\\n>

⚙️ Run configuration

\\n> \\n> **Configuration used**: defaults\\n> \\n> **Review profile**: CHILL\\n> \\n> **Plan**: Pro Plus\\n> \\n> **Run ID**: `df2270c5-c17a-4286-bf3b-5edf54ef1102`\\n> \\n>

\\n> \\n>

\\n>

📥 Commits

\\n> \\n> Reviewing files that changed from the base of the PR and between 07e6c31d9c768072427ff510a1daadc5c68cb1c0 and 57846a0075effbc28aab0e63e1da5f76b51ec501.\\n> \\n>

\\n> \\n>

\\n>

📒 Files selected for processing (7)

\\n> \\n> * `docs/leakfix-mine.md`\\n> * `leakmine/bigquery.py`\\n> * `leakmine/cli.py`\\n> * `leakmine/collect.py`\\n> * `leakmine/mine.py`\\n> * `leakmine/signals.py`\\n> * `leakmine/tests/test_leakmine.py`\\n> \\n>

\\n> \\n> ```ascii\\n> ___________________________________________________________\\n> < Veni, Vidi, Codici Fixi. I came, I saw, I fixed the code. >\\n> -----------------------------------------------------------\\n> \\\\\\n> \\\\ (\\\\__/)\\n> (•ㅅ•)\\n> / 　づ\\n> ```\\n\\n\\n\\n\\n\\n

\\n

✨ Finishing Touches

\\n\\n

\\n

📝 Generate docstrings

\\n\\n- [ ] Create stacked PR\\n- [ ] Commit on current branch\\n\\n

\\n

🧪 Generate unit tests (beta)

\\n\\n- [ ] Create PR with unit tests\\n- [ ] Commit unit tests in branch `claude/own-language-market-research-jgrrgm`\\n\\n

\\n\\n

\\n\\n\\n\\n\\n---\\n\\n\\n\\n\\n_{Comment `@coderabbitai help` to get the list of available commands.}\\n\\n\",\"path\":null,\"side\":null,\"line\":null}{\"type\":\"pull_request_comment\",\"body\":\"**_{_{![P2 Badge](https://img.shields.io/badge/P2-yellow?style=flat)}} Do not persist unknown languages after GraphQL failures**\\n\\nWhen the GraphQL request fails for a whole batch (for example missing/expired `GITHUB_TOKEN`, rate limiting, or a transient 5xx), this handler turns the failure into an empty `data` object, so `enrich_languages` records every repo in that batch as `classifier='language'`/`unknown`. Because the enrichment query skips candidates that already have a language label, a later rerun with a valid token will not retry those repos and `classify-store` will still fetch all their diffs, defeating the new budget-saving stage until the labels are manually deleted.\\n\\nUseful? React with 👍 / 👎.\",\"path\":\"leakmine/collect.py\",\"side\":\"RIGHT\",\"line\":null}{\"type\":\"pull_request_comment\",\"body\":\"\\n### 💡 Codex Review\\n\\nHere are some automated review suggestions for this pull request.\\n\\n**Reviewed commit:** `57846a0075`\\n \\n\\n

ℹ️ About Codex in GitHub

\\n
\\n\\n[Your team has set up Codex to review pull requests in this repo](https://chatgpt.com/codex/cloud/settings/general). Reviews are triggered when you\\n- Open a pull request for review\\n- Mark a draft as ready\\n- Comment \\\"@codex review\\\".\\n\\nIf Codex has suggestions, it will comment; otherwise it will react with 👍.\\n\\n\\n\\n\\nCodex can also answer questions or update the PR. Try commenting \\\"@codex address that feedback\\\".\\n \\n

\",\"path\":null,\"side\":null,\"line\":null}","text_elements",[],{"_398":601,"_379":380,"_400":386},{"_401":602,"_403":-5,"_42":-5,"_404":-5,"_405":-5,"_96":326,"_92":407},[603],{"_46":597,"_92":604},"input_text",{"_398":606,"_379":380,"_400":386},{"_401":607,"_403":-5,"_42":-5,"_404":-5,"_405":-5,"_96":326,"_92":407},[608],{"_46":609,"_92":604},"\n /workspace/OwnAudit\n bash\n 2026-06-28\n Etc/UTC\n",{"_398":611,"_379":380,"_400":386},{"_401":612,"_403":-5,"_42":-5,"_404":-5,"_405":-5,"_96":613,"_92":407},[614,615],"developer",{"_46":617,"_92":604},{"_46":616,"_92":604},"\n## Skills\nA skill is a set of local instructions to follow that is stored in a `SKILL.md` file. Below is the list of skills that can be used. Each entry includes a name, description, and file path so you can open the source for full instructions when using a specific skill.\n### Available skills\n- imagegen: Generate or edit raster images when the task benefits from AI-created bitmap visuals such as photos, illustrations, textures, sprites, mockups, or transparent-background cutouts. Use when Codex should create a brand-new image, transform an existing image, or derive visual variants from references, and the output should be a bitmap asset rather than repo-native code or vector. Do not use when the task is better handled by editing existing SVG/vector/code-native assets, extending an established icon or logo system, or building the visual directly in HTML/CSS/canvas. (file: /opt/codex/skills/.system/imagegen/SKILL.md)\n- openai-docs: Use when the user asks how to build with OpenAI products or APIs and needs up-to-date official documentation with citations, help choosing the latest model for a use case, or model upgrade and prompt-upgrade guidance; prioritize OpenAI docs MCP tools, use bundled references only as helper context, and restrict any fallback browsing to official OpenAI domains. (file: /opt/codex/skills/.system/openai-docs/SKILL.md)\n- plugin-creator: Create and scaffold plugin directories for Codex with a required `.codex-plugin/plugin.json`, optional plugin folders/files, and baseline placeholders you can edit before publishing or testing. Use when Codex needs to create a new local plugin, add optional plugin structure, or generate or update repo-root `.agents/plugins/marketplace.json` entries for plugin ordering and availability metadata. (file: /opt/codex/skills/.system/plugin-creator/SKILL.md)\n- skill-creator: Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Codex's capabilities with specialized knowledge, workflows, or tool integrations. (file: /opt/codex/skills/.system/skill-creator/SKILL.md)\n- skill-installer: Install Codex skills into $CODEX_HOME/skills from a curated list or a GitHub repo path. Use when a user asks to list installable skills, install a curated skill, or install a skill from another repo (including private repos). (file: /opt/codex/skills/.system/skill-installer/SKILL.md)\n### How to use skills\n- Discovery: The list above is the skills available in this session (name + description + file path). Skill bodies live on disk at the listed paths.\n- Trigger rules: If the user names a skill (with `$SkillName` or plain text) OR the task clearly matches a skill's description shown above, you must use that skill for that turn. Multiple mentions mean use them all. Do not carry skills across turns unless re-mentioned.\n- Missing/blocked: If a named skill isn't in the list or the path can't be read, say so briefly and continue with the best fallback.\n- How to use a skill (progressive disclosure):\n 1) After deciding to use a skill, open its `SKILL.md`. Read only enough to follow the workflow.\n 2) When `SKILL.md` references relative paths (e.g., `scripts/foo.py`), resolve them relative to the skill directory listed above first, and only consider other paths if needed.\n 3) If `SKILL.md` points to extra folders such as `references/`, load only the specific files needed for the request; don't bulk-load everything.\n 4) If `scripts/` exist, prefer running or patching them instead of retyping large code blocks.\n 5) If `assets/` or templates exist, reuse them instead of recreating from scratch.\n- Coordination and sequencing:\n - If multiple skills apply, choose the minimal set that covers the request and state the order you'll use them.\n - Announce which skill(s) you're using and why (one short line). If you skip an obvious skill, say why.\n- Context hygiene:\n - Keep context small: summarize long sections instead of pasting them; only load extra files when needed.\n - Avoid deep reference-chasing: prefer opening only files directly linked from `SKILL.md` unless you're blocked.\n - When variants exist (frameworks, providers, domains), pick only the relevant reference file(s) and note that choice.\n- Safety and fallback: If a skill can't be applied cleanly (missing files, unclear instructions), state the issue, pick the next-best approach, and continue.\n","\nFilesystem sandboxing defines which files can be read or written. `sandbox_mode` is `danger-full-access`: No filesystem sandboxing - all commands are permitted. Network access is enabled.\nApproval policy is currently never. Do not provide the `sandbox_permissions` for any reason, commands will be rejected.\n","turn/started",{"_379":380,"_87":620},{"_103":-5,"_42":386,"_387":621,"_389":390,"_391":622},[],"inProgress",{"_391":624,"_379":380},{"_625":626,"_92":627},"activeFlags",[],"active","thread/started",{"_630":631},"thread",{"_632":-5,"_633":-5,"_634":635,"_636":390,"_446":447,"_637":15,"_638":-5,"_639":-5,"_640":-5,"_641":642,"_42":380,"_643":644,"_181":-5,"_645":-5,"_161":646,"_647":199,"_648":-5,"_451":649,"_391":650,"_651":-5,"_77":652,"_653":390},"agentNickname","agentRole","cliVersion","0.128.0","createdAt","ephemeral","extra","forkedFromId","gitInfo","historyMode","legacy","modelProvider","openai","parentThreadId","/opt/codex/sessions/2026/06/28/rollout-2026-06-28T06-06-26-019f0cd6-1bcf-7da1-81ed-6ac0ddd20ecf.jsonl","preview","recencyAt","vscode",{"_92":395},"threadSource",[],"updatedAt",[],"messages",[657],{"_42":658,"_659":660,"_661":662,"_663":662,"_401":664,"_391":665,"_403":-5,"_666":32,"_667":668,"_669":670,"_671":-5},"a5f7674a-c84a-40ed-97f8-c8c938e1aa69","author",{"_96":326,"_181":-5,"_667":680},"create_time",1782626785.429741,"update_time",{"_676":46,"_677":678},"finished_successfully","weight","metadata",{"_672":673},"recipient","all","channel","l1239dk1",{"_674":675},"inline_comments",[],"content_type","parts",[679],"Good catch @chatgpt-codex-connector — fixed in 11da9fd. A whole-batch GraphQL failure (bad/expired token, rate limit, 5xx, or an errors-only response with no `data`) is now **omitted** from `fetch_repo_languages`'s result map instead of being coerced to `unknown`. `enrich_languages` only labels repos actually present in the map, so failed ones stay unenriched and a later run retries them — no more silently persisting `unknown` and skipping them forever. Genuine per-repo nulls (repo resolved, no `primaryLanguage`) are still recorded as `\"\"` so `classify-store` still attempts them.\n\nAdded two tests: failed/errors-only batch → empty map; and enrich records nothing on failure then succeeds on a subsequent working fetch. 36 tests pass under `python3` and `-O`.\n\n---\n_Generated by [Claude Code](https://claude.ai/code)_\nleakmine: language enrichment (batched GraphQL) to cut cross-language diff waste\n\n## What\n\nLanguage **enrichment** (hydration/saturation) to fix the cross-language waste the real ClickHouse verdict run exposed: it fetched **284 diffs to keep 13** (~95% wasted) because ClickHouse `github_events` has no repo-language column, so discovery was language-blind.\n\nThis adds a stage that hydrates the missing language and pre-filters *before* the expensive diff fetch.\n\n## How\n\n- **`collect.fetch_repo_languages`** — batched **GraphQL** (~100 repos per request, aliased `repository(...)` fields) → primary language. This is the key to it actually saving budget: language costs ~3 calls per ~300 candidates, versus one REST `/repos` call per repo, which would cost the *same* as the diff it's trying to save (so naive per-repo enrichment would be a wash).\n- **`mine.enrich_languages` + `enrich-store` CLI** — writes a `classifier='language'` label per candidate, and for repos whose language isn't in `signals.LANGS`, also a `classifier='patch'` `wrong-language` label. `classify-store`'s existing resume anti-join already skips `classifier='patch'` rows, so **off-language PRs are never diff-fetched**. Resumable; commits per repo.\n- `signals.LANGS` is now the canonical ecosystem→languages map (`bigquery` reuses it).\n\nPipeline: `bq-ingest` → **`enrich-store`** → `classify-store`.\n\n```\npython3 -m leakmine.cli enrich-store --store corpus.db --ecosystem react_ts # batched GraphQL\npython3 -m leakmine.cli classify-store --store corpus.db --ecosystem react_ts # only matching-lang diffs\n```\n\n## Economics (honest)\n\nThe win is real **because GraphQL batches**: enrichment turns \"fetch 284 diffs, keep 13\" into \"~3 GraphQL calls + diff-fetch only the JS/TS repos.\" It also makes the corpus language-correct regardless. For a *truly* free language filter at discovery there are still two better-when-available options (noted in §14): the BigQuery GH-Archive path (`base.repo.language`, already filtered server-side) and the Search-API `mine` path (`language:` qualifier) — both need no per-repo enrichment, but the first needs GCP and the second is capped at 1000/query.\n\n## Tests / CI\n\n35 tests under `python3` and `python3 -O` (+2: GraphQL parse incl. null/missing repos; end-to-end proof that an off-language repo is marked and then **not** diff-fetched by `classify-store`). Docs §14 updated.\n\n🤖 Generated with [Claude Code](https://claude.com/claude-code)\n\n---\n_Generated by [Claude Code](https://claude.ai/code/session_01W6Lbeeoqy7v3wjzV43iLun)_\n\n\n\n## Summary by CodeRabbit\n\n* **New Features**\n * Added a new language enrichment step to the mining pipeline.\n * Added a CLI command to run language enrichment and generate a summary report.\n * Added batched repository language lookup support for GitHub data.\n\n* **Bug Fixes**\n * Repositories detected as the wrong language are now skipped before patch diff fetching, reducing unnecessary work.\n * Language filtering now uses ecosystem-specific language values consistently.\n\n* **Tests**\n * Added coverage for batched language lookup and the new enrichment flow.\n\n\n\n\n\n[![Review Change Stack](https://storage.googleapis.com/coderabbit_public_assets/review-stack-in-coderabbit-ui.svg)](https://app.coderabbit.ai/change-stack/PhysShell/OwnAudit/pull/28?utm_source=github_walkthrough&utm_medium=github&utm_campaign=change_stack)\n\n\n\n\n> [!NOTE]\n> Currently processing new changes in this PR. This may take a few minutes, please wait...\n> \n>

