bc0a076e13
Fixes #7514 — the prefix-validation piece of the #5456 auto-linker 404-storm umbrella. The remaining pieces (404 retry guard, word-boundary tightening, comment edit/soft-delete) stay open under #5456. Part of #5456. ## Thinking Path > - Paperclip orchestrates AI agents for zero-human companies > - Agents and humans cross-reference work in markdown — issue descriptions, comments, documents — which the UI renders through a shared `MarkdownBody` > - That renderer auto-links any `IDENT-123`-shaped token to an internal `/issues/IDENT-123` link > - But foreign tracker keys share that exact shape: a Jira `TREE-604` (or any external `ORG-123`) mentioned in prose becomes a link to a Paperclip issue that does not exist — it 404s, and the renderer also fires a wasted issue-fetch for the bogus identifier > - The set of real issue prefixes is already in the browser: every company carries an `issuePrefix`, exposed via `CompanyContext` > - This PR gates bare-token auto-linking to known company prefixes, leaving explicit `issue://` / `/issues/` references and real markdown links untouched > - The benefit is no dead internal links from foreign keys, with no new query, cache, or server change — and zero regression when prefixes aren't yet known ## What Changed - **`ui/src/lib/issue-reference.ts`** — `parseIssueReferenceFromHref` takes an optional `knownPrefixes` set and rejects a bare `IDENT-123` token whose prefix isn't in it; threaded through `remarkLinkIssueReferences(options)` → tree rewrite → text and inline-code paths. An omitted/empty set keeps the legacy permissive behavior. Explicit `issue://` scheme and `/issues/` path forms are never gated. - **`ui/src/context/CompanyContext.tsx`** — adds `useOptionalCompany()`, a non-throwing variant of `useCompany()` (returns `null` outside a provider). - **`ui/src/components/MarkdownBody.tsx`** — reads company prefixes via `useOptionalCompany()` and passes them to the linkifier. The non-throwing read keeps `MarkdownBody` renderable in provider-less surfaces (e.g. standalone/exported markdown). - Tests extended in `issue-reference.test.ts` (gating + remark-plugin cases) and `MarkdownBody.test.tsx` (gating, empty-companies permissive, explicit-path bypass). ## Verification - `pnpm --filter @paperclipai/ui exec vitest run src/lib/issue-reference.test.ts src/components/MarkdownBody.test.tsx` — green (17 + 40 tests). - Full UI suite: `pnpm --filter @paperclipai/ui exec vitest run` — **1161 passed / 183 files**; pre-existing `MarkdownBody` link tests pass unmodified (they hit the permissive `null`-context path), confirming no regression. - `pnpm --filter @paperclipai/ui run typecheck` — clean. - _Screenshots pending — opening as draft; before/after images to follow before marking ready._ - Manual (before/after): in an issue description containing both a real Paperclip identifier and a Jira key — _before_ both render as `/issues/...` links (the Jira one dead); _after_ only the real identifier links and the Jira key is plain text. ## Risks - **Low risk.