Add nitpicker-style sticky PR comment bot (#8053)

## Summary
- Ports the logic of
[ethanis/nitpicker](https://github.com/ethanis/nitpicker) into the
existing torchci probot app: when `.github/nitpick.yml` is present, the
bot evaluates each rule's `pathFilter` against the PR's changed files
and posts a single sticky comment.
- Patterns use the upstream syntax: `+pat` includes, `-pat` excludes,
no-prefix is treated as include; matching is via `minimatch` (already a
dep). Comment is wrapped in `<!-- nitpick-bot-start -->` / `<!--
nitpick-bot-end -->` markers, updated on `pull_request.synchronize`, and
deleted when no rules match.
- Gated to `pytorch/pytorch` via `isPyTorchPyTorch` initially. Easy to
broaden to `isPyTorchbotSupportedOrg` later if other repos want to opt
in.

### Notes
- Config is fetched directly via `repos.getContent` rather than
`context.config()` because `@probot/octokit-plugin-config` deep-merges
as objects and silently flattens top-level YAML arrays (the upstream
nitpicker schema).
- `js-yaml` was already a transitive dep through probot; added it
explicitly to `torchci/package.json` along with `@types/js-yaml`.

### Example `.github/nitpick.yml`

```yaml
- markdown: |
    ## Did you update the docs?
    Please update docs when changing the public API.
  pathFilter:
    - "+torch/**"
    - "-torch/**/*.test.py"
```

## Test plan
- [x] `npx tsc --noEmit` clean
- [x] `npx jest test/nitpickBot.test.ts` — 12 new tests pass (parser,
file matcher, comment formation, plus probot integration covering: posts
on match, no-op when no match, exclude rules suppress matches, updates
existing comment, deletes stale comment when nothing matches, skips when
no `nitpick.yml`, skips for non-pytorch/pytorch repos)
- [ ] After merge, drop a `.github/nitpick.yml` into `pytorch/pytorch`
and observe behavior on a draft PR

Author: malfet

torchci/lib/bot/index.ts

@@ -7,6 +7,7 @@ import checkLabelsBot from "./checkLabelsBot";
 import ciflowPushTrigger from "./ciflowPushTrigger";
 import codevNoWritePerm from "./codevNoWritePermBot";
 import drciBot from "./drciBot";
+import nitpickBot from "./nitpickBot";
 import pytorchBot from "./pytorchBot";
 import retryBot from "./retryBot";
 import stripApprovalBot from "./stripApprovalBot";
@@ -22,6 +23,7 @@ export default function bot(app: Probot) {
   ciflowPushTrigger(app);
   codevNoWritePerm(app);
   drciBot(app);
+  nitpickBot(app);
   pytorchBot(app);
   retryBot(app);
   stripApprovalBot(app);