\n>

⚙️ Run configuration

\n> \n> **Configuration used**: defaults\n> \n> **Review profile**: CHILL\n> \n> **Plan**: Pro Plus\n> \n> **Run ID**: `df2270c5-c17a-4286-bf3b-5edf54ef1102`\n> \n>

\n> \n>

\n>

📥 Commits

\n> \n> Reviewing files that changed from the base of the PR and between 07e6c31d9c768072427ff510a1daadc5c68cb1c0 and 57846a0075effbc28aab0e63e1da5f76b51ec501.\n> \n>

\n> \n>

\n>

📒 Files selected for processing (7)

\n> \n> * `docs/leakfix-mine.md`\n> * `leakmine/bigquery.py`\n> * `leakmine/cli.py`\n> * `leakmine/collect.py`\n> * `leakmine/mine.py`\n> * `leakmine/signals.py`\n> * `leakmine/tests/test_leakmine.py`\n> \n>

\n> \n> ```ascii\n> ___________________________________________________________\n> < Veni, Vidi, Codici Fixi. I came, I saw, I fixed the code. >\n> -----------------------------------------------------------\n> \\\n> \\ (\\__/)\n> (•ㅅ•)\n> / 　づ\n> ```\n\n\n\n\n\n

✨ Finishing Touches

\n\n

📝 Generate docstrings

\n\n- [ ] Create stacked PR\n- [ ] Commit on current branch\n\n

🧪 Generate unit tests (beta)

\n\n- [ ] Create PR with unit tests\n- [ ] Commit unit tests in branch `claude/own-language-market-research-jgrrgm`\n\n

\n\n

\n\n\n\n\n---\n\n\n\n\n_{Comment `@coderabbitai help` to get the list of available commands.}\n\n\n**_{_{![P2 Badge](https://img.shields.io/badge/P2-yellow?style=flat)}} Do not persist unknown languages after GraphQL failures**\n\nWhen the GraphQL request fails for a whole batch (for example missing/expired `GITHUB_TOKEN`, rate limiting, or a transient 5xx), this handler turns the failure into an empty `data` object, so `enrich_languages` records every repo in that batch as `classifier='language'`/`unknown`. Because the enrichment query skips candidates that already have a language label, a later rerun with a valid token will not retry those repos and `classify-store` will still fetch all their diffs, defeating the new budget-saving stage until the labels are manually deleted.\n\nUseful? React with 👍 / 👎.\n\n### 💡 Codex Review\n\nHere are some automated review suggestions for this pull request.\n\n**Reviewed commit:** `57846a0075`\n \n\n

ℹ️ About Codex in GitHub

\n
\n\n[Your team has set up Codex to review pull requests in this repo](https://chatgpt.com/codex/cloud/settings/general). Reviews are triggered when you\n- Open a pull request for review\n- Mark a draft as ready\n- Comment \"@codex review\".\n\nIf Codex has suggestions, it will comment; otherwise it will react with 👍.\n\n\n\n\nCodex can also answer questions or update the PR. Try commenting \"@codex address that feedback\".\n \n

