Skip to content

Move the maintenance notes out of the evaluation repo - #56

Merged
sqlparser merged 1 commit into
masterfrom
move-out-maintenance-notes
Aug 25, 2026
Merged

Move the maintenance notes out of the evaluation repo#56
sqlparser merged 1 commit into
masterfrom
move-out-maintenance-notes

Conversation

@sqlparser

Copy link
Copy Markdown
Owner

docs/maintenance-notes.md was maintainer history — a 31 KB account of the 2026-07 cleanup, opening with its own disclaimer that "none of this is needed to use the demos".

It was linked six times from the README, including second in the top nav bar. So the earliest thing a visitor could click, before running anything, was a table reading: 28 open Dependabot advisories, 25 committed jars, 3 tests failing and documented as expected, 263 files whose path contradicted their own package line. Then the section titles — "The 3 known failures that were not", "four breakages and a merge", "Two demos that crashed instead of explaining themselves".

All of that is fixed, and the candour is what made the document good. But this repository is what people clone to decide whether to buy the parser, and a public catalogue of past defects in the nav bar works against that. The audience for the content is whoever maintains this repo next — not the audience the repo is for.

It also showed the cost of being on the ALLOW list

check-stale-docs.sh had to exempt it — its job is naming things that were deleted. Which meant the one guard we have could not see it drift, and it had:

ALLOW is now README.md alone, with a comment explaining why to keep it that way.

Where it went

docs/gsp_demo_java-maintenance-notes.md in the private sqlparser/gsp_java repo, with a header recording where it came from, the two facts corrected on the way over, and that its change log stops at 2026-08-12 (so it does not cover #51#55). That copy is currently uncommitted in that checkout — committing to a second repository is your call, not something to fold into this PR. This repo’s git log keeps the full history either way.

Nothing is lost from the README

The two deep links that carried substance are inlined where they were referenced, so both warnings now stand on their own:

  • the four package roots and 263 contradicting files, at the -Dexec.mainClass note
  • all four pom_dlineage.xml breakages, at the "don’t add a second POM" warning

Two broken links, fixed on the way

Nothing checks relative links, and two had rotted: columnImpact pointed at ../dlineage, which from that directory resolves to antiSQLInjection/dlineage; search pointed at ./visitors from inside search/. Every relative link in all 64 tracked markdown files resolves now.

I checked that with a throwaway script rather than a committed one. Wiring a link check into build.yml alongside check-stale-docs.sh is the obvious follow-up — it found two real breakages in seconds — but that is another permanent CI step, so I left it out of this PR.

Verified

156 tests · stale-docs self-test 5/5 and clean scan over 63 files · licensed-only guard 3/3 · dlineage jar in JSON and XML · parser version consistent at 4.2.6.

🤖 Generated with Claude Code

https://claude.ai/code/session_01UK3dBXFXqiDRYJxz1YxWCr

docs/maintenance-notes.md was maintainer history: a 31 KB account of the
2026-07 cleanup, opening with its own disclaimer that "none of this is needed
to use the demos". It was linked six times from the README, including second in
the top nav bar, so the earliest thing a visitor could click -- before running
anything -- was a table reading 28 open Dependabot advisories, 25 committed
jars, 3 tests failing and documented as expected, and 263 files whose path
contradicted their own package line.

All of that is fixed, and the candour is what made the document good. But this
repository is what people clone to decide whether to buy the parser, and a
public catalogue of past defects in the nav bar works against that. The
audience for the content is whoever maintains this repo next, which is not the
same audience the repo is for.

It also demonstrated the cost of being on check-stale-docs.sh's ALLOW list. It
had to be there -- its job is naming things that were deleted -- which meant
the one guard we have could not see it drift, and it had: a test count two
releases stale (144, now 156) and five references to connector/, renamed the
day before. That list is now README.md alone, with a comment explaining why to
keep it that way.

The file now lives at docs/gsp_demo_java-maintenance-notes.md in the private
sqlparser/gsp_java repository, with a header recording where it came from, the
two facts corrected on the way over, and the fact that its change log stops at
2026-08-12. Its full history stays in this repository's git log.

Nothing is lost from the README: the two deep links that carried substance are
inlined where they were referenced -- the four package roots and 263
contradicting files, and all four pom_dlineage.xml breakages -- so both
warnings now stand on their own.

Also fixes two relative links that had been broken for some time and that
nothing was checking: columnImpact pointed at ../dlineage from a directory
where that resolves to antiSQLInjection/dlineage, and search pointed at
./visitors from inside search/. Every relative link in all 64 tracked markdown
files resolves now.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UK3dBXFXqiDRYJxz1YxWCr
@sqlparser
sqlparser merged commit 6d49258 into master Aug 25, 2026
3 checks passed
@sqlparser
sqlparser deleted the move-out-maintenance-notes branch August 25, 2026 04:17
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant