Delete the dead Ant builds and scansql, fix the inner docs that contradict the README - #57
Open
sqlparser wants to merge 2 commits into
Open
Delete the dead Ant builds and scansql, fix the inner docs that contradict the README#57sqlparser wants to merge 2 commits into
sqlparser wants to merge 2 commits into
Conversation
…adict the README
Finding P4 of the 2026-08-24 first-visit evaluation: documents a visitor reaches
after "clone and look around" rather than by following the first code block, all
of them describing a repository that no longer exists.
Deleted, because they are not fixable into anything worth having:
- build.xml. Its DEFAULT target compiled src/main/java/demos against lib/ for
the main class demos.listGSPInfo -- three references to things deleted in
July, in four lines. Typing `ant` in the repository root ran that.
- gettablecolumns/build.xml. Same lib/ classpath, and pinned to parser
2.0.2.4 against 4.2.6 sources.
- demos/scansql/. A 141-byte readme, no Java source anywhere in the tree, and
a download URL whose host does not resolve. A demo directory with no demo,
advertising a binary nobody can get.
No markdown file mentioned any of the three; nothing ran them. Maven and the
.bat scripts are the two supported build routes and both are under CI, which is
why a third that nothing exercises is worth removing rather than repairing --
the same reasoning that deleted buildJar.sh, buildJar.bat and MANIFEST.MF on
2026-07-28.
Fixed:
- formatsql/readme.md said `java formatsql sqlfile.sql`. Now the real Maven
invocation, plus what /tolerant does and which checked-in sample
demonstrates it. Both commands were run before being written down.
- gettablecolumns/readme.md pointed at an article on support.sqlparser.com
(NXDOMAIN) and a zip at www.gudusoft.com (HTTP 404), and gave three
`java -cp .;lib/*;external_lib/*` commands using JDBC flags removed from
this demo on 2026-07-28, against a lib/ directory deleted the same day. The
binary section now explains what happened, and the flags are documented as
gone rather than as usage. Also dropped a stale
-Dexec.classpathScope=compile the root README already says is unnecessary.
check-stale-docs.sh grows to cover this class:
- four more dead strings: `.;lib/*`, build.xml, and the two hosts that do not
resolve.
- every relative markdown link must resolve, across all files including the
ALLOW-listed README.md. A file may have reason to name something removed;
none has reason to link at a path that is not there.
The link check earned itself immediately: it rejected this very commit, because
a licensed-only/README.md link written six directories up needed seven. That is
the point -- nothing else in the repository would have noticed.
Self-test now covers 12 cases, including that a resolving link passes, a broken
one fails, and an http link is not treated as a file path.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UK3dBXFXqiDRYJxz1YxWCr
The root README has said since the system-scope dependencies were removed that the flag is unnecessary. Twenty demo readmes went on telling people to pass it anyway, which is the same contradiction as the rest of this branch, just a benign one: the flag is inert, not wrong. Removed from every demo readme, including the two forms that needed care -- one where the flag rode on the end of an args line, and one where it was the last line of a continued command and dropping it left a dangling backslash on the line above. Nothing ends a fenced block with a stray continuation now, and three of the rewritten commands were run verbatim to confirm plain exec:java does the job. sqlrefactor's was =runtime rather than =compile, which is worth noting only because it shows nobody was reading these. check-stale-docs.sh now treats -Dexec.classpathScope as a dead string, so it cannot come back. README.md is the one file allowed to name it, since its job is to tell people they do not need it. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UK3dBXFXqiDRYJxz1YxWCr
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
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.
Finding P4 of the 2026-08-24 first-visit evaluation: the documents a visitor reaches after clone and look around, rather than by following the first code block. Two of the seven rows were already fixed (
demos/readme.mdin #54, the README version claim in #52); these are the rest.Deleted — not fixable into anything worth having
build.xmlsrc/main/java/demosagainstlib/for main classdemos.listGSPInfo— three references to things deleted in July, in four lines. Typingantin the repo root ran that.gettablecolumns/build.xmllib/classpath, and<property name="gsp.version" value="2.0.2.4"/>against 4.2.6 sourcesdemos/scansql/No markdown file mentioned any of the three, and nothing ran them. Maven and the
.batscripts are the two supported routes and both are under CI; a third that nothing exercises is worth removing rather than repairing — the same call that deletedbuildJar.sh,buildJar.batandMANIFEST.MFon 2026-07-28.Fixed
formatsql/readme.mdsaidjava formatsql sqlfile.sql. Now the real Maven invocation, plus what/tolerantdoes and which checked-in sample demonstrates it. Both commands were run before being written down — the plain form on the mixed sample stops at the bad statement, so the example usesyour.sqland the/tolerantexample uses the sample.gettablecolumns/readme.md— I resolved all three of its links:http://support.sqlparser.com/tutorials/gsp-demo-get-table-column/https://www.gudusoft.com/gsp_java/gettablecolumn.ziphttp://ftp.gudusoft.com/dl/scansql/scanSQL_trial.zip(inscansql/)It also gave three
java -cp .;lib/*;external_lib/*commands using/h /P /u /pflags removed from the demo on 2026-07-28, against alib/deleted the same day. The binary section now explains what happened and points atlicensed-only/; the flags are documented as gone rather than as usage.check-stale-docs.shgrows to cover the class.;lib/*,build.xml, and the two non-resolving hostsREADME.md— a file may have reason to name something removed, none has reason to link at a path that is not thereThe link check earned itself immediately: it rejected this very commit. A
licensed-only/README.mdlink I wrote six directories up needed seven:Nothing else in the repository would have noticed. Self-test is now 12 cases, including that a resolving link passes, a broken one fails, and an
http://link is not treated as a file path.The
-Dexec.classpathScopesweep (second commit)Done as well — all 20 remaining demo readmes. The root README has said since the
system-scope dependencies went that the flag is unnecessary; those readmes went on telling people to pass it. Inert rather than wrong, but the same contradiction as everything else here.Two forms needed care: the flag riding on the end of an
-Dexec.argsline, and the flag as the last line of a continued command, where dropping it leaves a dangling\on the line above. Nothing ends a fenced block with a stray continuation now, and three of the rewritten commands were run verbatim.sqlrefactor's was=runtime, not=compile— worth mentioning only as evidence nobody was reading these.check-stale-docs.shnow treats-Dexec.classpathScopeas a dead string, so it cannot come back;README.mdis the one file allowed to name it, since its job is to say you don't need it. Self-test is 13 cases.Verified
156 tests · 79 demos start · 24/24 demo cases · dlineage jar JSON+XML and
hr_miniunder the trial cap · stale-docs self-test 12/12, clean scan over 62 files, 63 files link-checked · licensed-only guard 3/3 · parser version consistent at 4.2.6.🤖 Generated with Claude Code
https://claude.ai/code/session_01UK3dBXFXqiDRYJxz1YxWCr