",{},"created",1782626827.6121879,"Completing the task","event_type","info",{"_92":407,"_96":93,"_401":1559},{"_92":688,"_689":690},"partial_repo_snapshot","files",[691,692,693],{"_161":1409,"_695":-5,"_696":1410,"_698":-5},{"_161":1195,"_695":-5,"_696":1196,"_698":-5},{"_161":694,"_695":-5,"_696":697,"_698":-5},"leakmine/tests/test_leakmine.py","contents","line_range_contents",[699],"missing_reason",{"_700":522,"_701":702,"_401":703,"_704":59},"line_range_start","line_range_end",648,[705,199,706,199,707,708,709,710,711,712,713,714,715,716,717,199,718,719,720,199,721,722,723,199,199,724,725,726,199,199,727,199,728,729,730,731,732,733,734,735,736,737,738,739,740,741,711,199,199,742,743,744,745,746,747,748,749,750,751,752,199,199,753,754,755,756,757,758,759,760,761,199,199,762,763,764,765,766,758,767,768,199,199,769,199,770,771,772,773,774,775,199,199,776,777,778,779,780,781,199,199,782,783,784,785,786,773,787,199,199,788,789,790,791,792,793,794,199,199,795,199,796,797,199,199,798,799,800,801,199,802,803,804,805,199,199,806,807,808,809,810,811,199,812,813,814,815,816,817,199,199,818,819,820,821,822,823,734,735,824,825,741,711,199,199,826,827,828,829,830,831,832,833,834,199,199,835,199,836,837,838,839,840,841,842,199,199,843,844,845,846,847,848,849,850,851,852,853,854,855,856,857,858,848,859,860,199,861,862,863,864,865,199,199,866,199,867,868,869,870,871,872,873,874,875,876,877,878,879,880,199,199,881,882,883,884,885,886,869,887,888,889,874,890,891,892,893,199,199,894,895,896,897,898,899,900,901,874,902,903,904,905,900,906,874,907,199,199,908,909,869,910,911,874,875,912,913,914,199,199,915,199,916,917,918,919,920,921,922,923,924,925,926,927,928,929,930,931,932,933,934,935,199,199,936,199,937,938,939,940,941,942,943,944,945,199,946,947,948,949,950,951,952,953,954,955,956,199,199,957,958,959,947,948,960,952,961,962,199,199,963,199,964,965,966,967,199,968,969,970,971,199,199,972,973,974,975,976,977,978,979,199,199,980,981,982,983,984,985,986,987,988,989,990,991,992,993,199,199,994,995,996,997,998,999,926,1000,1001,1002,199,1003,1004,199,1005,1006,199,1007,1008,1009,1010,1011,1012,1013,1014,1015,199,199,1016,199,1017,1007,1018,1019,1020,1021,1022,1023,1024,1025,1026,199,199,1027,1028,1029,1007,1030,1031,1032,1033,1034,1035,1036,199,199,1037,1038,1007,1039,1040,1023,1041,1042,199,199,1043,1044,1007,1045,1046,1047,1048,1049,1050,1051,199,199,1052,1053,1054,969,1055,1056,1057,1058,1059,1060,1061,1062,1063,1064,1065,199,199,1066,1067,1068,1069,1070,1071,1072,1073,1074,1075,1046,1076,1077,1050,1078,199,199,1079,1080,1081,1082,1083,1084,1085,1086,1087,1088,199,199,1089,1090,1091,1092,1093,1094,1095,1096,1097,1007,1098,1099,1100,1101,1102,1103,199,199,1104,1105,1106,1107,1108,1109,1110,1111,199,199,1112,1007,1113,1114,1115,1116,1117,1118,1119,1120,1121,1122,199,1005,1123,1006,199,1124,1125,1126,1127,1128,1129,1130,1131,1132,1133,1134,199,1135,1136,1137,1138,199,199,1139,1140,1141,1142,1143,1144,1145,1146,1147,1148,1149,1150,199,1151,1152,1153,1154,1155,1156,1157,199,199,1158,1007,1159,1160,1161,1162,199,1163,1164,199,1165,1166,1167,1122,199,1005,1168,1169,199,1124,1170,1171,199,199,1172,1007,1173,1114,1118,1174,1175,1176,1177,1178,1179,1180,1181,1182,199,199,1183,1184,1185,1075,1046,1186,1077,1050,1187,199,199,1188,1189,1190,1191,1192,199,199,1193,1194],"contains_end_of_file","\"\"\"LeakFixMine pipeline tests (docs/leakfix-mine.md). Bare python3 or pytest:"," PYTHONPATH=. python3 leakmine/tests/test_leakmine.py","Covers the honest core: diff parsing, patch-signal classification (incl. penalties), SZZ","attribution + before/after correspondence + lead-time over a REAL temp git repo, the","verdict combiner (unique catch / unique miss), metrics aggregation, the over-vetted sweep","selection, and deterministic query/SQL generation. -O-safe (explicit raises, no bare assert).","\"\"\"","import json","import os","import sqlite3","import subprocess","import sys","import tempfile","HERE = os.path.dirname(os.path.abspath(__file__))","ROOT = os.path.dirname(os.path.dirname(HERE))","sys.path.insert(0, ROOT)","from leakmine import ( # noqa: E402"," bigquery, collect, confirm, diffparse, metrics, mine, schema, signals, sweep, szz,",")","def _expect(cond, msg):"," if not cond:"," raise AssertionError(msg)","# ---- diffparse -------------------------------------------------------------------","REACT_PATCH = \"\"\"diff --git a/src/Widget.tsx b/src/Widget.tsx","--- a/src/Widget.tsx","+++ b/src/Widget.tsx","@@ -10,5 +10,9 @@ export function Widget() {"," useEffect(() => {"," const onResize = () => setW(window.innerWidth);"," window.addEventListener('resize', onResize);","- }, []);","+ return () => {","+ window.removeEventListener('resize', onResize);","+ };","+ }, []);","+ // cleanup added"," }","def test_diffparse():"," fds = diffparse.parse_patch(REACT_PATCH)"," _expect(len(fds) == 1, \"one file diff\")"," fd = fds[0]"," _expect(fd.path == \"src/Widget.tsx\", f\"path {fd.path}\")"," added = \"\\n\".join(fd.added_text())"," _expect(\"removeEventListener\" in added, \"added removeEventListener\")"," # the fix removed old line 13 (`}, []);`) — it must register as touched."," _expect(fd.touches_old_line(13, window=0), \"old line 13 removed/replaced\")"," _expect(not fd.touches_old_line(10, window=0), \"context line 10 untouched\")"," _expect(fd.touches_old_line(11, window=2), \"window admits near line\")","def test_diffparse_body_lines_with_header_prefixes():"," # a removed SQL comment renders as \"--- ...\" at column 0; inside the hunk budget it"," # must be body content, not misread as a file header."," patch = (\"diff --git a/q.sql b/q.sql\\n--- a/q.sql\\n+++ b/q.sql\\n\""," \"@@ -1,2 +1,2 @@\\n--- old comment\\n+-- new comment\\n SELECT 1;\\n\")"," fds = diffparse.parse_patch(patch)"," _expect(len(fds) == 1, f\"one file, got {len(fds)}\")"," _expect(\"-- old comment\" in fds[0].removed_text(), \"removed comment captured as body\")"," _expect(\"-- new comment\" in fds[0].added_text(), \"added comment captured as body\")","def test_diffparse_bare_multifile():"," # bare patch (no `diff --git`): the second file's \"--- \" must start a NEW FileDiff,"," # not overwrite the first."," patch = (\"--- a/one.py\\n+++ b/one.py\\n@@ -1,1 +1,1 @@\\n-a\\n+b\\n\""," \"--- a/two.py\\n+++ b/two.py\\n@@ -1,1 +1,1 @@\\n-c\\n+d\\n\")"," _expect(len(fds) == 2, f\"two files, got {len(fds)}\")"," _expect({f.path for f in fds} == {\"one.py\", \"two.py\"}, f\"paths {[f.path for f in fds]}\")","# ---- signals ---------------------------------------------------------------------","def test_signals_react():"," cls = signals.classify(\"react_ts\", title=\"Fix memory leak in Widget\","," body=\"useEffect never removed the resize listener\", patch=REACT_PATCH)"," _expect(cls.category == signals.SUBSCRIPTION, f\"category {cls.category}\")"," _expect(cls.is_likely_fix, f\"likely fix, score={cls.score}\")"," _expect(\"title:leak-keyword\" in cls.evidence, \"title keyword scored\")","def test_signals_docs_penalty():"," docs = (\"diff --git a/README.md b/README.md\\n--- a/README.md\\n+++ b/README.md\\n\""," \"@@ -1,1 +1,2 @@\\n context\\n+a note about leaks\\n\")"," cls = signals.classify(\"react_ts\", title=\"docs: mention memory leak\", body=\"\", patch=docs)"," _expect(\"penalty:docs-only\" in cls.evidence, \"docs-only penalised\")"," _expect(not cls.is_candidate, f\"docs-only not a candidate, score={cls.score}\")","def test_signals_dotnet_event():"," patch = (\"diff --git a/A.xaml.cs b/A.xaml.cs\\n--- a/A.xaml.cs\\n+++ b/A.xaml.cs\\n\""," \"@@ -5,3 +5,4 @@\\n void Wire() {\\n- svc.Tick += OnTick;\\n\""," \"+ svc.Tick += OnTick;\\n+ Unloaded += (s,e) => svc.Tick -= OnTick;\\n }\\n\")"," cls = signals.classify(\"dotnet_wpf\", title=\"Fix event handler leak\", body=\"\", patch=patch)"," _expect(cls.is_candidate, f\"candidate, score={cls.score}\")","def test_signals_java_executor_single_form():"," # a fix that adds ONLY .shutdown() (not both forms) must still clear the threshold."," patch = (\"diff --git a/Svc.java b/Svc.java\\n--- a/Svc.java\\n+++ b/Svc.java\\n\""," \"@@ -5,2 +5,3 @@\\n void stop() {\\n+ pool.shutdown();\\n }\\n\")"," cls = signals.classify(\"java_spring\", title=\"fix ExecutorService memory leak\", body=\"\", patch=patch)"," _expect(cls.category == signals.TASK, f\"category {cls.category}\")"," _expect(cls.is_candidate, f\"single-form shutdown is a candidate, score={cls.score}\")","# ---- szz attribution + correspondence --------------------------------------------","def _finding(tool, rule, file, line, **kw):"," return szz.Finding(tool=tool, rule=rule, file=file, line=line, **kw)","def test_attribution_causal_vs_coincidental():"," f_on_fix = _finding(\"ownaudit\", \"OWN-EFFECT\", \"src/Widget.tsx\", 13)"," att = szz.attribute(REACT_PATCH, f_on_fix, window=0)"," _expect(att.fix_touches, \"finding on removed line is causal\")"," f_elsewhere = _finding(\"eslint\", \"no-leak\", \"src/Widget.tsx\", 200)"," att2 = szz.attribute(REACT_PATCH, f_elsewhere, window=2)"," _expect(not att2.fix_touches, \"far-away finding not causal\")"," _expect(att2.in_changed_file, \"but still in the changed file\")","def test_correspondence_confirmed_vs_lucky():"," target = _finding(\"ownaudit\", \"OWN-EFFECT\", \"src/Widget.tsx\", 13, resolution=\"semantic\")"," before = [target]"," after = [] # gone after"," corr = szz.correspond(before, after, REACT_PATCH, target, window=0)"," _expect(corr.confirmed_catch, \"flagged before, gone after, causal -> confirmed\")"," # coincidental: finding far from the fix also 'disappears' (e.g. file reshuffled)."," lucky = _finding(\"eslint\", \"x\", \"src/Widget.tsx\", 999)"," corr2 = szz.correspond([lucky], [], REACT_PATCH, lucky, window=2)"," _expect(corr2.detected_before and corr2.gone_after, \"before/after set\")"," _expect(not corr2.causal, \"not on fixed lines\")"," _expect(not corr2.confirmed_catch, \"gone-but-not-causal is NOT a confirmed catch\")","RENAME_PATCH = \"\"\"diff --git a/src/Old.tsx b/src/New.tsx","rename from src/Old.tsx","rename to src/New.tsx","--- a/src/Old.tsx","+++ b/src/New.tsx","@@ -10,3 +10,4 @@ export function Widget() {","+ }, []); // touched, but the leak is NOT actually fixed","+ // (rule still fires on the new path below)","def test_correspondence_rename_not_confirmed():"," # fix renames Old.tsx -> New.tsx and edits the line, but the SAME rule still fires on"," # the renamed file. That is a file-move, not a fix: it must NOT count as confirmed."," before = [_finding(\"ownaudit\", \"OWN-EFFECT\", \"src/Old.tsx\", 11)]"," after = [_finding(\"ownaudit\", \"OWN-EFFECT\", \"src/New.tsx\", 11)] # persists under new path"," corr = szz.correspond(before, after, RENAME_PATCH, before[0], window=2)"," _expect(corr.detected_before, \"flagged before\")"," _expect(not corr.gone_after, \"rule still fires on the renamed path -> not gone\")"," _expect(not corr.confirmed_catch, \"rename+persisting finding is NOT a confirmed catch\")","# ---- lead-time over a real temp git repo -----------------------------------------","def _git(repo, *args, date=\"2024-01-01T00:00:00\"):"," env = {**os.environ,"," \"GIT_AUTHOR_NAME\": \"t\", \"GIT_AUTHOR_EMAIL\": \"t@t\","," \"GIT_COMMITTER_NAME\": \"t\", \"GIT_COMMITTER_EMAIL\": \"t@t\","," \"GIT_AUTHOR_DATE\": date, \"GIT_COMMITTER_DATE\": date}"," return subprocess.run([\"git\", \"-C\", repo, *args], env=env,"," capture_output=True, text=True, check=True)","def test_lead_time():"," with tempfile.TemporaryDirectory() as repo:"," _git(repo, \"init\", \"-q\", \"-b\", \"main\")"," path = os.path.join(repo, \"svc.cs\")"," # commit 1: introduce a leaking line."," with open(path, \"w\") as fh:"," fh.write(\"class S {\\n void M() { Tick += OnTick; }\\n}\\n\")"," _git(repo, \"add\", \".\")"," _git(repo, \"commit\", \"-qm\", \"intro\", date=\"2024-01-01T00:00:00\")"," leak_sha = _git(repo, \"rev-parse\", \"HEAD\").stdout.strip()"," # a couple of unrelated commits."," for i, d in enumerate((\"2024-02-01T00:00:00\", \"2024-03-01T00:00:00\")):"," with open(os.path.join(repo, f\"f{i}.txt\"), \"w\") as fh:"," fh.write(\"x\\n\")"," _git(repo, \"add\", \".\"); _git(repo, \"commit\", \"-qm\", f\"noise{i}\", date=d)"," # the human fix: change line 2."," fh.write(\"class S {\\n void M() { Tick += OnTick; Unloaded += () => Tick -= OnTick; }\\n}\\n\")"," _git(repo, \"add\", \".