petl 1.7.30 is a patch release of the Python table ETL library, tagged v1.7.30 and published on October 17, 2026. It is a bug fix release, not a feature release. The change most likely to affect a pipeline is the handling of explicit empty headers in dictionary generators.
The full release notes and downloads are on the GitHub release page. This is a stable release, not a prerelease.
Dictionary headers and boolean parsing ¶
Two fixes touch how raw input becomes table values.
PR 742 makes dictionary generators respect an explicit empty header. Before this change, an empty header passed by the caller could be treated as if no header had been given. The notes do not spell out the old fallback, so check your own sources if you build tables from lists of dicts. A pipeline that passes an empty header on purpose should now get that header back, not an inferred one.
PR 745 changes boolparser(strict=False) so it returns the original value when the input is not recognized as a boolean. The notes call this a return of the original value. In practice that means a lenient parse no longer replaces an unknown string with something else. If a downstream step counted on a coerced result, it will now see the raw value. Add a validation step after the parse if you need a guarantee.
The tradeoff is simple. Code that relied on the old lenient output may see different values. Code that wants to detect bad input gets a cleaner signal.
Sort checks and column selection ¶
PR 743 fixes issorted() on tables that have a header but no data rows. An empty table is trivially sorted, and the check should say so. This matters in pipelines that guard a merge or a join on issorted(). An empty extract on a quiet day is a normal event, and it should not fail the guard.
PR 744 lets cutout accept field sequences. You can now pass a list or tuple of field names to drop columns in one call, in the same way other field selection functions take a sequence. This is a convenience for code that builds the column list at runtime from a config file or a schema.
Aggregation keys ¶
PR 730 preserves singleton keys in multiple aggregations. This is the first contribution from its author. When you aggregate on a key and compute several results at once, a key made of a single field was not preserved as expected. After the fix, the singleton key survives in the output.
If you aggregate with several output fields, compare the key column layout before and after the upgrade. A report job that reads the output by position is the most likely place to notice a difference.
What is not in this release ¶
There are no new connectors, no new table functions, and no dependency changes in the notes. Nothing is listed as a breaking change. The release also credits two new contributors, one for PR 742 and one for PR 730. The full diff is in the compare view against v1.7.29.
Upgrade notes ¶
The notes list no breaking changes or migration steps. Still, three of the five fixes change visible output in edge cases:
- Explicit empty headers in dictionary generators are now kept.
boolparser(strict=False)returns the original value for unrecognized input.- Multiple aggregations keep singleton keys.
Run your existing test suite against v1.7.30 before rolling it out. Pay attention to tests with empty tables, mixed type boolean columns, and grouped output compared by column order. The release page lists all five pull requests if you need to trace a specific difference.
Where to get it ¶
- Release page: https://github.com/petl-developers/petl/releases/tag/v1.7.30
- Repository: https://github.com/petl-developers/petl
- Tag:
v1.7.30