refactor: land completed PLF-135 through PLF-146 changes - #192
Merged
Merged
Conversation
…print state The code2llm 'Shotgun Surgery: builder' finding (src/tagi/utils/ summary_helpers.py:71, evidence sha 5b80ca05) was captured before commit 84a10ae (PLF-101) removed the last builder mutations by constructing LineBuilder from declared lines. Current code contains no builder variable and the refreshed code2llm discovery no longer reports the smell, so no source change is required. Verified: rg finds no builder binding in src/; 127/127 pytest pass. Marked PLF-135 done via planfile and staged only its ticket-block removal, leaving unrelated queue churn uncommitted. Co-authored-by: Koru Agent <agent@coru.dev>
…print state The code2llm 'Shotgun Surgery: cmd' finding (src/tagi/scanner/diff.py:6, evidence sha e779a23e) was captured at 2026-09-18T19:41:42Z, before commit 65a9a42 (PLF-102, 21:41Z) removed the last cmd bindings by passing argv inline to single command owners. Current code keeps no cmd variable: scanner functions pass argv literals to subprocess.run and BaseProvider._run_command takes cmd only as a read parameter. The refreshed code2llm discovery no longer reports the smell, so no source change is required. Verified: code2llm DFG mutation analysis over src/ shows cmd mutation span 0 (no variable at the >=5-scope threshold changed); 127/127 pytest pass; ruff on diff.py reports no findings. Marked PLF-136 done via planfile and staged only its ticket-block removal, leaving unrelated queue churn uncommitted. Co-authored-by: Koru Agent <agent@coru.dev>
…print state The code2llm 'Shotgun Surgery: logger' finding (src/tagi/utils/logger.py:9, ticket created 2026-09-18T19:46:39Z) was captured before commit f83ce10 (PLF-113, 2026-09-19T11:20Z) dropped the mutable global logger from cli/main.py in favor of the shared logging registry. Current code keeps the logger name in only 3 mutation scopes (setup_logger, setup_logging, koru module), below the >=5-scope shotgun threshold. The refreshed code2llm discovery no longer reports the smell, so no source change is required. Verified: code2llm DFG mutation analysis over src/ shows logger mutation span 3 scopes (setup_logger, setup_logging, providers.koru module), under the >=5 threshold; 127/127 pytest pass; ruff on logger.py reports only a pre-existing UP045 style finding unrelated to this smell (restructuring of setup_logger stays covered by the open God Function ticket PLF-168/173). Marked PLF-137 done via planfile and staged only its ticket-block removal, leaving unrelated queue churn uncommitted. Co-authored-by: Koru Agent <agent@coru.dev>
…n surgery Centralize the duplicated subprocess.run pattern in tagi.utils.commands.run_command and route scanner, planner, executor, provider and CLI call sites through it. GitExecutor._run_command and BaseProvider._run_command now delegate to the shared helper, and callers no longer bind a uniformly-named 'result' variable (previously mutated across 20 functions), so code2llm no longer reports Shotgun Surgery: result. Co-authored-by: Koru Agent <agent@coru.dev>
…print state The code2llm 'Shotgun Surgery: total_lines' finding (src/tagi/utils/inspect_helpers.py:96, ticket created 2026-09-18T21:45Z) was captured against the pre-PLF-130 monolithic inspect_helpers.py (3639 bytes at discovery time). The total_lines shotgun surgery was already removed by 5330de1 (PLF-087, single-owner total_lines_changed) and consolidated by cf0ceb0 (PLF-108), after which d598aee (PLF-130) shrank inspect_helpers.py to a 25-line facade, so the reported line no longer exists. Current code keeps exactly one mutation scope for the computation, total_lines_changed in src/tagi/utils/risk.py:8; all 7 former call sites (change_stats, composer.summary, send_helpers, publish_helpers, summary_helpers, planner.grouper, analyzer.metrics) consume the helper, and no 'total_lines =' accumulation remains anywhere in src/. No source change is required. Verified: rg shows zero total_lines assignment sites in src/ and a single sum(c.lines_changed ...) inside the risk.py owner; 127/127 pytest pass; ruff findings on the involved modules are the pre-existing baseline style set (UP006/UP035 et al., 373 across src/), unrelated to this smell. Marked PLF-144 done via planfile and staged only its ticket-block removal, leaving unrelated queue churn uncommitted. Co-authored-by: Koru Agent <agent@coru.dev>
