Fix filldown RuntimeError on header-only tables

[sarathfrancis90]

filldown reads the first data row up front to seed its fill values, so a table that has a header but no data rows blows up with RuntimeError: generator raised StopIteration instead of just passing the header through. fillright and fillleft don't have this problem — they return the header unchanged — so it's really an inconsistency between the three fill transforms.

It's easy to hit, e.g. after a select that filters everything out:

>>> import petl as etl
>>> empty = etl.select([['foo', 'bar'], [1, 2]], lambda r: r.foo > 100)
>>> list(etl.filldown(empty))
RuntimeError: generator raised StopIteration

Same PEP 479 class as the iterrowslice fix in #613: the StopIteration from next() escapes the generator. I wrapped the seed read in the same try/except StopIteration: return the function already uses for the header read just above it, so a header-only table now yields just the header like the other fill transforms.

Added a regression test next to the existing *_headerless ones and a changelog note. Verified with pytest --cov=petl petl and pytest --doctest-modules petl (all green) plus the sphinx-build -W docs build.

[qodo-free-for-open-source-projects]

PR Summary by Qodo

Fix filldown StopIteration crash on header-only tables
🐞 Bug fix 🧪 Tests 📝 Documentation 🕐 10-20 Minutes

Description

• Guard filldown’s initial seed-row read so header-only tables don’t raise RuntimeError.
• Add regression test to ensure header-only input passes through unchanged.
• Note the behavioral fix in the 1.7.17 changelog.

Diagram

graph TD
  A["Client code"] --> B(["filldown() API"]) --> C["FillDownView"] --> D(["iterfilldown()"]) --> E[("Input table iterator")] --> F[("Output rows")]
  T["test_fills.py"] --> B

Loading

High-Level Assessment

The following are alternative approaches to this PR:

1. Lazy seed initialization inside the main loop

• ➕ Avoids a separate next(it) call and its StopIteration handling
• ➕ Keeps all row processing in one loop, potentially simpler to reason about
• ➖ Requires additional branching to detect and handle the first data row
• ➖ Slightly larger refactor for a small behavioral fix

2. Use itertools.tee (peek) to avoid consuming the iterator

• ➕ Can “peek” at the first data row without advancing the original iterator
• ➖ Extra memory/overhead and complexity for minimal benefit
• ➖ Still needs explicit handling for the no-data case

Recommendation: The current approach (catching StopIteration around the seed-row next(it)) is the best fit: it is minimal, consistent with the existing header read guard, aligns filldown behavior with fillright/fillleft on header-only input, and avoids a larger refactor.

Files changed (3) +22 / -1

Bug fix (1) +4 / -1

fills.pyHandle missing seed row in iterfilldown to avoid PEP 479 RuntimeError +4/-1

Handle missing seed row in iterfilldown to avoid PEP 479 RuntimeError

• Wraps the initial seed-row read (next(it)) in a try/except StopIteration. For header-only tables, the generator now exits cleanly after yielding the header instead of leaking StopIteration and triggering a RuntimeError under PEP 479.

petl/transform/fills.py

Tests (1) +13 / -0

test_fills.pyAdd regression test for filldown on header-only tables +13/-0

Add regression test for filldown on header-only tables

• Introduces test_filldown_header_only to ensure filldown returns the header unchanged when there are no data rows. Verifies behavior both with default fill fields and with an explicit field selection.

petl/test/transform/test_fills.py

Documentation (1) +5 / -0

changes.rstDocument filldown header-only behavior fix in changelog +5/-0

Document filldown header-only behavior fix in changelog

• Adds a 1.7.17 changelog entry noting that filldown no longer raises RuntimeError for header-only tables. Clarifies behavior consistency with fillright/fillleft.

docs/changes.rst

[qodo-free-for-open-source-projects]

Code Review by Qodo

🐞 Bugs (0) 📘 Rule violations (0) 📎 Requirement gaps (0)

Great, no issues found!

Qodo reviewed your code and found no material issues that require review

[sarathfrancis90]

Heads up that the red on the Python 3.8/3.9 jobs is AttributeError: ignore_egg_info_in_manifest during pip install -e . (egg-info generation) — a recent setuptools release triggers it on those legacy interpreters via the vcs-versioning backend. It happens before the tests run and hits every PR's 3.8/3.9 matrix entries, not this change; the fix is green on 3.10–3.14 (and locally: 572 tests + 725 doctests). Glad to send a separate PR pinning setuptools for the 3.8/3.9 jobs if that'd help.