let componentScoped = false;
if (ref.language === 'cfml' || ref.language === 'cfscript') {
const scoped = receiverName.match(/^(variables|this|local|arguments)\.(.+)$/i);
if (scoped) {
scanReceiver = scoped[2]!;
const scope = scoped[1]!.toLowerCase();
componentScoped = scope === 'variables' || scope === 'this';
}
}
const patterns = localReceiverTypePatterns(
ref.language,
receiverName.replace(/[.*+?^${}()|[\]\\]/g, '\\$&'),
scanReceiver.replace(/[.*+?^${}()|[\]\\]/g, '\\$&'),
);
if (patterns.length === 0) return null;
@@ -1230,16 +1279,17 @@ function inferLocalReceiverType(
if (!lines || lines.length === 0) return null;
const callIdx = Math.max(0, Math.min(lines.length - 1, ref.line - 1));
const startIdx = Math.max(0, enclosingScopeStartLine(ref, context) - 1);
const startIdx = componentScoped
? 0
: Math.max(0, enclosingScopeStartLine(ref, context) - 1);
// Nearest declaration wins: scan backward from the call to the scope start.
for (let i = callIdx; i >= startIdx; i--) {
const matchLine = (i: number): string | null => {
const line = lines[i];
if (!line) continue;
if (!line) return null;
// A generated/minified line (one multi-KB statement) is not something a
// human-written local declaration lives on, and regexing it per ref is
// pure waste — skip it rather than scan it.
if (line.length > 10_000) continue;
if (line.length > 10_000) return null;
for (const re of patterns) {
const m = line.match(re);
if (m && m[1]) {
@@ -1247,6 +1297,23 @@ function inferLocalReceiverType(
if (type) return type;
}
}
return null;
};
// Nearest declaration wins: scan backward from the call to the scope start.
for (let i = callIdx; i >= startIdx; i--) {
const type = matchLine(i);
if (type) return type;
}
// A component-scoped field's declaration is position-independent — the
// `variables.svc = new X()` pseudoconstructor assignment or `property`
// declaration may sit BELOW the calling function in the file — so when the
// backward pass finds nothing, sweep the remainder of the file too.
if (componentScoped) {
for (let i = callIdx + 1; i < lines.length; i++) {
const type = matchLine(i);
if (type) return type;
}
}
return null;
}