\"); _git(repo, \"commit\", \"-qm\", \"fix leak\", date=\"2024-04-01T00:00:00\")"," lt = szz.lead_time(repo, leak_sha, \"svc.cs\", 2)"," _expect(lt.found, \"fix located in history\")"," # 2 intervening commits (noise0, noise1); the fix commit itself is excluded."," _expect(lt.commits_between == 2, f\"2 intervening commits, got {lt.commits_between}\")"," _expect(lt.fix_date.startswith(\"2024-04-01\"), f\"fix date {lt.fix_date}\")","# ---- confirm (verdict combiner) --------------------------------------------------","def test_judge_unique_catch_and_miss():"," own_before = [_finding(\"ownaudit\", \"OWN-EFFECT\", \"src/Widget.tsx\", 13, resolution=\"interproc\")]"," cand = confirm.Candidate("," id=\"c1\", ecosystem=\"react_ts\", title=\"fix memory leak\", body=\"listener leak\","," patch=REACT_PATCH,"," before={\"ownaudit\": own_before, \"eslint\": []},"," after={\"ownaudit\": [], \"eslint\": []},"," )"," v = confirm.judge(cand, ownaudit_tool=\"ownaudit\", baseline_tools=(\"eslint\",))"," _expect(v.is_real_fix, \"real fix\")"," _expect(\"ownaudit\" in v.caught_by, \"ownaudit caught\")"," _expect(v.unique_to_ownaudit, \"unique to ownaudit (eslint missed)\")"," _expect(\"eslint\" in v.missed_by, \"eslint in missed_by\")"," _expect(v.own_resolution == \"interproc\", f\"resolution {v.own_resolution}\")","def test_judge_tool_catch_does_not_define_real_fix():"," # a borderline patch (candidate but score < 10) that a tool confirm-catches must NOT be"," # promoted to a \"real fix\" — ground truth is the patch signal, not the tool's catch."," patch = (\"diff --git a/a.tsx b/a.tsx\\n--- a/a.tsx\\n+++ b/a.tsx\\n\""," \"@@ -1,3 +1,3 @@\\n function setup() {\\n- window.addEventListener('x', f);\\n\""," \"+ window.removeEventListener('x', f);\\n }\\n\")"," id=\"b1\", ecosystem=\"react_ts\", title=\"improve effect cleanup\", body=\"\", patch=patch,"," before={\"ownaudit\": [_finding(\"ownaudit\", \"OWN-X\", \"a.tsx\", 2)]},"," after={\"ownaudit\": []},"," v = confirm.judge(cand, ownaudit_tool=\"ownaudit\")"," _expect(\"ownaudit\" in v.caught_by, \"catch still recorded\")"," _expect(not v.is_real_fix, \"tool catch must not define a real fix\")"," _expect(\"borderline-send-to-review\" in v.notes, \"routed to review instead\")","def test_fp_after_matches_rule_and_file():"," patch = (\"diff --git a/a.cs b/a.cs\\n--- a/a.cs\\n+++ b/a.cs\\n\""," \"@@ -5,2 +5,3 @@\\n void M() {\\n+ DoThing();\\n }\\n\")"," # same (rule, file) survives on a touched file -> precision smell."," same = confirm.Candidate("," id=\"f1\", ecosystem=\"dotnet_wpf\", title=\"t\", body=\"\", patch=patch,"," before={\"ownaudit\": [_finding(\"ownaudit\", \"R1\", \"a.cs\", 5)]},"," after={\"ownaudit\": [_finding(\"ownaudit\", \"R1\", \"a.cs\", 5)]},"," _expect(confirm._fp_after(same, \"ownaudit\"), \"surviving same rule+file is a smell\")"," # a DIFFERENT rule on the same file must not count."," diff = confirm.Candidate("," id=\"f2\", ecosystem=\"dotnet_wpf\", title=\"t\", body=\"\", patch=patch,"," after={\"ownaudit\": [_finding(\"ownaudit\", \"R2\", \"a.cs\", 9)]},"," _expect(not confirm._fp_after(diff, \"ownaudit\"), \"unrelated rule must not inflate fp_after\")","def test_judge_unique_miss():"," # a real fix (strong patch signal) that NO tool caught -> the precious blind-spot bucket."," id=\"c2\", ecosystem=\"react_ts\", title=\"fix memory leak\", body=\"listener leak\","," patch=REACT_PATCH, before={\"ownaudit\": [], \"eslint\": []}, after={\"ownaudit\": [], \"eslint\": []},"," _expect(v.is_real_fix, \"still a real fix by patch signal\")"," _expect(not v.caught_by, \"no tool caught it\")"," _expect(\"real-fix-no-tool-caught\" in v.notes, \"flagged as shared blind spot\")","# ---- metrics ---------------------------------------------------------------------","def test_metrics_aggregate():"," vs = ["," confirm.Verdict(\"a\", \"react_ts\", \"subscription-leak\", 12, True,"," caught_by=[\"ownaudit\"], missed_by=[\"eslint\"], unique_to_ownaudit=True,"," own_resolution=\"interproc\"),"," confirm.Verdict(\"b\", \"react_ts\", \"timer-leak\", 11, True,"," caught_by=[\"ownaudit\", \"eslint\"], unique_to_ownaudit=False,"," own_resolution=\"syntactic\"),"," confirm.Verdict(\"c\", \"react_ts\", \"subscription-leak\", 9, True, caught_by=[],"," missed_by=[\"ownaudit\", \"eslint\"], notes=[\"real-fix-no-tool-caught\"]),"," ]"," rep = metrics.aggregate(vs, ownaudit_tool=\"ownaudit\", baseline_tools=(\"eslint\",))"," _expect(rep.n_real_fixes == 3, \"3 real fixes\")"," _expect(rep.catches[\"ownaudit\"] == 2, \"ownaudit caught 2\")"," _expect(rep.unique_to_ownaudit == 1, \"1 unique to ownaudit\")"," _expect(rep.unique_miss == 1, \"1 shared blind spot\")"," _expect(abs(rep.recall_on_corpus[\"ownaudit\"] - 2 / 3) < 1e-3, \"ownaudit recall 2/3\")"," _expect(rep.by_tier.get(\"interproc\") == 1 and rep.by_tier.get(\"syntactic\") == 1, \"tier split\")"," md = metrics.render_markdown(rep)"," _expect(\"LeakFixMine\" in md and \"by analysis tier\" in md, \"markdown renders\")","# ---- sweep (over-vetted selection) -----------------------------------------------","def test_sweep_over_vetted():"," polished = sweep.Package(\"Serilog\", \"nuget\", downloads=10**8, stars=6000, maintainers=8,"," open_issues=10, age_days=3000, has_ci=True, shape=\"library\")"," appish = sweep.Package(\"acme-dashboard\", \"npm\", downloads=10**6, stars=400, maintainers=2,"," open_issues=120, age_days=400, has_ci=False, shape=\"application\")"," sp = sweep.over_vetted_score(polished)"," sa = sweep.over_vetted_score(appish)"," _expect(sp.vetted > sa.vetted, \"polished lib scores more vetted\")"," _expect(\"high-stars-per-issue\" in sp.reason, \"stars/issue signal fires\")"," # many polished libs + a few apps; cap must let apps through despite higher downloads."," pkgs = [sweep.Package(f\"lib{i}\", \"nuget\", downloads=10**8, stars=8000, maintainers=9,"," open_issues=5, age_days=3000, has_ci=True, shape=\"library\")"," for i in range(8)]"," pkgs += [appish, sweep.Package(\"ctrl-plane\", \"npm\", downloads=5 * 10**5, stars=200,"," open_issues=90, age_days=300, shape=\"application\")]"," chosen = sweep.select_targets(pkgs, n=5, max_vetted_fraction=0.4)"," names = {c.pkg.name for c in chosen}"," n_vetted = sum(1 for c in chosen if c.vetted >= 0.6)"," _expect(n_vetted <= 2, f\"vetted cap (<=40% of 5) respected, got {n_vetted}\")"," _expect(names & {\"acme-dashboard\", \"ctrl-plane\"}, \"an app-shaped repo got in\")","def test_sweep_underfill_is_intentional():"," # an all-vetted pool can't fill n without breaching the cap: underfill is intentional,"," # NOT a backfill with the very libs the cap excludes."," for i in range(10)]"," _expect(len(chosen) == 2, f\"hard cap -> 2 of 5 (intentional underfill), got {len(chosen)}\")"," _expect(all(c.vetted >= 0.6 for c in chosen), \"the two admitted are the vetted libs\")","# ---- collect (deterministic query/SQL gen) ---------------------------------------","def test_queries_and_sql():"," qs = collect.github_search_queries(\"react_ts\", merged_after=\"2023-01-01\")"," _expect(all(\"merged:>=2023-01-01\" in q for q in qs), \"date qualifier appended\")"," _expect(any(\"useEffect\" in q for q in qs), \"useEffect query present\")"," sql = collect.gharchive_sql(\"dotnet_wpf\", date_from=\"20240101\", date_to=\"20241231\")"," _expect(\"_TABLE_SUFFIX BETWEEN '20240101' AND '20241231'\" in sql, \"partition-scoped\")"," _expect(\"PullRequestEvent\" in sql and \"memory leak\" in sql, \"PR + keyword filter\")"," _expect(\"pull_request.title\" in sql and \"pull_request.body\" in sql, \"title AND body matched\")","def test_fetch_search_with_fake_http():"," fake = json.dumps({\"items\": ["," {\"number\": 7, \"title\": \"fix leak\", \"html_url\": \"https://x/pr/7\","," \"repository_url\": \"https://api.github.com/repos/acme/widget\", \"pull_request\": {}},"," ]}).encode()"," got = collect.fetch_search(\"q\", http=lambda url, token: fake)"," _expect(len(got) == 1 and got[0].kind == \"pr\", \"parsed one PR\")"," _expect(got[0].repo == \"acme/widget\", f\"repo parsed: {got[0].repo}\")","def test_fetch_patch_with_fake_http():"," calls = {}"," def fake_http(url, token):"," calls[\"url\"] = url"," calls[\"token\"] = token"," return \"diff --git a/a b/a\\n@@ -1 +1 @@\\n-x\\n+y\\n\""," out = collect.fetch_patch(\"acme/w\", 7, token=\"T\", http=fake_http)"," _expect(out.startswith(\"diff --git\"), \"returns diff text\")"," _expect(calls[\"url\"].endswith(\"/repos/acme/w/pulls/7\"), f\"url {calls['url']}\")"," _expect(calls[\"token\"] == \"T\", \"token threaded through\")"," # a fetch error must degrade to \"\" (one bad PR can't abort a mining run), not raise."," def boom(url, token):"," raise RuntimeError(\"network\")"," _expect(collect.fetch_patch(\"a/b\", 1, http=boom) == \"\", \"fetch error -> empty string\")","def test_mine_run_orchestration():"," cands = ["," collect.Candidate(\"\", \"acme/w\", 1, \"pr\", \"fix memory leak\", \"u1\", \"listener leak\"),"," collect.Candidate(\"\", \"acme/w\", 1, \"pr\", \"dup of #1\", \"u1\", \"\"), # dedup by (repo,#)"," collect.Candidate(\"\", \"acme/x\", 2, \"pr\", \"chore: tidy\", \"u2\", \"\"), # weak -> dropped"," collect.Candidate(\"\", \"acme/y\", 3, \"issue\", \"an issue\", \"u3\", \"\"), # not a PR -> skip"," weak = (\"diff --git a/x.tsx b/x.tsx\\n--- a/x.tsx\\n+++ b/x.tsx\\n\""," \"@@ -1,1 +1,1 @@\\n-const a = 1;\\n+const a = 2;\\n\")"," patches = {(\"acme/w\", 1): REACT_PATCH, (\"acme/x\", 2): weak}"," def fake_search(query, *, token=\"\", per_page=50):"," return cands # query pack repeats the same hits; dedup must collapse them"," def fake_patch(repo, number, *, token=\"\"):"," return patches.get((repo, number), \"\")"," conn = schema.connect(\":memory:\")"," res = mine.run(\"react_ts\", search=fake_search, fetch_patch=fake_patch, conn=conn, min_score=7)"," _expect(res.seen == 2, f\"PR-only + deduped examined, got {res.seen}\")"," _expect(res.kept == 1, f\"only the real leak fix kept, got {res.kept}\")"," _expect(res.rows and res.rows[0][\"repo\"] == \"acme/w\", \"kept the leak fix\")"," _expect(res.rows[0][\"category\"] == signals.SUBSCRIPTION, \"classified as subscription\")"," _expect(schema.count(conn, \"candidates\") == 1, \"kept candidate stored\")"," _expect(schema.count(conn, \"labels\") == 1, \"kept label stored\")"," _expect(\"mining run\" in res.summary_md(), \"summary renders\")","# ---- schema (store round-trip) ---------------------------------------------------","def test_schema_store():"," schema.insert_candidate(conn, {\"id\": \"c1\", \"ecosystem\": \"react_ts\", \"repo\": \"a/b\","," \"number\": 1, \"kind\": \"pr\", \"title\": \"fix leak\", \"merged\": 1})"," schema.insert_label(conn, \"c1\", \"subscription-leak\", 12, [\"patch:subscription-leak\"], \"patch\")"," v = confirm.Verdict(\"c1\", \"react_ts\", \"subscription-leak\", 12, True,"," caught_by=[\"ownaudit\"], unique_to_ownaudit=True)"," schema.insert_verdict(conn, v)"," _expect(schema.count(conn, \"candidates\") == 1, \"candidate stored\")"," _expect(schema.count(conn, \"labels\") == 1, \"label stored\")"," _expect(schema.count(conn, \"verdicts\") == 1, \"verdict stored\")","def test_candidate_upsert_preserves_children():"," # re-inserting a candidate must UPDATE in place, not DELETE+reinsert — otherwise (with"," # FK on) it would orphan/cascade the child label below."," schema.insert_candidate(conn, {\"id\": \"c1\", \"ecosystem\": \"react_ts\", \"title\": \"old\", \"merged\": 1})"," schema.insert_label(conn, \"c1\", \"subscription-leak\", 12, [\"e\"], \"patch\")"," schema.insert_candidate(conn, {\"id\": \"c1\", \"ecosystem\": \"react_ts\", \"title\": \"new\", \"merged\": 1})"," _expect(schema.count(conn, \"candidates\") == 1, \"still one candidate\")"," _expect(schema.count(conn, \"labels\") == 1, \"child label survived the re-insert\")"," title = conn.execute(\"SELECT title FROM candidates WHERE id='c1'\").fetchone()[0]"," _expect(title == \"new\", f\"row updated in place, title={title}\")","def test_schema_verdict_idempotent():"," # re-confirming the same candidate upserts, so resume never double-counts verdicts."," schema.insert_candidate(conn, {\"id\": \"c1\", \"ecosystem\": \"react_ts\", \"title\": \"t\", \"merged\": 1})"," v = confirm.Verdict(\"c1\", \"react_ts\", \"subscription-leak\", 12, True, caught_by=[\"ownaudit\"])"," schema.insert_verdict(conn, v) # second confirm of the same candidate"," _expect(schema.count(conn, \"verdicts\") == 1, \"verdict upserts, not duplicated\")","def test_schema_foreign_keys_enforced():"," # FK enforcement is ON: a verdict for an unknown candidate must be rejected."," v = confirm.Verdict(\"ghost\", \"react_ts\", \"subscription-leak\", 12, True)"," try:"," schema.insert_verdict(conn, v)"," raised = False"," except sqlite3.IntegrityError: # specifically the FK rejection, not any error"," raised = True"," _expect(raised, \"orphan verdict rejected by foreign-key constraint\")","def test_bq_gharchive_sql():"," sql = bigquery.gharchive_discovery_sql(\"dotnet_wpf\", date_from=\"20240101\", date_to=\"20241231\","," max_changed_files=80)"," # assert the actual OR boolean, not mere field presence (an AND regression must fail)."," _expect((\"title')) LIKE '%memory leak%' OR \""," \"LOWER(COALESCE(JSON_EXTRACT_SCALAR(payload,'$.pull_request.body'),'')) \""," \"LIKE '%memory leak%'\") in sql, \"title OR body, not AND\")"," _expect(\"base.repo.language')) = 'c#'\" in sql, \"language filter\")"," _expect(\"changed_files') AS INT64) <= 80\" in sql, \"size cap applied\")"," _expect(\"QUALIFY ROW_NUMBER()\" in sql, \"deduped to one row per PR\")"," # size cap can be disabled."," nocap = bigquery.gharchive_discovery_sql(\"react_ts\", date_from=\"20240101\", date_to=\"20240131\","," max_changed_files=0)"," _expect(\"changed_files') AS INT64) <=\" not in nocap, \"cap omitted when 0\")","def test_bq_contents_sweep_sql():"," s = bigquery.contents_sweep_sql(\"react_ts\", sample=True)"," _expect(\"sample_files\" in s and \"sample_contents\" in s, \"uses cheap sample tables by default\")"," _expect(\"REGEXP_CONTAINS(c.content, r'addEventListener\\\\(')\" in s, \"acquire regex present\")"," _expect(\"NOT REGEXP_CONTAINS(c.content, r'removeEventListener')\" in s, \"cleanup-absence test\")"," full = bigquery.contents_sweep_sql(\"react_ts\", sample=False)"," _expect(\"github_repos.files`\" in full and \"github_repos.contents`\" in full, \"full tables\")"," _expect(\"2.7TB\" in full, \"full scan is cost-warned\")"," # an ecosystem without sweep pairs raises rather than emit garbage."," raised = False"," bigquery.contents_sweep_sql(\"nim\")"," except ValueError:"," _expect(raised, \"missing sweep pairs -> explicit error\")","def test_bq_metadata_score():"," hi, ev = bigquery.metadata_score(\"dotnet_wpf\", title=\"Fix memory leak in view\","," body=\"event handler retained\", changed_files=4)"," _expect(hi >= 4 and \"small-pr\" in ev, f\"focused leak fix ranks high, {hi}\")"," lo, ev2 = bigquery.metadata_score(\"dotnet_wpf\", title=\"Fix memory leak\", body=\"\","," changed_files=500)"," # +3 title keyword, -3 mega penalty -> 0; assert the score, not just the evidence."," _expect(lo == 0 and \"penalty:mega-pr\" in ev2, f\"mega-PR penalty zeroes the score, {lo}\")"," none, _ = bigquery.metadata_score(\"dotnet_wpf\", title=\"add feature\", body=\"\", changed_files=3)"," _expect(none < 4, \"no-keyword PR below threshold\")","def test_bq_ingest_ndjson():"," ndjson = \"\\n\".join(["," json.dumps({\"repo\": \"a/b\", \"number\": 1, \"title\": \"Fix memory leak\", \"body\": \"x\","," \"changed_files\": 3, \"html_url\": \"u1\"}),"," json.dumps({\"repo\": \"a/b\", \"number\": 1, \"title\": \"dup\", \"body\": \"\", \"changed_files\": 3}),"," json.dumps({\"repo\": \"c/d\", \"number\": 9, \"title\": \"add feature\", \"body\": \"\","," \"changed_files\": 2}), # no keyword -> dropped"," ])"," rows = bigquery.read_ndjson(ndjson)"," res = bigquery.ingest_rows(rows, \"dotnet_wpf\", min_meta_score=4, conn=conn)"," _expect(res.seen == 2, f\"deduped by (repo,number), got {res.seen}\")"," _expect(res.kept == 1, f\"only the keyword'd small PR kept, got {res.kept}\")"," _expect(res.rows[0][\"repo\"] == \"a/b\", \"kept the leak PR\")"," _expect(schema.count(conn, \"candidates\") == 