-p legacy is the only thing that uses mediawiki-utilities, and it exists
solely to reproduce numbers from research predating the mwxml and
mwtypes split. The package is a 2015 monolith that also contains
mw.database and mw.api, so it declares requests and pymysql
unconditionally: every wikiq install pulled a MySQL client and six other
packages to support one flag, building them from an sdist that has no
wheel.
Move it to a `legacy` extra. mw.lib.persistence imports nothing from
mw.database or mw.api, so nothing about -p legacy changes for people who
install the extra, and everyone else stops paying for it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
(cherry picked from commit 66dcc6d8cea620a619fc19121b5c27a085b11991)
pywikidiff2 was imported at module scope, so every invocation of wikiq
required it -- including `wikiq --help`. It is not on PyPI, compiles a
C++ extension, and needs libthai, which made a compiler a hard
requirement for installing a tool that mostly does not need one.
Only --diff and -p wikidiff2 actually use it. Import it inside those two
code paths instead, and report what to install when it is missing rather
than failing with an ImportError traceback. The check also runs once at
startup, since both use sites sit deep in the per-revision loop and a run
can stream for hours before reaching them.
Dropping the dependency also removes the PEP 508 direct reference from
the package metadata, and with it the need for hatchling's
allow-direct-references. PyPI rejects uploads whose metadata contains a
direct URL, so this is a prerequisite for publishing wikiq there.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
(cherry picked from commit 12325afd46670e949abf1c0e14799e212d7ff6ce)
Nate has given up on getting wikiq to output parquet directly because it
makes too many unpredictable large memory allocations. The workflow now
outputs JSONL in a single pass and then uses spark (wikiq_spark) to index
it as parquet in a second pass.
Remove the parquet output path along with the machinery that existed only
to support it: checkpoint files, resume temp-file merging, namespace
partitioning, and file rotation (--partition-namespaces,
--max-revisions-per-file). Resume support remains for JSONL output, where
the resume point is derived from the last complete line of the output
file. Also remove the parquet tests and baseline files.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
(cherry picked from commit 72a8851373)
build_table returns (table, reverts_column, wikitext_parser) but the
jsonl test helper still unpacked two values, so test_jsonl_noargs and
test_jsonl_tsv_equivalence failed with a ValueError before reading any
output. Also update the docstring, which still described the two-value
return.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
(cherry picked from commit 7c27942cb0)
WikiqPage unconditionally prepended the namespace name to titles of
pages outside the main namespace. mwxml >= 0.3.7 keeps the namespace
prefix on the title when the dump carries <ns> tags (it previously
stripped it), which made wikiq emit double-prefixed titles like
"Category:Category:Alaska". Check for the prefix before adding it.
mwxml still strips the prefix on its fallback path for dumps without
<ns> tags, so those titles are still prefixed as before.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
(cherry picked from commit ec09641545)
RegexPair.matchmake crashed with a TypeError when a capture-group
pattern was applied to a revision whose text or comment was deleted or
suppressed (content is None). Guard both matching paths against None,
and make the no-capture-group path emit a None column for such
revisions instead of omitting the key entirely.
This carries forward the fix Kaylea Champion and Mako Hill made on the
mako_changes-20230429 branch (7e6cd5b), which predated the rewrite.
Adds a unit test for matchmake(None) and an end-to-end test against the
ikwiki dump, which contains revisions with deleted text and comments.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
(cherry picked from commit dbcae5c64e)
Per-page adaptive parse skip: once mwparserfromhell times out on a
revision of size N bytes, force-timeout future revisions on the same
page whose text is >= N bytes instead of waiting another 60s on each
one. State resets per page, so well-behaved pages are unaffected.
Motivated by zhwiki bot-vandalism log pages (e.g. ns=4
'Wikipedia:当前的破坏/...' at 4.6GB across 6133 revisions) where every
revision triggers the 60s SIGALRM, blocking the worker for hours.
- Implement per-namespace resume points (dict mapping namespace -> (pageid, revid))
to correctly handle interleaved dump ordering in partitioned output
- Extract resume functionality to dedicated resume.py module
- Add graceful shutdown handling via shutdown_requested flag (CLI-level only)
- Use lazy ParquetWriter creation to avoid empty files on early exit
- Refactor writing logic to _write_batch() helper method
- Simplify control flow by replacing continue statements with should_write flag
