Fix wildcard columns not matching when combined with optional (?)

[mborodii-prog]

Fix wildcard columns not matching when combined with optional (?)

Problem

write:
  - file:
      name: file.xlsx
      columns:
        - Col*?

Any column selector combining a wildcard (*) with the optional marker (?) matched nothing, silently — the columns were just missing from the output, with no error.

Root cause

wildcard_expansion() converts * patterns into escaped regex before checking for the trailing ? optional marker. So Col*? got turned into the regex Col(.*)\?, which requires the column name to literally end in a question mark character — never true for a real column — instead of being recognized as "wildcard Col*, marked optional."

>>> from wrangles.utils import wildcard_expansion
>>> wildcard_expansion(['Col1', 'Col2', 'Other'], ['Col*'])
['Col1', 'Col2']          # works fine on its own
>>> wildcard_expansion(['Col1', 'Col2', 'Other'], ['Col*?'])
KeyError: 'Wildcard expansion pattern did not find any matching columns'

Fix

1. Strip the ? marker before the *→regex conversion, and track optionality per-entry so a wildcard that matches zero columns is treated like a missing optional column (no error) instead of guaranteed to match nothing.

>>> wildcard_expansion(['Col1', 'Col2', 'Other'], ['Col*?'])
['Col1', 'Col2']                          # now expands correctly
>>> wildcard_expansion(['Col1', 'Col2'], ['NoMatch*?'])
[]                                        # optional + zero matches = no error
>>> wildcard_expansion(['Col1', 'Col2'], ['NoMatch*'])
KeyError: 'Wildcard expansion pattern did not find any matching columns'   # non-optional still errors, unchanged

2. Second bug found while testing the first fix: wildcard_expansion() gets called twice on the same columns list within a single write step (once generically in recipe.py, again inside the file connector). The original code's optional-handling used a local variable and never mutated the caller's list, so this double-call was always safe. My first attempt at fix Allow setting credentials with environment variables #1 mutated the list in place to strip ?, which broke the second call — it saw the marker already gone and raised anyway.

>>> cols = ['Find', 'Replace', 'find?']
>>> wildcard_expansion(['Find', 'Replace'], cols)
['Find', 'Replace']
>>> cols   # before this fix: mutated to ['Find', 'Replace', 'find'] - lost the "optional" marker
['Find', 'Replace', 'find?']   # now: untouched, safe to call again
>>> wildcard_expansion(['Find', 'Replace'], cols)   # second call, same list
['Find', 'Replace']   # still correct

Fixed by having the function operate on an internal copy — it no longer mutates the caller's list under any circumstance, regardless of how many times it's called with the same input.

Tests

Added to tests/connectors/test_file.py:

• test_write_file_optional_wildcard_col_matches — Fi*? against a 2-column file correctly writes only the matching column.
• test_write_file_optional_wildcard_col_no_matches — NoSuchColumn*? alongside a real column writes only the real column, no error.

Both read back the actual written file rather than trusting recipe.run()'s return value, since write-step column filtering is a side effect on the output file, not on the pipeline dataframe.