1 and schema.count(conn, \"labels\") == 1, \"stored\")"," _expect(\"metadata tier\" in res.summary_md(), \"summary flags the tier\")","def test_signals_no_oom_substring_noise():"," # bare \"oom\" was removed (it matched zoom/room/doom under substring) — junk titles must"," # not register a leak keyword, but a real leak title still must."," for junk in (\"Refactor Room screen\", \"bump react-zoom-pan-pinch\", \"New Gloomier Bangs\"):"," cls = signals.classify(\"react_ts\", title=junk, body=\"\", patch=\"\")"," _expect(\"title:leak-keyword\" not in cls.evidence, f\"{junk!r} must not match a keyword\")"," cls = signals.classify(\"react_ts\", title=\"fix memory leak\", body=\"\", patch=\"\")"," _expect(\"title:leak-keyword\" in cls.evidence, \"real leak title still matches\")","def test_classify_from_store():"," schema.insert_candidate(conn, {\"id\": \"a/b#1\", \"ecosystem\": \"react_ts\", \"repo\": \"a/b\","," \"number\": 1, \"kind\": \"pr\", \"title\": \"fix memory leak\","," \"body\": \"listener leak\", \"merged\": 1})"," schema.insert_candidate(conn, {\"id\": \"c/d#2\", \"ecosystem\": \"react_ts\", \"repo\": \"c/d\","," \"number\": 2, \"kind\": \"pr\", \"title\": \"chore: tidy\","," \"body\": \"\", \"merged\": 1})"," patches = {(\"a/b\", 1): REACT_PATCH,"," (\"c/d\", 2): \"diff --git a/x.tsx b/x.tsx\\n--- a/x.tsx\\n+++ b/x.tsx\\n\""," \"@@ -1,1 +1,1 @@\\n-const a = 1;\\n+const a = 2;\\n\"}"," fetched = []"," fetched.append((repo, number))"," res = mine.classify_from_store(conn, \"react_ts\", fetch_patch=fake_patch, min_score=7)"," _expect(res.seen == 2, f\"both stored PRs examined, {res.seen}\")"," _expect(res.fetched == 2, f\"both diffs fetched, {res.fetched}\")"," _expect(res.kept == 1, f\"only the real leak fix kept at patch tier, {res.kept}\")"," _expect(res.rows[0][\"repo\"] == \"a/b\" and res.rows[0][\"category\"] == signals.SUBSCRIPTION,"," \"patch-tier category assigned\")"," # every attempt is labelled at the patch tier: the real category + a below-threshold miss."," total = conn.execute(\"SELECT COUNT(*) FROM labels WHERE classifier='patch'\").fetchone()[0]"," real = conn.execute(\"SELECT COUNT(*) FROM labels WHERE classifier='patch' \""," \"AND label NOT IN ('below-threshold','fetch-failed')\").fetchone()[0]"," _expect(total == 2 and real == 1, f\"2 patch labels, 1 real verdict; got {total}/{real}\")"," # resume: a second run must skip everything already attempted — no re-fetch, no dupes."," res2 = mine.classify_from_store(conn, \"react_ts\", fetch_patch=fake_patch, min_score=7)"," _expect(res2.seen == 0, f\"resume processes nothing, got {res2.seen}\")"," _expect(len(fetched) == 2, f\"no diffs re-fetched on resume, got {len(fetched)}\")","def test_fetch_repo_languages_batched_graphql():"," captured = {}"," def fake_http(body, token):"," captured[\"body\"] = body"," return json.dumps({\"data\": {"," \"r0\": {\"primaryLanguage\": {\"name\": \"TypeScript\"}},"," \"r1\": {\"primaryLanguage\": None}, # empty repo"," \"r2\": None, # missing/inaccessible repo"," }}).encode()"," out = collect.fetch_repo_languages([\"a/ts\", \"b/empty\", \"c/gone\"], token=\"T\", http=fake_http)"," _expect(out == {\"a/ts\": \"typescript\", \"b/empty\": \"\", \"c/gone\": \"\"}, f\"parsed langs: {out}\")"," _expect(b\"repository(owner:\" in captured[\"body\"], \"batched aliased query built\")"," # whole-batch failure must OMIT the repos (so the caller retries), not record \"unknown\"."," def boom(body, token):"," raise RuntimeError(\"401\")"," _expect(collect.fetch_repo_languages([\"x/y\"], http=boom) == {}, \"failed batch -> empty map\")"," # an errors-only response (no data) is likewise treated as failure."," no_data = collect.fetch_repo_languages([\"x/y\"], http=lambda b, t: b'{\"errors\":[{\"m\":\"x\"}]}')"," _expect(no_data == {}, \"errors-only response -> empty map\")","def test_enrich_skips_wrong_language_before_fetch():"," for repo, num, title in [(\"ts/app\", 1, \"fix memory leak\"), (\"rb/app\", 2, \"fix memory leak\")]:"," schema.insert_candidate(conn, {\"id\": f\"{repo}#{num}\", \"ecosystem\": \"react_ts\","," \"repo\": repo, \"number\": num, \"kind\": \"pr\","," \"title\": title, \"body\": \"\", \"merged\": 1})"," def fake_langs(repos, *, token=\"\", batch=100):"," return {\"ts/app\": \"typescript\", \"rb/app\": \"ruby\"}"," er = mine.enrich_languages(conn, \"react_ts\", fetch_languages=fake_langs)"," _expect(er.repos == 2 and er.skipped == 1, f\"one wrong-language marked, {er.skipped}\")"," # the ruby repo now carries a patch-tier 'wrong-language' label -> classify-store skips it."," fetched.append(repo)"," return REACT_PATCH # would score high IF fetched"," _expect(fetched == [\"ts/app\"], f\"only the TS repo's diff fetched, got {fetched}\")"," _expect(res.fetched == 1, f\"ruby repo skipped before fetch, fetched={res.fetched}\")","def test_enrich_retries_after_fetch_failure():"," schema.insert_candidate(conn, {\"id\": \"x/y#1\", \"ecosystem\": \"react_ts\", \"repo\": \"x/y\","," # whole-batch failure -> fetcher returns {} -> nothing recorded, nothing persisted."," er = mine.enrich_languages(conn, \"react_ts\", fetch_languages=lambda r, *, token=\"\", batch=100: {})"," _expect(er.repos == 0 and er.labeled == 0, \"failure records nothing\")"," n = conn.execute(\"SELECT COUNT(*) FROM labels WHERE classifier='language'\").fetchone()[0]"," _expect(n == 0, \"no 'unknown' label persisted on failure (so it stays retryable)\")"," # a later run with a working fetch enriches it."," er2 = mine.enrich_languages(conn, \"react_ts\","," fetch_languages=lambda r, *, token=\"\", batch=100: {x: \"typescript\" for x in r})"," _expect(er2.repos == 1 and er2.labeled == 1, \"retry succeeds after token is fixed\")","def test_bq_ingest_rejects_contents_rows():"," # contents-sweep rows (repo/path/signal, no PR number) must fail fast, not silently drop."," rows = [{\"repo\": \"a/b\", \"path\": \"src/x.tsx\", \"signal\": \"listener\"}]"," bigquery.ingest_rows(rows, \"react_ts\")"," _expect(raised, \"contents-sweep rows rejected by bq-ingest\")","def main():"," tests = [v for k, v in sorted(globals().items()) if k.startswith(\"test_\") and callable(v)]"," for t in tests:"," t()"," print(f\"ok - {len(tests)} leakmine tests passed\")","if __name__ == \"__main__\":"," main()","leakmine/mine.py",[1197],{"_700":522,"_701":1198,"_401":1199,"_704":59},272,[1200,199,1201,1202,1203,1204,1205,199,1206,1207,1208,1209,199,1210,1211,711,1212,199,1213,1214,199,1215,199,199,1216,1217,1218,1219,1220,1221,1222,199,1223,1224,1225,1226,1227,1228,1229,1230,1231,1232,1229,1233,1229,1234,1235,1236,1237,1238,1239,1240,1241,1242,1243,1244,199,199,1245,1246,1247,1248,1249,1250,1251,1252,1253,1254,1255,1256,1257,1258,1259,1260,1261,1262,199,1263,1264,199,1265,1266,1267,1268,1269,1270,1271,1272,1273,1274,199,1275,1276,1277,1278,1272,1279,199,1280,1281,1272,1282,1283,1284,1285,1286,1287,1288,1289,1290,1291,1292,1293,1294,1295,1296,1297,1298,1299,1300,1301,1302,1303,1304,1305,199,199,1216,1306,1218,1307,1308,1309,1310,199,1223,1227,1311,1229,1312,1313,1314,1229,1315,1236,1316,1317,1244,199,199,1318,1319,1320,1247,1248,1321,1322,1323,1324,1325,1326,199,1327,1328,1329,1330,1331,1332,1333,1334,1335,1336,1337,1338,1339,1340,1341,874,1342,1343,1344,1345,1346,1347,1348,1349,1350,1351,1352,1353,1354,1355,1356,1357,1358,1359,1360,1361,1362,1363,1364,1365,1305,199,199,1366,1319,1320,1247,1248,1251,1252,1322,1255,1256,1367,1368,1369,1370,199,1371,1372,1373,199,1374,1375,1376,1377,1378,1379,1332,1262,1380,1381,1382,1383,1384,1336,1385,1386,1387,1388,874,1342,1343,1344,1345,1389,1390,1391,1392,1267,1268,1393,1394,1395,1396,1397,1398,1399,1400,1401,1396,1402,1403,1404,1405,1406,1407,1408,1365,1305],"\"\"\"Mining orchestrator — discover → fetch patch → classify → store, one ecosystem at a time.","This is the CI-runnable arm of the pipeline (docs/leakfix-mine.md §4/§10). It chains the","GitHub-Search query pack (`collect.github_search_queries`) → per-PR patch fetch","(`collect.fetch_patch`) → patch-signal classification (`signals.classify`), dedups by","(repo, number), keeps everything scoring at/above the candidate threshold, and writes the","result to the SQLite store plus a JSON dataset and a markdown summary.","Deliberately scoped to **discovery + classification** — the part that runs cleanly on a","hosted runner with only `GITHUB_TOKEN`. The before/after tool comparison (`confirm` /","`metrics`) needs per-repo checkout + analyzer builds, so it stays a local / self-hosted","step; see docs/leakfix-mine.md §7.","Network access is isolated behind the injected `search` / `fetch_patch` callables, so the","orchestration unit-tests offline with fakes (CI never reaches out).","from __future__ import annotations","import time","from dataclasses import dataclass, field","from . import collect, schema, signals","@dataclass","class MineResult:"," ecosystem: str"," rows: list[dict] = field(default_factory=list) # one per kept candidate"," seen: int = 0 # candidates examined (post-dedup)"," fetched: int = 0 # patches successfully fetched"," kept: int = 0 # candidates >= min_score"," def summary_md(self) -> str:"," by_cat: dict[str, int] = {}"," for r in self.rows:"," by_cat[r[\"category\"]] = by_cat.get(r[\"category\"], 0) + 1"," lines = ["," f\"# LeakFixMine — `{self.ecosystem}` mining run\","," \"\","," f\"- candidates examined: **{self.seen}**\","," f\"- patches fetched: **{self.fetched}**\","," f\"- kept (>= threshold): **{self.kept}**\","," \"## kept by category\","," \"| category | kept |\","," \"|---|---:|\","," ]"," for cat, c in sorted(by_cat.items(), key=lambda kv: -kv[1]):"," lines.append(f\"| {cat} | {c} |\")"," lines += [\"\", \"## kept candidates\", \"\", \"| score | category | repo#PR | title |\","," \"|---:|---|---|---|\"]"," for r in sorted(self.rows, key=lambda r: -r[\"score\"])[:100]:"," title = (r[\"title\"] or \"\").replace(\"|\", \"\\\\|\")[:80]"," lines.append(f\"| {r['score']} | {r['category']} | {r['repo']}#{r['number']} | {title} |\")"," return \"\\n\".join(lines) + \"\\n\"","def run("," ecosystem: str,"," *,"," token: str = \"\","," merged_after: str = \"\","," per_query: int = 50,"," min_score: int = 7,"," sleep: float = 0.0,"," conn=None,"," search=None,"," fetch_patch=None,",") -> MineResult:"," \"\"\"Mine one ecosystem. `search(query, *, token, per_page) -> list[Candidate]` and"," `fetch_patch(repo, number, *, token) -> str` default to the live `collect` helpers but"," are injected as fakes in tests. `sleep` throttles between API calls to stay under the"," Search API's ~30 req/min; pass 0 in tests.\"\"\""," search = search or collect.fetch_search"," fetch_patch = fetch_patch or collect.fetch_patch"," res = MineResult(ecosystem=ecosystem)"," seen: set[tuple[str, int]] = set()"," for query in collect.github_search_queries(ecosystem, merged_after=merged_after):"," candidates = search(query, token=token, per_page=per_query)"," if sleep:"," time.sleep(sleep)"," for cand in candidates:"," key = (cand.repo, cand.number)"," if cand.kind != \"pr\" or key in seen:"," continue"," seen.add(key)"," res.seen += 1"," patch = fetch_patch(cand.repo, cand.number, token=token)"," if sleep:"," time.sleep(sleep)"," if not patch:"," res.fetched += 1"," cls = signals.classify(ecosystem, title=cand.title, body=cand.body, patch=patch)"," if cls.score < min_score:"," res.kept += 1"," row = {"," \"ecosystem\": ecosystem,"," \"repo\": cand.repo,"," \"number\": cand.number,"," \"url\": cand.url,"," \"title\": cand.title,"," \"category\": cls.category,"," \"score\": cls.score,"," \"is_candidate\": cls.is_candidate,"," \"is_likely_fix\": cls.is_likely_fix,"," \"evidence\": cls.evidence,"," }"," res.rows.append(row)"," if conn is not None:"," schema.insert_candidate(conn, {"," \"id\": f\"{cand.repo}#{cand.number}\", \"ecosystem\": ecosystem,"," \"query\": query, \"repo\": cand.repo, \"number\": cand.number,"," \"kind\": \"pr\", \"title\": cand.title, \"body\": cand.body,"," \"state\": \"merged\", \"merged\": 1, \"url\": cand.url,"," })"," schema.insert_label(conn, f\"{cand.repo}#{cand.number}\", cls.category,"," cls.score, cls.evidence, \"patch\")"," return res","class EnrichResult:"," repos: int = 0 # repos looked up this run"," labeled: int = 0 # candidates given a language label"," skipped: int = 0 # candidates marked wrong-language (won't be diff-fetched)"," by_lang: dict = field(default_factory=dict)"," f\"# LeakFixMine — language enrichment (`{self.ecosystem}`)\","," f\"- repos enriched: **{self.repos}**\","," f\"- candidates labelled: **{self.labeled}**\","," f\"- marked wrong-language (skipped before diff fetch): **{self.skipped}**\","," \"| language | candidates |\", \"|---|---:|\","," for lang, n in sorted(self.by_lang.items(), key=lambda kv: -kv[1]):"," lines.append(f\"| {lang or 'unknown'} | {n} |\")","def enrich_languages("," conn,"," eco_key: str,"," batch: int = 100,"," limit=None,"," fetch_languages=None,",") -> EnrichResult:"," \"\"\"Enrich stored candidates with their repo's primary language (batched GraphQL) and mark"," cross-language ones so `classify_from_store` skips them BEFORE the expensive diff fetch."," Run between `bq-ingest` and `classify-store`. For each candidate it writes a"," `classifier='language'` label (record), and if the repo language is known and NOT in this"," ecosystem's `signals.LANGS`, also a `classifier='patch'` label `wrong-language` — which the"," classify-store anti-join already excludes, so no diff is fetched for it. Unknown languages"," are left for classify-store to try. Resumable: repos already language-labelled are skipped."," \"\"\""," fetch_languages = fetch_languages or collect.fetch_repo_languages"," target = set(signals.LANGS.get(eco_key, ()))"," res = EnrichResult(ecosystem=eco_key)"," sql = ("," \"SELECT DISTINCT c.repo FROM candidates c \""," \"LEFT JOIN labels lg ON lg.candidate_id = c.id AND lg.classifier = 'language' \""," \"LEFT JOIN labels lp ON lp.candidate_id = c.id AND lp.classifier = 'patch' \""," \"WHERE c.ecosystem = ? AND c.kind = 'pr' \""," \"AND lg.candidate_id IS NULL AND lp.candidate_id IS NULL ORDER BY c.repo\""," params = [eco_key]"," if limit is not None:"," sql += \" LIMIT ?