** No server/API/migration change; pure client rendering logic. - A referenced issue whose company isn't in the viewer's `companies` list stops auto-linking — acceptable, since that internal link wouldn't resolve for that viewer anyway; explicit `/issues/IDENT` references still render. - During initial load (companies not yet fetched) behavior is identical to today (permissive), so no new flicker. ## Model Used - **Anthropic Claude Opus 4.8** (`claude-opus-4-8`), 1M-token context, extended thinking + tool use. Plan authored and implemented with the model; all decisions reviewed by the human contributor. ## Checklist - [x] I have included a thinking path that traces from project context to this change - [x] I have specified the model used (with version and capability details) - [x] I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work - [x] I have run tests locally and they pass - [x] I have added or updated tests where applicable - [ ] If this change affects the UI, I have included before/after screenshots - [x] I have updated relevant documentation to reflect my changes (no doc changes required — behavior gated, no public API/doc surface affected) - [x] I have considered and documented any risks above - [x] I will address all Greptile and reviewer comments before requesting merge --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
180 lines
6.1 KiB
TypeScript
180 lines
6.1 KiB
TypeScript
type MarkdownNode = {
|
|
type: string;
|
|
value?: string;
|
|
url?: string;
|
|
children?: MarkdownNode[];
|
|
};
|
|
|
|
const BARE_ISSUE_IDENTIFIER_RE = /^[A-Z][A-Z0-9]*-\d+$/i;
|
|
const ISSUE_SCHEME_RE = /^issue:\/\/:?([^?#\s]+)(?:[?#].*)?$/i;
|
|
const ISSUE_REFERENCE_TOKEN_RE = /issue:\/\/:?[^\s<>()]+|https?:\/\/[^\s<>()]+|\/(?:[^\s<>()/]+\/)*issues\/[A-Z][A-Z0-9]*-\d+(?=$|[\s<>)\],.;!?:])|\b[A-Z][A-Z0-9]*-\d+\b/gi;
|
|
|
|
export function parseIssuePathIdFromPath(pathOrUrl: string | null | undefined): string | null {
|
|
if (!pathOrUrl) return null;
|
|
const pathname = pathOrUrl.trim();
|
|
if (!pathname) return null;
|
|
if (/^https?:\/\//i.test(pathname)) return null;
|
|
|
|
const segments = pathname.split("/").filter(Boolean);
|
|
const issueIndex = segments.findIndex((segment) => segment === "issues");
|
|
if (issueIndex === -1 || issueIndex === segments.length - 1) return null;
|
|
const issuePathId = decodeURIComponent(segments[issueIndex + 1] ?? "");
|
|
if (!issuePathId || issuePathId.startsWith(":")) return null;
|
|
return BARE_ISSUE_IDENTIFIER_RE.test(issuePathId) ? issuePathId.toUpperCase() : issuePathId;
|
|
}
|
|
|
|
export function parseIssueReferenceFromHref(
|
|
href: string | null | undefined,
|
|
knownPrefixes?: Set<string>,
|
|
) {
|
|
if (!href) return null;
|
|
const trimmed = href.trim();
|
|
const issueSchemeMatch = trimmed.match(ISSUE_SCHEME_RE);
|
|
if (issueSchemeMatch?.[1]) {
|
|
const issuePathId = decodeURIComponent(issueSchemeMatch[1]);
|
|
return {
|
|
issuePathId,
|
|
href: `/issues/${encodeURIComponent(issuePathId)}`,
|
|
};
|
|
}
|
|
|
|
const pathId = parseIssuePathIdFromPath(href);
|
|
if (pathId) {
|
|
return {
|
|
issuePathId: pathId,
|
|
href: `/issues/${encodeURIComponent(pathId)}`,
|
|
};
|
|
}
|
|
|
|
if (!BARE_ISSUE_IDENTIFIER_RE.test(trimmed)) return null;
|
|
const normalized = trimmed.toUpperCase();
|
|
// Only auto-link a bare IDENT-123 token when its prefix belongs to a known
|
|
// company. An empty or omitted set means "prefixes unknown" -> stay permissive
|
|
// (preserves behavior during initial load and in provider-less render
|
|
// surfaces). The issue:// scheme and /issues/ path forms handled above are
|
|
// never gated -- they are deliberate references, not accidental foreign keys.