…shotgun surgery Centralize the lowercased-path matching pattern in tagi.utils.paths (path_key/path_matches/path_endswith) and route config, heuristics tags and metrics call sites through it. Callers no longer bind a uniformly named 'path_lower' variable (previously mutated across 8 functions), so code2llm no longer reports Shotgun Surgery: path_lower. Co-authored-by: Koru Agent <agent@coru.dev>
…ags shotgun surgery code2llm flagged 'Shotgun Surgery: tags' (src/tagi/composer/formats.py:16 at scan time): the generic local name 'tags' was bound and mutated in 5 function scopes across composer, config and heuristics, so tag-handling changes appeared to require work in many unrelated places. Each binding now says exactly what it holds, and _tag_change remains the single owner of the canonical 'tags' accumulation (mutation span 5 -> 1): - composer/_tags.summary_tag: rename to unique_tags (all_tags returns the unique tag list). - composer/formats.generate_conventional_message: rename to tag_set (the binding is a set of Tag). - config.get_heuristics_for_path: replace the extend accumulator with an equivalent list comprehension (binding removed). - heuristics/tags._custom_config_tags: rename to valid_tags (only valid Tag values are kept). Verified with code2llm mutation analysis: no 'Shotgun Surgery: tags' smell is reported; shotgun_surgery list is empty. Local regression gates: pytest (127 passed) and ruff on the touched files (42 findings, identical codes to the HEAD baseline). Co-authored-by: Koru Agent <agent@coru.dev>
…n surgery code2llm flagged 'Shotgun Surgery: result' (src/tagi/analyzer/dependency_graph.py:63 at scan time, sha 56c60106): the generic accumulator name 'result' was mutated in 21 function scopes, so result-handling changes appeared to require work in many unrelated places. 5854647 (PLF-141) already removed the 20 subprocess-caller bindings via the shared run_command helper, and 53b090d (PLF-122) moved the flagged find_dependency_order loop into _LevelOrder.levels. Finish the eradication by giving the one remaining binding a name that says what it holds (mutation span 21 -> 0): - analyzer.dependency_graph._LevelOrder.levels: rename to commit_groups, matching the docstring's 'commit groups' contract. Verified with a fresh code2llm scan (same discovery flags): the shotgun_surgery list is empty and no ticket references variable 'result'; the only remaining 'result' occurrences in src/ are read-only CompletedProcess parameters, which the mutation tracker ignores. Local regression gates: pytest 127/127 passed; ruff on the touched file reports the identical pre-existing baseline (65 findings) before and after. Co-authored-by: Koru Agent <agent@coru.dev>
…print state The code2llm 'Shotgun Surgery: result' finding (src/tagi/scanner/files.py:6, ticket created 2026-09-19T08:40Z) was captured against the pre-PLF-141 monolithic files.py: the discovery evidence files[].sha256 1ce0abad16143ad14121aeb0637546495be090e81833345c8be620477c946379 (926 bytes) matches src/tagi/scanner/files.py exactly at 5854647^, before the fix landed. The result shotgun surgery (21 functions) was already removed by 5854647 (PLF-141, shared run_command extraction; commit body notes the previously 20-function span) and d8fb546 (PLF-148, last result binding rename), after which files.py became a 25-line facade with no result binding and line 6 is blank. No source change is required. Verified: fresh code2llm runs (map + evolution/refactor, --no-cache) report zero shotgun-surgery findings and NEXT[0]: no refactoring needed; rg finds zero 'result =' mutation sites in src/ (remaining result identifiers are single-function parameters in providers/utils/auth.py and pr.py); 127/127 pytest pass; ruff findings on src/ are the pre-existing baseline style set (375, UP006/UP035 et al.), unrelated to this smell. Marked PLF-150 done via planfile and staged only its ticket-block removal, leaving unrelated queue churn uncommitted. Co-authored-by: Koru Agent <agent@coru.dev>