\""," params.append(limit)"," repos = [row[0] for row in conn.execute(sql, params)]"," if not repos:"," return res"," langs = fetch_languages(repos, token=token, batch=batch)"," for repo in repos:"," if repo not in langs:"," continue # lookup failed for this repo's batch — leave it for a retry"," lang = langs[repo]"," res.repos += 1"," rows = conn.execute("," \"SELECT id FROM candidates WHERE ecosystem = ? AND repo = ?\", (eco_key, repo),"," ).fetchall()"," for (cid,) in rows:"," schema.insert_label(conn, cid, lang or \"unknown\", 0, [], \"language\")"," res.labeled += 1"," res.by_lang[lang] = res.by_lang.get(lang, 0) + 1"," if target and lang and lang not in target:"," schema.insert_label(conn, cid, \"wrong-language\", 0, [], \"patch\")"," res.skipped += 1"," conn.commit()","def classify_from_store("," \"\"\"Close the loop: take the candidate PRs already in the store (from a BigQuery /"," GH-Archive ingest, which only had title/body metadata), fetch each diff, and run the"," REAL patch-tier `signals.classify`. The metadata tier was a fetch *queue*; this is the"," verdict — category + patch-signal score, kept at `min_score`."," `fetch_patch(repo, number, *, token) -> str` defaults to the live `collect` helper and"," is injected as a fake in tests. Labels are committed PER ATTEMPT (not just at the end),"," so a crash/kill mid-run doesn't roll back resume progress and re-burn diff quota."," RESUMABLE: every attempted PR gets a `classifier='patch'` label — the real category when"," kept, `below-threshold` for a fetched-but-low-score miss, `fetch-failed` when no diff came"," back — and any PR that already has such a label is skipped BEFORE `--limit` is counted. So"," `--limit`-sized batches (and reruns after an interruption) advance through the queue"," instead of re-burning diff quota on the same head. To retry failures, delete their"," `fetch-failed` labels."," res = MineResult(ecosystem=eco_key)"," # Exclude already-attempted candidates IN SQL (anti-join on a patch-tier label) and push"," # --limit into the query, so reruns/batches stream only the unprocessed head instead of"," # loading the whole table and re-burning diff quota. ORDER BY c.id makes the batch order"," # stable (and materialises the result, so the in-loop label inserts can't disturb it)."," \"SELECT c.repo, c.number, c.title, c.body FROM candidates AS c \""," \"LEFT JOIN labels AS l ON l.candidate_id = c.id AND l.classifier = 'patch' \""," \"WHERE c.ecosystem = ? AND c.kind = 'pr' AND l.candidate_id IS NULL \""," \"ORDER BY c.id\""," for repo, number, title, body in conn.execute(sql, params):"," cid = f\"{repo}#{number}\""," res.seen += 1"," patch = fetch_patch(repo, number, token=token)"," if not patch:"," schema.insert_label(conn, cid, \"fetch-failed\", 0, [], \"patch\")"," conn.commit() # persist progress per attempt (crash-safe resume)"," continue"," res.fetched += 1"," cls = signals.classify(eco_key, title=title or \"\", body=body or \"\", patch=patch)"," if cls.score < min_score:"," schema.insert_label(conn, cid, \"below-threshold\", cls.score, cls.evidence, \"patch\")"," conn.commit()"," res.kept += 1"," res.rows.append({"," \"ecosystem\": eco_key, \"repo\": repo, \"number\": number,"," \"title\": title, \"category\": cls.category, \"score\": cls.score,"," \"is_likely_fix\": cls.is_likely_fix, \"evidence\": cls.evidence,"," })"," schema.insert_label(conn, cid, cls.category, cls.score, cls.evidence, \"patch\")","leakmine/collect.py",[1411],{"_700":522,"_701":1412,"_401":1413,"_704":59},195,[1414,199,1415,199,1416,1417,1418,1419,1420,1421,1422,199,1423,1424,1425,711,1212,199,712,1426,1427,1428,199,1429,199,199,1430,1431,1432,1433,1434,1435,1436,1437,199,199,1438,1320,1247,1439,1440,1441,1442,1443,1444,199,1445,1446,1447,1332,1433,1448,1449,1450,1451,1452,1453,1454,874,1455,1456,1457,1458,1459,1460,1461,1462,1463,1464,1465,1466,1467,1468,1469,1470,1471,1472,1473,874,199,199,1216,1474,1218,1475,1476,1477,1478,1479,1480,199,199,1481,1482,1483,1484,1485,1486,874,1487,1488,1489,1490,1491,1492,1493,1494,1495,1496,1497,1498,1499,1500,1501,199,199,1502,1503,1504,1505,1506,1507,1046,1508,1509,1510,199,199,1511,1512,1513,1514,1515,1516,1517,1518,1519,1520,1521,1522,1523,1524,1525,1526,1527,1528,1529,1530,1531,1532,1533,1534,1535,1396,1521,1536,1537,1538,1501,199,199,1539,1540,1541,1542,1543,874,1544,1545,199,199,1546,1547,1548,1549,1550,1551,1544,1552,199,199,1553,1547,1554,1549,1550,1551,1544,1545,199,199,1555,1556,1557,1558],"\"\"\"Discovery — generate the candidate-finding queries, deterministically.","Two collection backends, by scale (docs/leakfix-mine.md §4):"," - GitHub Search API — fine for a first slice, but brutal at scale: ~30 req/min, 1000"," results per query hard cap, and `language:` is the *repo's* language, not the diff's."," - GH Archive (gharchive.org, mirrored as the public BigQuery `githubarchive` dataset) —"," the whole event firehose. We filter PR/issue events by leak keywords in SQL, no rate"," limit. The data is free; querying via BigQuery needs a GCP account (1 TB/month free),"," so we ALWAYS scope by day-partition to stay inside the free tier. GH Archive replaces"," *discovery* only — the patch itself still comes per-PR from the API or a clone.","Both query forms are generated here as pure strings so they unit-test with no network.","The actual fetch is isolated behind `fetch_search`, which takes an injectable HTTP getter","(so tests pass a fake and CI never reaches out).","import urllib.parse","import urllib.request","from dataclasses import dataclass","from . import signals as sig","def github_search_queries(eco_key: str, *, merged_after: str = \"\") -> list[str]:"," \"\"\"The GitHub-search query pack for an ecosystem, optionally date-bounded."," `merged_after` is an ISO date (YYYY-MM-DD); it appends `merged:>=DATE`.\"\"\""," eco = sig.ECOSYSTEMS[eco_key]"," qs = list(eco.queries)"," if merged_after:"," qs = [f\"{q} merged:>={merged_after}\" for q in qs]"," return qs","def gharchive_sql("," date_from: str,"," date_to: str,"," limit: int = 2000,",") -> str:"," \"\"\"A BigQuery query over `githubarchive.day.*` for merged PRs whose title/body carry a"," leak keyword and whose repo language matches. Day-partition scan keeps it cheap."," `date_from`/`date_to` are YYYYMMDD (the `day` table suffix). We pass them through"," `_TABLE_SUFFIX BETWEEN` so BigQuery only scans those partitions — the difference"," between a few cents and a blown free tier."," # match the keyword in EITHER the PR title or body — a generic title with the leak"," # described in the body must still enter the corpus (COALESCE guards a null body)."," kw = \" OR \".join("," \"(LOWER(JSON_EXTRACT_SCALAR(payload, '$.pull_request.title')) LIKE '%{k}%' OR \""," \"LOWER(COALESCE(JSON_EXTRACT_SCALAR(payload, '$.pull_request.body'), '')) \""," \"LIKE '%{k}%')\".format(k=k)"," for k in eco.keywords"," exts = \" OR \".join(f\"f LIKE '%{e}'\" for e in eco.file_ext)"," return ("," \"-- LeakFixMine discovery: merged leak-fix PRs, partition-scoped.\\n\""," \"SELECT\\n\""," \" repo.name AS repo,\\n\""," \" JSON_EXTRACT_SCALAR(payload, '$.number') AS number,\\n\""," \" JSON_EXTRACT_SCALAR(payload, '$.pull_request.title') AS title,\\n\""," \" JSON_EXTRACT_SCALAR(payload, '$.pull_request.merged_at') AS merged_at,\\n\""," \" JSON_EXTRACT_SCALAR(payload, '$.pull_request.base.repo.language') AS lang\\n\""," \"FROM `githubarchive.day.*`\\n\""," f\"WHERE _TABLE_SUFFIX BETWEEN '{date_from}' AND '{date_to}'\\n\""," \" AND type = 'PullRequestEvent'\\n\""," \" AND JSON_EXTRACT_SCALAR(payload, '$.action') = 'closed'\\n\""," \" AND JSON_EXTRACT_SCALAR(payload, '$.pull_request.merged') = 'true'\\n\""," f\" AND ({kw})\\n\""," \" -- repo-language prefilter; the real diff-language check happens on changed files:\\n\""," f\" -- expect changed paths matching: ({exts})\\n\""," f\"ORDER BY merged_at DESC\\n\""," f\"LIMIT {limit}\\n\"","class Candidate:"," repo: str"," number: int"," kind: str # \"pr\" | \"issue\""," title: str"," url: str"," body: str = \"\"","def fetch_search(query: str, *, token: str = \"\", http=None, per_page: int = 50) -> list[Candidate]:"," \"\"\"Run one GitHub issue/PR search. `http` is an injectable opener returning bytes"," (tests pass a fake; production uses urllib). Network-touching by nature — never called"," in CI. Maps results into Candidates; classification happens downstream in `signals`.\"\"\""," url = \"https://api.github.com/search/issues?\" + urllib.parse.urlencode("," {\"q\": query, \"per_page\": per_page}"," raw = (http or _urllib_get)(url, token)"," data = json.loads(raw)"," out: list[Candidate] = []"," for it in data.get(\"items\", []):"," is_pr = \"pull_request\" in it"," out.append(Candidate("," ecosystem=\"\","," repo=_repo_of(it.get(\"repository_url\", \"\")),"," number=it.get(\"number\", 0),"," kind=\"pr\" if is_pr else \"issue\","," title=it.get(\"title\", \"\"),"," url=it.get(\"html_url\", \"\"),"," body=it.get(\"body\") or \"\","," ))"," return out","def fetch_patch(repo: str, number: int, *, token: str = \"\", http=None) -> str:"," \"\"\"Fetch a PR's unified diff. `http` is an injectable getter returning the raw diff"," text (tests pass a fake; production asks GitHub for the `.diff` media type). Network by"," nature — never called in CI tests. Returns \"\" on any fetch error so one bad PR doesn't"," abort a mining run.\"\"\""," url = f\"https://api.github.com/repos/{repo}/pulls/{number}\""," return (http or _urllib_get_diff)(url, token)"," except Exception:"," return \"\"","def fetch_repo_languages(repos, *, token: str = \"\", http=None, batch: int = 100) -> dict:"," \"\"\"Enrich repos with their primary language via BATCHED GraphQL — ~100 repos per request"," (aliased `repository(...)` fields), so language is far cheaper than one REST call each."," Returns {repo: lowercased-language} (\"\" when unknown/missing). `http(body, token) -> bytes`"," is injectable for tests; production posts to the GraphQL endpoint (token required).\"\"\""," out: dict = {}"," uniq = list(dict.fromkeys(repos))"," for i in range(0, len(uniq), batch):"," chunk = uniq[i:i + batch]"," parts = []"," for j, repo in enumerate(chunk):"," owner, _, name = repo.partition(\"/\")"," parts.append("," f\"r{j}: repository(owner: {json.dumps(owner)}, name: {json.dumps(name)}) \""," \"{ primaryLanguage { name } }\""," )"," body = json.dumps({\"query\": \"query {\" + \" \".join(parts) + \"}\"}).encode()"," try:"," data = json.loads((http or _graphql_post)(body, token)).get(\"data\")"," except Exception:"," data = None"," if not data:"," # whole-batch failure (bad token / rate limit / 5xx / errors-only response):"," # OMIT these repos from the result so the caller leaves them unenriched and a"," # later run retries — never record them as a persisted \"unknown\"."," node = data.get(f\"r{j}\") if isinstance(data, dict) else None"," pl = (node or {}).get(\"primaryLanguage\") if isinstance(node, dict) else None"," out[repo] = ((pl or {}).get(\"name\") or \"\").lower()","def _graphql_post(body: bytes, token: str) -> bytes:"," req = urllib.request.Request("," \"https://api.github.com/graphql\", data=body, method=\"POST\","," headers={\"Authorization\": f\"Bearer {token}\", \"User-Agent\": \"leakmine\","," \"Content-Type\": \"application/json\"},"," with urllib.request.urlopen(req, timeout=30) as r: # noqa: S310 (https only)"," return r.read()","def _urllib_get_diff(url: str, token: str) -> str:"," req = urllib.request.Request(url, headers={"," \"Accept\": \"application/vnd.github.v3.diff\","," \"User-Agent\": \"leakmine\","," **({\"Authorization\": f\"Bearer {token}\"} if token else {}),"," })"," return r.read().decode(\"utf-8\", \"replace\")","def _urllib_get(url: str, token: str) -> bytes:"," \"Accept\": \"application/vnd.github+json\",","def _repo_of(repository_url: str) -> str:"," # https://api.github.com/repos/{owner}/{name} -> owner/name"," parts = repository_url.rstrip(\"/\").split(\"/\")"," return \"/\".join(parts[-2:]) if len(parts) >= 2 else repository_url",[1560,1561,1562,1563,1564,1565,1566,1567],{"_676":46,"_46":1580},{"_676":1569,"_161":1409,"_700":1578,"_701":1579},{"_676":46,"_46":1577},{"_676":1569,"_161":1195,"_700":1575,"_701":1576},{"_676":46,"_46":1574},{"_676":1569,"_161":694,"_700":1572,"_701":1573},{"_676":1569,"_161":694,"_700":1570,"_701":1571},{"_676":46,"_46":1568},"\n* No code changes were needed in this follow-up, so I did not create a new commit or PR.\n\n**Testing**\n* ✅ `python3 -m leakmine.tests.test_leakmine`\n* ✅ `python3 -O -m leakmine.tests.test_leakmine`\n* ✅ `git status --short`","repo_file_citation",613,626,580,586,"\n* Verified tests cover both failure paths: failed/errors-only GraphQL batches return an empty map, and enrichment records nothing on failure before succeeding on a later retry. ",186,190,"\n* Verified `enrich_languages` skips repos absent from the fetched language map, leaving failed lookups unenriched and retryable. ",146,159,"**Summary**\n* Verified the PR already contains the requested fix: `fetch_repo_languages` now treats whole-batch GraphQL failures or errors-only responses as omitted results rather than persisted `unknown` labels, allowing later retries. ",{"_92":326,"_42":84,"_96":326,"_1583":1584,"_1585":1586,"_1587":1588,"_132":-5,"_119":1589},[83],"input_items",[1608,1609,1610,1611,1612],"external_post_config",{"_1590":1591,"_263":196,"_1592":305,"_274":-5,"_1593":257,"_1594":-5,"_1595":-5,"_1596":1597,"_1598":1599,"_1600":-5,"_1601":-5,"_1602":1603,"_1604":15,"_1605":59,"_1606":54,"_1607":15},"user_model_slug","slug",1782626777.795363,"installation_id",131220309,"pr_number","repo_id","gitlab_service_account_project_id","review_trigger","comment_id",4825072873,"comment_type","issue","pr_write_guard_token","pending_pr_reaction_id","pending_comment_reaction_id",375638274,"comment_if_no_findings","comment_on_error","share_url","is_parallel_review",{"_92":407,"_96":326,"_401":1623},{"_312":313,"_314":315,"_1620":319,"_1621":129,"_322":-5,"_92":1622},{"_92":1613,"_314":1619,"_161":-5,"_1615":-5,"_1616":-5},{"_92":1613,"_314":1617,"_161":1409,"_1615":1618,"_1616":-5},{"_92":1613,"_314":1614,"_161":-5,"_1615":-5,"_1616":-5},"pull_request_comment","\n### 💡 Codex Review\n\nHere are some automated review suggestions for this pull request.\n\n**Reviewed commit:** `57846a0075`\n \n\n