|
|
if (knownPrefixes && knownPrefixes.size > 0) {
|
|
const prefix = normalized.split("-")[0];
|
|
if (!prefix || !knownPrefixes.has(prefix)) return null;
|
|
}
|
|
return {
|
|
issuePathId: normalized,
|
|
href: `/issues/${encodeURIComponent(normalized)}`,
|
|
};
|
|
}
|
|
|
|
function splitTrailingPunctuation(token: string) {
|
|
let core = token;
|
|
let trailing = "";
|
|
|
|
while (core.length > 0) {
|
|
const lastChar = core.at(-1);
|
|
if (!lastChar || !/[),.;!?:\]]/.test(lastChar)) break;
|
|
if (lastChar === ")") {
|
|
const openCount = (core.match(/\(/g) ?? []).length;
|
|
const closeCount = (core.match(/\)/g) ?? []).length;
|
|
if (closeCount <= openCount) break;
|
|
}
|
|
if (lastChar === "]") {
|
|
const openCount = (core.match(/\[/g) ?? []).length;
|
|
const closeCount = (core.match(/\]/g) ?? []).length;
|
|
if (closeCount <= openCount) break;
|
|
}
|
|
trailing = `${lastChar}${trailing}`;
|
|
core = core.slice(0, -1);
|
|
}
|
|
|
|
return { core, trailing };
|
|
}
|
|
|
|
function createIssueLinkNode(value: string, href: string, childType: "text" | "inlineCode" = "text"): MarkdownNode {
|
|
return {
|
|
type: "link",
|
|
url: href,
|
|
children: [{ type: childType, value }],
|
|
};
|
|
}
|
|
|
|
function linkifyIssueReferencesInText(value: string, knownPrefixes?: Set<string>): MarkdownNode[] | null {
|
|
const nodes: MarkdownNode[] = [];
|
|
let cursor = 0;
|
|
let matched = false;
|
|
|
|
for (const match of value.matchAll(ISSUE_REFERENCE_TOKEN_RE)) {
|
|
const raw = match[0];
|
|
if (!raw) continue;
|
|
|
|
const start = match.index ?? 0;
|
|
const end = start + raw.length;
|
|
const { core, trailing } = splitTrailingPunctuation(raw);
|
|
const issueRef = parseIssueReferenceFromHref(core, knownPrefixes);
|
|
if (!issueRef) continue;
|
|
|
|
matched = true;
|
|
if (start > cursor) {
|
|
nodes.push({ type: "text", value: value.slice(cursor, start) });
|
|
}
|
|
nodes.push(createIssueLinkNode(core, issueRef.href));
|
|
if (trailing) {
|
|
nodes.push({ type: "text", value: trailing });
|
|
}
|
|
cursor = end;
|
|
}
|
|
|
|
if (!matched) return null;
|
|
if (cursor < value.length) {
|
|
nodes.push({ type: "text", value: value.slice(cursor) });
|
|
}
|
|
return nodes;
|
|
}
|
|
|
|
function rewriteMarkdownTree(node: MarkdownNode, knownPrefixes?: Set<string>) {
|
|
if (!Array.isArray(node.children) || node.children.length === 0) return;
|
|
if (node.type === "link" || node.type === "linkReference" || node.type === "code" || node.type === "definition" || node.type === "html") {
|
|
return;
|
|
}
|
|
|
|
const nextChildren: MarkdownNode[] = [];
|
|
for (const child of node.children) {
|
|
if (child.type === "inlineCode" && typeof child.value === "string") {
|
|
const issueRef = parseIssueReferenceFromHref(child.value, knownPrefixes);
|
|
if (issueRef) {
|
|
nextChildren.push(createIssueLinkNode(child.value, issueRef.href, "inlineCode"));
|
|
continue;
|
|
}
|
|
}
|
|
|
|
if (child.type === "text" && typeof child.value === "string") {
|
|
const linked = linkifyIssueReferencesInText(child.value, knownPrefixes);
|
|
if (linked) {
|
|
nextChildren.push(...linked);
|
|
continue;
|
|
}
|
|
}
|
|
|
|
rewriteMarkdownTree(child, knownPrefixes);
|
|
nextChildren.push(child);
|
|
}
|
|
node.children = nextChildren;
|
|
}
|
|
|
|
export interface RemarkLinkIssueReferencesOptions {
|
|
/**
|
|
* Company issue prefixes that are eligible for auto-linking. When provided
|
|
* and non-empty, a bare IDENT-123 token only becomes an issue link if its
|
|
* prefix is in this set -- this keeps foreign tracker keys (e.g. a Jira
|
|
* "TREE-604") from linking to non-existent Paperclip issues. Omit or pass an
|
|
* empty list to keep the legacy permissive behavior (link any IDENT-123).
|
|
*/
|
|
knownPrefixes?: string[];
|
|
}
|
|
|
|
export function remarkLinkIssueReferences(options?: RemarkLinkIssueReferencesOptions) {
|
|
const knownPrefixes = options?.knownPrefixes && options.knownPrefixes.length > 0
|
|
? new Set(options.knownPrefixes.map((prefix) => prefix.toUpperCase()))
|
|
: undefined;
|
|
return (tree: MarkdownNode) => {
|
|
rewriteMarkdownTree(tree, knownPrefixes);
|
|
};
|
|
}
|