Repository navigation
release duckdb-sqlalchemy 1.5.6.10 - #225
Conversation
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 12a0ce5ea2
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
|
||
| ### Highlights | ||
|
|
||
| - Large INSERTs no longer store booleans as `1.0` or `0.0` in text columns that also receive floats. PRLINK |
There was a problem hiding this comment.
Replace PRLINK placeholders before publishing
When this version bump is merged to main, .github/workflows/publish.yaml extracts the 1.5.6.10 changelog section verbatim and uses it as the GitHub release description, so all seven PRLINK tokens here will become literal public release-note text instead of links; the two copies in docs/breaking-changes.md will also remain unresolved. Replace every placeholder with the actual pull-request link before publishing.
Useful? React with 👍 / 👎.
Summary
Releases 1.5.6.10 with one bulk-insert fix and two performance improvements.
executemanyINSERT that reachesduckdb_copy_thresholdand mixesfloatandboolvalues in a non-numeric column stored booleans as1.0/0.0. Arrow infers DOUBLE for such a column. These batches now use the ordinary insert path and storetrue/false.get_multi_*call resolves are kept in the Inspector'sinfo_cache.MetaData.reflect()on 20 tables takes 10 statements instead of 24. Names that do not resolve are not cached, so a table created after a failed lookup still reflects on the same Inspector.FixedArray: results are copied withlist(value)when items need no conversion, includingFloatitems that DuckDB already returns as floats.pyproject.tomlanduv.lock, withCHANGELOG.mdanddocs/breaking-changes.mdentries.Test evidence
pytest duckdb_sqlalchemy: 896 passed, 10 skipped (DuckDB 1.5.6, SQLAlchemy 2.0.52, Python from the local venv).pre-commit(ruff, ruff format, ty) passes on the changed files.FixedArrayitems with and without NULLs.