ℹ️ About Codex in GitHub

","side","line","**_{_{![P2 Badge](https://img.shields.io/badge/P2-yellow?style=flat)}} Do not persist unknown languages after GraphQL failures**\n\nWhen the GraphQL request fails for a whole batch (for example missing/expired `GITHUB_TOKEN`, rate limiting, or a transient 5xx), this handler turns the failure into an empty `data` object, so `enrich_languages` records every repo in that batch as `classifier='language'`/`unknown`. Because the enrichment query skips candidates that already have a language label, a later rerun with a valid token will not retry those repos and `classify-store` will still fetch all their diffs, defeating the new budget-saving stage until the labels are manually deleted.\n\nUseful? React with 👍 / 👎.","RIGHT","\n\n\n[![Review Change Stack](https://storage.googleapis.com/coderabbit_public_assets/review-stack-in-coderabbit-ui.svg)](https://app.coderabbit.ai/change-stack/PhysShell/OwnAudit/pull/28?utm_source=github_walkthrough&utm_medium=github&utm_campaign=change_stack)\n\n\n\n\n> [!NOTE]\n> Currently processing new changes in this PR. This may take a few minutes, please wait...\n> \n>

\n>

⚙️ Run configuration

\n> \n> **Configuration used**: defaults\n> \n> **Review profile**: CHILL\n> \n> **Plan**: Pro Plus\n> \n> **Run ID**: `df2270c5-c17a-4286-bf3b-5edf54ef1102`\n> \n>

\n> \n>

\n>

📥 Commits

\n> \n> Reviewing files that changed from the base of the PR and between 07e6c31d9c768072427ff510a1daadc5c68cb1c0 and 57846a0075effbc28aab0e63e1da5f76b51ec501.\n> \n>

\n> \n>

\n>

📒 Files selected for processing (7)

✨ Finishing Touches

\n\n

📝 Generate docstrings

\n\n- [ ] Create stacked PR\n- [ ] Commit on current branch\n\n

🧪 Generate unit tests (beta)

\n\n- [ ] Create PR with unit tests\n- [ ] Commit unit tests in branch `claude/own-language-market-research-jgrrgm`\n\n

