Use SQL-aware source cleaning for duplication analysis #17

Problem

FamixRepCLikeCleaner retains SQL -- comments and truncates SELECT 'https://example.test'; at //. Even after source anchors are fixed, duplicate results are semantically unreliable.

Found during the 2026-09-20 FAMIXNGSQL/Moose 13 compatibility audit.

Implementation scope

Implement/select a SQL-aware cleaner preserving quoted identifiers, escaped strings and dollar-quoted bodies, handling SQL line/block comments, and retaining source line mapping. Depends on the source-anchor fix for end-to-end validation.

Implementation PR target: deem0n/FAMIXNGSQL, in the deem0n fork first. Keep this issue separate from the other v3.0.0 fixes. The PR must explain the previous failure, resulting behavior, compatibility implications, and actual validation.

Acceptance criteria and tests

Line comments; nested block comments; single/double/dollar quoting; // inside literals; multiline literals and line mapping; positive/negative duplicate fixtures; fragment display.

• Regression tests demonstrate the reported problem and pass with the fix.
• Relevant existing tests pass; validation environment and limitations are recorded in the PR.
• A separate implementation PR is linked here and reviewed.
• Integrated v3.0.0 acceptance is verified before this release blocker is closed.

Release

Part of the v3.0.0 milestone. An open PR is not a completed fix; release qualification is tracked separately.

Implementation

#30

Combined qualification: #32. The implementation is available for review; this issue remains open until its integrated release acceptance is verified.