Fix where skipping wrangles when all rows are filtered ou

[mborodii-prog]

Fix: where skips wrangles when all rows are filtered out

Problem

When a where clause matched no rows, the resulting empty DataFrame was still passed into the wrangle for execution. This became an issue for chained wrangles where an earlier step creates a column that a later step depends on — because on an empty DataFrame no new columns are ever produced.

Failing example:

wrangles:
  - recipe:
      where: successful_search == True   # all False → empty DataFrame
      wrangles:
        - split.dictionary:
            input: scored_results        # runs but creates no columns (empty df)
        - split.dictionary:
            input: summary               # KeyError: column 'summary' does not exist

With input data:

scored_results	successful_search
{}	False

Error produced:

KeyError: 'ERROR IN WRANGLE #1 recipe - "ERROR IN WRANGLE #2 split.dictionary - \'Column summary does not exist\'"'

Fix

In wrangles/recipe.py, after _filter_dataframe(), added a four-line guard — if where produces an empty DataFrame, restore the original and continue to the next wrangle without executing:

df = _filter_dataframe(df, where=params.get('where'), ...)

# If where filters out all rows, skip the wrangle entirely
if len(df) == 0:
    df = df_original
    continue

Behaviour after fix

All rows filtered out → wrangle skipped, data unchanged:

scored_results	successful_search
{}	False

No error, no spurious columns added.

Partial match → still works correctly:

scored_results	successful_search	summary	other key
{'summary': 'this is a summary', 'other key': '...'}	True	this is a summary	...
{}	False	``	``

Only the matching row is processed; unmatched rows are merged back unchanged.

[ebhills]

There a column shift bug. I am running this recipe w batch size = 1:

wrangles:
  - convert.case:
      input: Desc
      output: output_column_name
      case: upper
      where: WC = "A"

[mborodii-prog]

@ebhills now it must be working fine (batch_size = 1)