\n\n

\n\n\n\n\n---\n\n\n\n\n_{Comment `@coderabbitai help` to get the list of available commands.}\n\n","base_ref","head_ref","pull_request_info",[1624],{"_676":46,"_46":1625},"Good catch @chatgpt-codex-connector — fixed in 11da9fd. A whole-batch GraphQL failure (bad/expired token, rate limit, 5xx, or an errors-only response with no `data`) is now **omitted** from `fetch_repo_languages`'s result map instead of being coerced to `unknown`. `enrich_languages` only labels repos actually present in the map, so failed ones stay unenriched and a later run retries them — no more silently persisting `unknown` and skipping them forever. Genuine per-repo nulls (repo resolved, no `primaryLanguage`) are still recorded as `\"\"` so `classify-store` still attempts them.\n\nAdded two tests: failed/errors-only batch → empty map; and enrich records nothing on failure then succeeds on a subsequent working fetch. 36 tests pass under `python3` and `-O`.\n\n---\n_Generated by [Claude Code](https://claude.ai/code)_",[1831,1832],[],{"_681":682,"_46":683,"_684":685},{"_655":1822},{"_337":1821},{"_337":1661},{"_304":305,"_295":306,"_307":308,"_309":59,"_310":15,"_311":15,"_312":313,"_314":315,"_316":267,"_317":127,"_318":319,"_320":321,"_322":323,"_324":-5,"_325":-5,"_326":1660,"_328":329,"_330":331,"_332":333,"_334":335},{"_42":123,"_193":194,"_195":196,"_197":15,"_198":199,"_200":201,"_202":1649,"_204":-5,"_205":206,"_207":15,"_208":1650,"_210":199,"_211":199,"_212":1651,"_214":1652,"_216":1653,"_218":-5,"_219":220,"_221":222,"_223":224,"_225":1654,"_55":-5,"_119":227,"_228":1655,"_230":1656,"_232":233,"_234":15,"_235":15},{"_133":1642},[1636,1637,1638,1639,1640,1641],{"_148":171,"_150":165,"_152":166,"_154":163,"_156":172,"_158":159,"_160":32,"_161":-5},{"_148":170,"_150":165,"_152":166,"_154":163,"_156":167,"_158":159,"_160":32,"_161":-5},{"_148":168,"_150":165,"_152":166,"_154":163,"_156":169,"_158":159,"_160":32,"_161":-5},{"_148":164,"_150":165,"_152":166,"_154":163,"_156":167,"_158":159,"_160":32,"_161":-5},{"_148":162,"_150":151,"_152":153,"_154":163,"_156":157,"_158":159,"_160":32,"_161":-5},{"_148":149,"_150":151,"_152":153,"_154":155,"_156":157,"_158":159,"_160":32,"_161":-5},[1643,1644,1645],{"_177":1648,"_179":190},{"_177":1647,"_179":186},{"_177":1646,"_179":180},{"_181":182,"_92":183,"_119":184},{"_181":187,"_92":183,"_119":188},{"_181":191,"_92":183,"_119":192},{"_297":298,"_299":-5,"_300":-5,"_301":-5,"_302":-5,"_303":-5},[257],{"_257":1657},{"_239":240,"_241":242,"_243":244,"_245":246,"_247":248,"_249":250,"_251":252,"_253":254,"_255":256},{},[],{"_238":59},{"_236":59,"_237":199},{"_42":259,"_181":260,"_261":1658,"_263":196,"_264":265,"_266":267,"_268":269,"_270":271,"_272":273,"_274":64,"_275":15,"_276":15,"_277":59,"_278":59,"_279":59,"_280":15,"_281":-5,"_160":282,"_283":-5,"_55":1659},{"_289":290,"_42":291,"_92":292,"_293":294,"_295":296},{"_285":59,"_286":59,"_287":59,"_288":59},{"_289":290,"_181":290,"_336":-5,"_293":294,"_42":291},[1662,1663,1664,1665,1666,1667,1668,1669,1670,1671,1672,1673,1674,1675,1676,1677,1678,1679,1680,1681,1682,1683,1684,1685,1686,1687,1688,1689,1690,1691,1692,1693,1694,1695,1696,1697,1698],{"_154":628,"_377":1817},{"_154":392,"_377":1814},{"_154":618,"_377":1811},{"_154":396,"_377":1806},{"_154":396,"_377":1802},{"_154":396,"_377":1798},{"_154":411,"_377":1793},{"_154":411,"_377":1789},{"_154":396,"_377":1785},{"_154":396,"_377":1783},{"_154":411,"_377":1779},{"_154":396,"_377":1777},{"_154":396,"_377":1775},{"_154":411,"_377":1771},{"_154":396,"_377":1769},{"_154":411,"_377":1765},{"_154":396,"_377":1761},{"_154":396,"_377":1759},{"_154":411,"_377":1753},{"_154":396,"_377":1751},{"_154":396,"_377":1749},{"_154":396,"_377":1747},{"_154":411,"_377":1743},{"_154":411,"_377":1739},{"_154":396,"_377":1735},{"_154":396,"_377":1733},{"_154":411,"_377":1729},{"_154":396,"_377":1727},{"_154":396,"_377":1725},{"_154":411,"_377":1720},{"_154":396,"_377":1718},{"_154":411,"_377":1714},{"_154":396,"_377":1710},{"_154":411,"_377":1708},{"_154":396,"_377":1704},{"_154":392,"_377":1702},{"_154":376,"_377":1699},{"_379":380,"_87":1700},{"_382":383,"_384":385,"_103":-5,"_42":386,"_387":1701,"_389":390,"_391":113},[],{"_391":1703,"_379":380},{"_92":395},{"_398":1705,"_379":380,"_400":386},{"_401":1706,"_403":-5,"_42":-5,"_404":-5,"_405":406,"_96":93,"_92":407},[1707],{"_46":409,"_92":410},{"_398":1709,"_379":380,"_400":386},{"_42":414,"_405":406,"_46":409,"_92":415},{"_398":1711,"_379":380,"_400":386},{"_401":-5,"_418":419,"_42":-5,"_404":-5,"_420":1712,"_92":422},[1713],{"_46":424,"_92":425},{"_398":1715,"_379":380,"_400":386},{"_401":1716,"_42":429,"_420":1717,"_92":422},[],[424],{"_398":1719,"_379":380,"_400":386},{"_433":434,"_42":-5,"_404":-5,"_435":436,"_92":437},{"_398":1721,"_379":380,"_400":386},{"_440":441,"_442":443,"_444":1722,"_446":447,"_384":32,"_448":32,"_42":434,"_449":450,"_451":452,"_391":113,"_92":453},[1723,1724],{"_442":460,"_181":461,"_161":462,"_92":459},{"_442":456,"_181":457,"_161":458,"_92":459},{"_398":1726,"_379":380,"_400":386},{"_465":466,"_433":434,"_42":-5,"_404":-5,"_181":467,"_468":-5,"_92":469},{"_398":1728,"_379":380,"_400":386},{"_433":472,"_42":-5,"_404":-5,"_435":473,"_92":437},{"_398":1730,"_379":380,"_400":386},{"_440":476,"_442":477,"_444":1731,"_446":447,"_384":32,"_448":32,"_42":472,"_449":479,"_451":452,"_391":113,"_92":453},[1732],{"_442":481,"_92":482},{"_398":1734,"_379":380,"_400":386},{"_465":485,"_433":472,"_42":-5,"_404":-5,"_181":467,"_468":-5,"_92":469},{"_398":1736,"_379":380,"_400":386},{"_401":-5,"_418":488,"_42":-5,"_404":-5,"_420":1737,"_92":422},[1738],{"_46":491,"_92":425},{"_398":1740,"_379":380,"_400":386},{"_401":1741,"_42":495,"_420":1742,"_92":422},[],[491],{"_398":1744,"_379":380,"_400":386},{"_440":499,"_442":500,"_444":1745,"_446":447,"_384":502,"_448":32,"_42":503,"_449":504,"_451":452,"_391":113,"_92":453},[1746],{"_442":506,"_92":482},{"_398":1748,"_379":380,"_400":386},{"_433":503,"_42":-5,"_404":-5,"_435":509,"_92":437},{"_398":1750,"_379":380,"_400":386},{"_465":512,"_433":503,"_42":-5,"_404":-5,"_181":467,"_468":-5,"_92":469},{"_398":1752,"_379":380,"_400":386},{"_433":515,"_42":-5,"_404":-5,"_435":516,"_92":437},{"_398":1754,"_379":380,"_400":386},{"_440":519,"_442":520,"_444":1755,"_446":447,"_384":522,"_448":32,"_42":515,"_449":523,"_451":452,"_391":113,"_92":453},[1756,1757,1758],{"_442":533,"_181":461,"_161":462,"_92":459},{"_442":532,"_181":457,"_161":458,"_92":459},{"_442":527,"_161":528,"_529":530,"_92":531},{"_398":1760,"_379":380,"_400":386},{"_465":536,"_433":515,"_42":-5,"_404":-5,"_181":467,"_468":-5,"_92":469},{"_398":1762,"_379":380,"_400":386},{"_401":-5,"_418":539,"_42":-5,"_404":-5,"_420":1763,"_92":422},[1764],{"_46":542,"_92":425},{"_398":1766,"_379":380,"_400":386},{"_401":1767,"_42":546,"_420":1768,"_92":422},[],[542],{"_398":1770,"_379":380,"_400":386},{"_433":550,"_42":-5,"_404":-5,"_435":551,"_92":437},{"_398":1772,"_379":380,"_400":386},{"_440":-5,"_442":554,"_444":1773,"_446":447,"_384":32,"_448":32,"_42":550,"_449":556,"_451":452,"_391":113,"_92":453},[1774],{"_442":558,"_161":559,"_529":560,"_92":531},{"_398":1776,"_379":380,"_400":386},{"_465":563,"_433":550,"_42":-5,"_404":-5,"_181":467,"_468":-5,"_92":469},{"_398":1778,"_379":380,"_400":386},{"_433":566,"_42":-5,"_404":-5,"_435":567,"_92":437},{"_398":1780,"_379":380,"_400":386},{"_440":570,"_442":571,"_444":1781,"_446":447,"_384":32,"_448":32,"_42":566,"_449":573,"_451":452,"_391":113,"_92":453},[1782],{"_442":575,"_92":482},{"_398":1784,"_379":380,"_400":386},{"_465":578,"_433":566,"_42":-5,"_404":-5,"_181":467,"_468":-5,"_92":469},{"_398":1786,"_379":380,"_400":386},{"_401":-5,"_418":581,"_42":-5,"_404":-5,"_420":1787,"_92":422},[1788],{"_46":584,"_92":425},{"_398":1790,"_379":380,"_400":386},{"_401":1791,"_42":588,"_420":1792,"_92":422},[],[584],{"_398":1794,"_379":380,"_400":386},{"_592":-5,"_401":1795,"_42":594,"_92":595},[1796],{"_46":597,"_598":1797,"_92":46},[],{"_398":1799,"_379":380,"_400":386},{"_401":1800,"_403":-5,"_42":-5,"_404":-5,"_405":-5,"_96":326,"_92":407},[1801],{"_46":597,"_92":604},{"_398":1803,"_379":380,"_400":386},{"_401":1804,"_403":-5,"_42":-5,"_404":-5,"_405":-5,"_96":326,"_92":407},[1805],{"_46":609,"_92":604},{"_398":1807,"_379":380,"_400":386},{"_401":1808,"_403":-5,"_42":-5,"_404":-5,"_405":-5,"_96":613,"_92":407},[1809,1810],{"_46":617,"_92":604},{"_46":616,"_92":604},{"_379":380,"_87":1812},{"_103":-5,"_42":386,"_387":1813,"_389":390,"_391":622},[],{"_391":1815,"_379":380},{"_625":1816,"_92":627},[],{"_630":1818},{"_632":-5,"_633":-5,"_634":635,"_636":390,"_446":447,"_637":15,"_638":-5,"_639":-5,"_640":-5,"_641":642,"_42":380,"_643":644,"_181":-5,"_645":-5,"_161":646,"_647":199,"_648":-5,"_451":649,"_391":1819,"_651":-5,"_77":1820,"_653":390},{"_92":395},[],[],[1823],{"_42":658,"_659":1824,"_661":662,"_663":662,"_401":1825,"_391":665,"_403":-5,"_666":32,"_667":1826,"_669":670,"_671":-5},{"_96":326,"_181":-5,"_667":1830},{"_676":46,"_677":1829},{"_672":1827},{"_674":1828},[],[679],{},{"_92":407,"_96":93,"_401":1846},{"_92":688,"_689":1833},[1834,1835,1836],{"_161":1409,"_695":-5,"_696":1843,"_698":-5},{"_161":1195,"_695":-5,"_696":1840,"_698":-5},{"_161":694,"_695":-5,"_696":1837,"_698":-5},[1838],{"_700":522,"_701":702,"_401":1839,"_704":59},[705,199,706,199,707,708,709,710,711,712,713,714,715,716,717,199,718,719,720,199,721,722,723,199,199,724,725,726,199,199,727,199,728,729,730,731,732,733,734,735,736,737,738,739,740,741,711,199,199,742,743,744,745,746,747,748,749,750,751,752,199,199,753,754,755,756,757,758,759,760,761,199,199,762,763,764,765,766,758,767,768,199,199,769,199,770,771,772,773,774,775,199,199,776,777,778,779,780,781,199,199,782,783,784,785,786,773,787,199,199,788,789,790,791,792,793,794,199,199,795,199,796,797,199,199,798,799,800,801,199,802,803,804,805,199,199,806,807,808,809,810,811,199,812,813,814,815,816,817,199,199,818,819,820,821,822,823,734,735,824,825,741,711,199,199,826,827,828,829,830,831,832,833,834,199,199,835,199,836,837,838,839,840,841,842,199,199,843,844,845,846,847,848,849,850,851,852,853,854,855,856,857,858,848,859,860,199,861,862,863,864,865,199,199,866,199,867,868,869,870,871,872,873,874,875,876,877,878,879,880,199,199,881,882,883,884,885,886,869,887,888,889,874,890,891,892,893,199,199,894,895,896,897,898,899,900,901,874,902,903,904,905,900,906,874,907,199,199,908,909,869,910,911,874,875,912,913,914,199,199,915,199,916,917,918,919,920,921,922,923,924,925,926,927,928,929,930,931,932,933,934,935,199,199,936,199,937,938,939,940,941,942,943,944,945,199,946,947,948,949,950,951,952,953,954,955,956,199,199,957,958,959,947,948,960,952,961,962,199,199,963,199,964,965,966,967,199,968,969,970,971,199,199,972,973,974,975,976,977,978,979,199,199,980,981,982,983,984,985,986,987,988,989,990,991,992,993,199,199,994,995,996,997,998,999,926,1000,1001,1002,199,1003,1004,199,1005,1006,199,1007,1008,1009,1010,1011,1012,1013,1014,1015,199,199,1016,199,1017,1007,1018,1019,1020,1021,1022,1023,1024,1025,1026,199,199,1027,1028,1029,1007,1030,1031,1032,1033,1034,1035,1036,199,199,1037,1038,1007,1039,1040,1023,1041,1042,199,199,1043,1044,1007,1045,1046,1047,1048,1049,1050,1051,199,199,1052,1053,1054,969,1055,1056,1057,1058,1059,1060,1061,1062,1063,1064,1065,199,199,1066,1067,1068,1069,1070,1071,1072,1073,1074,1075,1046,1076,1077,1050,1078,199,199,1079,1080,1081,1082,1083,1084,1085,1086,1087,1088,199,199,1089,1090,1091,1092,1093,1094,1095,1096,1097,1007,1098,1099,1100,1101,1102,1103,199,199,1104,1105,1106,1107,1108,1109,1110,1111,199,199,1112,1007,1113,1114,1115,1116,1117,1118,1119,1120,1121,1122,199,1005,1123,1006,199,1124,1125,1126,1127,1128,1129,1130,1131,1132,1133,1134,199,1135,1136,1137,1138,199,199,1139,1140,1141,1142,1143,1144,1145,1146,1147,1148,1149,1150,199,1151,1152,1153,1154,1155,1156,1157,199,199,1158,1007,1159,1160,1161,1162,199,1163,1164,199,1165,1166,1167,1122,199,1005,1168,1169,199,1124,1170,1171,199,199,1172,1007,1173,1114,1118,1174,1175,1176,1177,1178,1179,1180,1181,1182,199,199,1183,1184,1185,1075,1046,1186,1077,1050,1187,199,199,1188,1189,1190,1191,1192,199,199,1193,1194],[1841],{"_700":522,"_701":1198,"_401":1842,"_704":59},[1200,199,1201,1202,1203,1204,1205,199,1206,1207,1208,1209,199,1210,1211,711,1212,199,1213,1214,199,1215,199,199,1216,1217,1218,1219,1220,1221,1222,199,1223,1224,1225,1226,1227,1228,1229,1230,1231,1232,1229,1233,1229,1234,1235,1236,1237,1238,1239,1240,1241,1242,1243,1244,199,199,1245,1246,1247,1248,1249,1250,1251,1252,1253,1254,1255,1256,1257,1258,1259,1260,1261,1262,199,1263,1264,199,1265,1266,1267,1268,1269,1270,1271,1272,1273,1274,199,1275,1276,1277,1278,1272,1279,199,1280,1281,1272,1282,1283,1284,1285,1286,1287,1288,1289,1290,1291,1292,1293,1294,1295,1296,1297,1298,1299,1300,1301,1302,1303,1304,1305,199,199,1216,1306,1218,1307,1308,1309,1310,199,1223,1227,1311,1229,1312,1313,1314,1229,1315,1236,1316,1317,1244,199,199,1318,1319,1320,1247,1248,1321,1322,1323,1324,1325,1326,199,1327,1328,1329,1330,1331,1332,1333,1334,1335,1336,1337,1338,1339,1340,1341,874,1342,1343,1344,1345,1346,1347,1348,1349,1350,1351,1352,1353,1354,1355,1356,1357,1358,1359,1360,1361,1362,1363,1364,1365,1305,199,199,1366,1319,1320,1247,1248,1251,1252,1322,1255,1256,1367,1368,1369,1370,199,1371,1372,1373,199,1374,1375,1376,1377,1378,1379,1332,1262,1380,1381,1382,1383,1384,1336,1385,1386,1387,1388,874,1342,1343,1344,1345,1389,1390,1391,1392,1267,1268,1393,1394,1395,1396,1397,1398,1399,1400,1401,1396,1402,1403,1404,1405,1406,1407,1408,1365,1305],[1844],{"_700":522,"_701":1412,"_401":1845,"_704":59},[1414,199,1415,199,1416,1417,1418,1419,1420,1421,1422,199,1423,1424,1425,711,1212,199,712,1426,1427,1428,199,1429,199,199,1430,1431,1432,1433,1434,1435,1436,1437,199,199,1438,1320,1247,1439,1440,1441,1442,1443,1444,199,1445,1446,1447,1332,1433,1448,1449,1450,1451,1452,1453,1454,874,1455,1456,1457,1458,1459,1460,1461,1462,1463,1464,1465,1466,1467,1468,1469,1470,1471,1472,1473,874,199,199,1216,1474,1218,1475,1476,1477,1478,1479,1480,199,199,1481,1482,1483,1484,1485,1486,874,1487,1488,1489,1490,1491,1492,1493,1494,1495,1496,1497,1498,1499,1500,1501,199,199,1502,1503,1504,1505,1506,1507,1046,1508,1509,1510,199,199,1511,1512,1513,1514,1515,1516,1517,1518,1519,1520,1521,1522,1523,1524,1525,1526,1527,1528,1529,1530,1531,1532,1533,1534,1535,1396,1521,1536,1537,1538,1501,199,199,1539,1540,1541,1542,1543,874,1544,1545,199,199,1546,1547,1548,1549,1550,1551,1544,1552,199,199,1553,1547,1554,1549,1550,1551,1544,1545,199,199,1555,1556,1557,1558],[1847,1848,1849,1850,1851,1852,1853,1854],{"_676":46,"_46":1580},{"_676":1569,"_161":1409,"_700":1578,"_701":1579},{"_676":46,"_46":1577},{"_676":1569,"_161":1195,"_700":1575,"_701":1576},{"_676":46,"_46":1574},{"_676":1569,"_161":694,"_700":1572,"_701":1573},{"_676":1569,"_161":694,"_700":1570,"_701":1571},{"_676":46,"_46":1568},[1857,1858,1859,1860,1861],{"_1590":1591,"_263":196,"_1592":305,"_274":-5,"_1593":257,"_1594":-5,"_1595":-5,"_1596":1597,"_1598":1599,"_1600":-5,"_1601":-5,"_1602":1603,"_1604":15,"_1605":59,"_1606":54,"_1607":15},{"_92":407,"_96":326,"_401":1862},{"_312":313,"_314":315,"_1620":319,"_1621":129,"_322":-5,"_92":1622},{"_92":1613,"_314":1619,"_161":-5,"_1615":-5,"_1616":-5},{"_92":1613,"_314":1617,"_161":1409,"_1615":1618,"_1616":-5},{"_92":1613,"_314":1614,"_161":-5,"_1615":-5,"_1616":-5},[1863],{"_676":46,"_46":1625},"task_e_6a40b9d900a883209d0cde88d5f26a7f",1782626777.002677,"GitHub Mention: leakmine: language enrichment (batched GraphQL) to cut cross-language diff waste","has_generated_title","has_unread_turn","denormalized_metadata",{"_1895":32,"_1896":1897,"_1898":-5,"_1899":-5,"_1890":1900,"_1901":-5,"_1902":-5,"_1903":-5,"_1904":-5},"task_status_display",{"_1883":1884,"_1885":196,"_130":127,"_1886":-5,"_1887":1888},"creator_workspace","external_pull_requests",[1876],{"_42":1877,"_1878":83,"_1879":1880,"_1881":129},"github-1280321876-28","assistant_turn_id","pull_request",{"_304":305,"_295":306,"_307":308,"_309":59,"_310":15,"_311":15,"_312":313,"_314":315,"_316":267,"_317":127,"_318":319,"_320":321,"_322":323,"_324":-5,"_325":-5,"_326":1882,"_328":329,"_330":331,"_332":333,"_334":335},"codex_updated_sha",{"_289":290,"_181":290,"_336":-5,"_293":294,"_42":291},"latest_turn_status_display",{"_1889":83,"_112":113,"_1890":-5,"_94":1891,"_1892":-5,"_1893":-5,"_1894":1888,"_114":-5},"environment_label","fix_task_id","initial_intent","pr","turn_id","diff_stats",[],"image_asset_pointer_citations","issue_counts","intent","num_follow_ups","repo_name_map",{},"last_user_turn_created_at","last_turn_status",{"_1905":-5,"_1906":-5,"_1907":-5},"last_pushed_turn_id","start_branch_name","pushed_branch_name","cached_pull_request_data","files_modified","lines_added","lines_removed","dd",{"_1919":1920,"_1921":1922},"disablePrefetch","shouldPrefetchAccount","shouldPrefetchUser","shouldPrefetchSystemHints","promoteCss","disableSSR","stripModulepreloadImports","statsigGateEvaluationsPromise",["P",1918],"traceId","1429466103795019079","traceTime",1785163004035]