torchci/lib/bot/nitpickBot.ts

@@ -0,0 +1,208 @@
+import * as yaml from "js-yaml";
+import { minimatch } from "minimatch";
+import { Context, Probot } from "probot";
+import { getFilesChangedByPr, isPyTorchPyTorch } from "./utils";
+
+// Implements logic similar to https://github.com/ethanis/nitpicker.
+// Reads `.github/nitpick.yml` from the repo's default branch and posts
+// (or updates) a single sticky comment on PRs whose changed files match
+// any of the configured rules.
+//
+// Config format (top-level YAML list):
+//
+//   - markdown: |
+//       ## Did you update the migration?
+//       Reminder text shown to the PR author.
+//     pathFilter:
+//       - "+migrations/**"
+//       - "-migrations/test/**"
+//
+// Each pattern is evaluated with `minimatch`. A pattern beginning with `+`
+// (or with no prefix) is an include pattern; a pattern beginning with `-`
+// is an exclude pattern. A file matches the rule if it matches any include
+// pattern and no exclude pattern. The rule fires (its `markdown` is added
+// to the comment) when at least one changed file matches.
+
+export const NITPICK_COMMENT_START = "<!-- nitpick-bot-start -->";
+export const NITPICK_COMMENT_END = "<!-- nitpick-bot-end -->";
+export const NITPICK_CONFIG_PATH = ".github/nitpick.yml";
+
+export interface NitpickRule {
+  markdown: string;
+  pathFilter: string[];
+}
+
+export function parseNitpickConfig(text: string): NitpickRule[] {
+  const parsed = yaml.load(text);
+  if (!Array.isArray(parsed)) {
+    return [];
+  }
+  const rules: NitpickRule[] = [];
+  for (const entry of parsed) {
+    if (
+      entry &&
+      typeof entry.markdown === "string" &&
+      Array.isArray(entry.pathFilter)
+    ) {
+      rules.push({
+        markdown: entry.markdown,
+        pathFilter: entry.pathFilter.filter((p: any) => typeof p === "string"),
+      });
+    }
+  }
+  return rules;
+}
+
+export function fileMatchesRule(file: string, rule: NitpickRule): boolean {
+  let included = false;
+  for (const pat of rule.pathFilter) {
+    if (pat.startsWith("-")) {
+      if (minimatch(file, pat.slice(1))) {
+        return false;
+      }
+    } else {
+      const glob = pat.startsWith("+") ? pat.slice(1) : pat;
+      if (minimatch(file, glob)) {
+        included = true;
+      }
+    }
+  }
+  return included;
+}
+
+export function getMatchingRules(
+  files: string[],
+  rules: NitpickRule[]
+): NitpickRule[] {
+  return rules.filter((rule) =>
+    files.some((file) => fileMatchesRule(file, rule))
+  );
+}
+
+export function formNitpickComment(rules: NitpickRule[]): string {
+  if (rules.length === 0) {
+    return "";
+  }
+  const body = rules.map((r) => r.markdown.trim()).join("\n\n---\n\n");
+  return `${NITPICK_COMMENT_START}\n${body}\n${NITPICK_COMMENT_END}`;
+}
+
+async function findExistingNitpickComment(
+  context: Context<"pull_request">,
+  owner: string,
+  repo: string,
+  prNum: number
+): Promise<{ id: number; body: string }> {
+  const res = await context.octokit.issues.listComments({
+    owner,
+    repo,
+    issue_number: prNum,
+  });
+  for (const c of res.data) {
+    if (c.body && c.body.includes(NITPICK_COMMENT_START)) {
+      return { id: c.id, body: c.body };
+    }
+  }
+  return { id: 0, body: "" };
+}
+
+async function loadNitpickConfig(
+  context: Context<"pull_request">,
+  owner: string,
+  repo: string
+): Promise<NitpickRule[] | null> {
+  try {
+    const res = await context.octokit.repos.getContent({
+      owner,
+      repo,
+      path: NITPICK_CONFIG_PATH,
+    });
+    const data = res.data as { content?: string; encoding?: string };
+    if (!data.content) {
+      return null;
+    }
+    const text = Buffer.from(
+      data.content,
+      (data.encoding as BufferEncoding) ?? "base64"
+    ).toString("utf8");
+    return parseNitpickConfig(text);
+  } catch (err: any) {
+    if (err.status === 404) {
+      return null;
+    }
+    throw err;
+  }
+}
+
+export default function nitpickBot(app: Probot): void {
+  app.on(
+    [
+      "pull_request.opened",
+      "pull_request.reopened",
+      "pull_request.synchronize",
+    ],
+    async (context) => {
+      const owner = context.payload.repository.owner.login;
+      const repo = context.payload.repository.name;
+      // Limit to pytorch/pytorch initially.
+      if (!isPyTorchPyTorch(owner, repo)) {
+        context.log(
+          `${__filename} only runs on pytorch/pytorch (got ${owner}/${repo})`
+        );
+        return;
+      }
+      if (context.payload.pull_request.state !== "open") {
+        return;
+      }
+      const prNum = context.payload.pull_request.number;
+
+      const rules = await loadNitpickConfig(context, owner, repo);
+      if (rules == null) {
+        context.log(`${NITPICK_CONFIG_PATH} not found, skipping`);
+        return;
+      }
+
+      const filesChanged = await getFilesChangedByPr(
+        context.octokit,
+        owner,
+        repo,
+        prNum
+      );
+      const matched = getMatchingRules(filesChanged, rules);
+      const newBody = formNitpickComment(matched);
+      const existing = await findExistingNitpickComment(
+        context,
+        owner,
+        repo,
+        prNum
+      );
+
+      if (newBody === "") {
+        if (existing.id !== 0) {
+          await context.octokit.issues.deleteComment({
+            owner,
+            repo,
+            comment_id: existing.id,
+          });
+        }
+        return;
+      }
+
+      if (existing.id === 0) {
+        await context.octokit.issues.createComment({
+          owner,
+          repo,
+          issue_number: prNum,
+          body: newBody,
+        });
+      } else if (existing.body !== newBody) {
+        await context.octokit.issues.updateComment({
+          owner,
+          repo,
+          comment_id: existing.id,
+          body: newBody,
+        });
+      }
+    }
+  );
+}

torchci/package.json

@@ -45,6 +45,7 @@
     "echarts-for-react": "^3.0.2",
     "eslint-plugin-unused-imports": "^3.2.0",
     "jaro-winkler-typescript": "^1.0.1",
+    "js-yaml": "^4.1.0",
     "lodash": "^4.17.21",
     "lz-string": "^1.5.0",
     "minimatch": "^9.0.3",
@@ -77,6 +78,7 @@
     "@types/d3": "^7.4.3",
     "@types/echarts": "^4.9.14",
     "@types/jest": "^29.5.14",
+    "@types/js-yaml": "^4.0.5",
     "@types/jsdom": "^16.2.14",
     "@types/lodash": "^4.14.182",
     "@types/node": "^17.0.40",