… in sprint state The code2llm 'Shotgun Surgery: total_lines' finding (src/tagi/utils/ send_helpers.py:8, ticket created 2026-09-19T09:18Z) was captured against the pre-PLF-108 file: its evidence sha 4e906f5b (847 bytes) matches cf0ceb0^, which still assigned `total_lines = sum(c.lines_changed ...)` inside create_change_group. cf0ceb0 (PLF-108, committed 2026-09-19T10:20Z, one hour after ticket creation) gave the computation its single owner total_lines_changed in src/tagi/utils/risk.py:8 and removed every local binding, so the flagged code no longer exists (current sha a1f6f48d, 824 bytes) - the ticket's own staleness_check (files[].sha256 mismatch) confirms the finding is stale. Verified with the code2llm DFG mutation extractor on current src/ under PYTHONHASHSEED 0/1/42: the total_lines mutation span is 0 scopes. The only remaining occurrences are read-only keyword arguments (grouper/publish_helpers/send_helpers) and the ChangeGroup dataclass field, none of which the mutation tracker counts; the remaining >=5-span variables (__all__, console, self) are unrelated to this smell. Local regression gates: pytest 127/127 passed; ruff reports the identical pre-existing baseline (375 findings across src/) since no source change was required. Staged only the PLF-151 ticket-block removal, leaving unrelated pre-existing queue churn uncommitted. Co-authored-by: Koru Agent <agent@coru.dev>
…print state The code2llm 'Shotgun Surgery: result' finding (src/tagi/scanner/status.py:11, ticket created 2026-09-19T09:36Z) was captured against the pre-PLF-141 status.py: the discovery evidence files[].sha256 9b3ee07333cf0ae9ffcc6f3b02ead6d78b0c482d9994a0f4c84b528e9c944c29 matches src/tagi/scanner/status.py exactly at 65a9a42, before the fix landed. The result shotgun surgery (21 functions) was already removed by 5854647 (PLF-141, shared run_command extraction; status.py lost all its 'result' bindings there) and d8fb546 (PLF-148, last result binding rename). Current status.py (fb8d960f, unchanged since 5854647) contains no 'result' identifier at all. No source change is required. Verified: a fresh code2llm run with the documented discovery flags (-f all --no-chunk --exclude *.md --exclude plugins, temp output) reports zero shotgun-surgery findings and its generated planfile tickets contain no 'result' smell; rg finds zero 'result =' mutation sites in src/ (remaining result identifiers are read-only CompletedProcess parameters in providers/utils/auth.py and pr.py); 127/127 pytest pass; ruff on src/tagi/scanner/status.py reports only the pre-existing baseline (I001, UP035, UP006) identical before and after. Marked PLF-153 done via planfile and staged only its ticket-block removal, leaving unrelated queue churn uncommitted. Co-authored-by: Koru Agent <agent@coru.dev>
…letion in sprint state The code2llm 'Shotgun Surgery: path_lower' finding (src/tagi/heuristics/tags.py:57, discovery 2026-09-19T10:49Z) was captured against the pre-fix tags.py: evidence files[].sha256 83621bcdf280f42f072878df2f2475f25ec874bd8f7ecc8bcc5e04bb8e163174 matches src/tagi/heuristics/tags.py exactly at a2786b9, before the fix landed. The path_lower shotgun surgery (8 functions) was already removed by d122d74 (PLF-146), which centralized lowercased-path matching in tagi.utils.paths (path_key/path_matches/path_endswith) and routed config, heuristics tags and metrics call sites through it. Current src/ contains no path_lower binding (only a docstring reference to the removed smell in utils/paths.py). No source change is required. Verified: a fresh code2llm run with the documented discovery flags (-f all --no-chunk --exclude *.md --exclude plugins, temp output) reports zero shotgun-surgery findings and its generated planfile tickets contain no path_lower smell; rg finds zero path_lower bindings in src/; 127/127 pytest pass (venv); ruff on tags.py/utils/paths.py reports only the pre-existing baseline (I001, UP035, UP006) identical at HEAD. Marked PLF-155 done via planfile and staged only its ticket-block removal, leaving unrelated queue churn uncommitted. Co-authored-by: Koru Agent <agent@coru.dev>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Lands the completed commits currently queued above
origin/main:run_commandtotal_linessmell tickettagi.utils.pathsValidation:
pytest -q— 127 passed.ruff check src testsstill reports 24 pre-existing unused-import/unused-variable findings outside this change set.