fix: auto-detect single-column CSV files

[harsh-thakkar7]

rows_from_file() fails to auto-detect single-column CSV files. Format detection sniffs the first bytes with csv.Sniffer().sniff(), which raises csv.Error: Could not determine delimiter for any file without a delimiter — so a bare list like this could never be imported:

>>> from sqlite_utils.utils import rows_from_file
>>> import io
>>> rows_from_file(io.BytesIO(b"id\n1\n2\n"))
Traceback (most recent call last):
  ...
csv.Error: Could not determine delimiter

Fix

When sniffing fails to find a delimiter, fall back to the default CSV dialect instead of crashing. csv.DictReader (with no custom dialect) then loads the file as a single column exactly as expected:

>>> rows, format_ = rows_from_file(io.BytesIO(b"id\n1\n2\n"))
>>> list(rows)
[{'id': '1'}, {'id': '2'}]

Multi-column CSV, TSV and JSON auto-detection are unaffected.

Tests

Added to tests/test_utils.py:

• test_rows_from_file_detects_single_column_csv[id... ] — parameterized over a single-column list of words and of integers, asserting Format.CSV and the one-key rows (both fail before this change).

pytest -q tests/test_utils.py   # 21 passed
pytest -q                       # 1502 passed, 16 skipped

---

📚 Documentation preview 📚: https://sqlite-utils--881.org.readthedocs.build/en/881/

[harsh-thakkar7]

Closing this — I don't think this change is mature enough to land right now. Thanks for the project!