fix(qual): fix format for qualitative multiple select questions DEV-2233

[Guitlle]

Summary

Refactor QualSelectMultipleField to export ordinal columns containing text values

Description

• Replaced the legacy behavior in QualSelectMultipleField where selected qualitative multiple-select responses were aggregated into a single, comma-separated string
• Implemented get_labels() and get_value_names() to dynamically yield a fixed number of ordinal positional columns (e.g., Option 1, Option 2, ... Option N). The number of columns generated matches the maximum possible number of choices mapped in the field schema
• Modified get_value_from_entry() to return an ordered list of the specific text labels (choices) the user selected
• Implemented a new format() method that maps the textual values into different columns for each option
• Updated the test_additional_field_exports_advanced unit tests in tests/test_additional_field_exports.py

Preview steps

Run pytest, this is covered by unit tests.

[coveralls]

coverage: 86.834% (-0.02%) from 86.852% — dev-2233-export-qual-qs into main

[noliveleger]

@greptileai review

[greptile-apps]

Greptile Summary

This PR refactors QualSelectMultipleField to emit one column per choice (like FormChoiceFieldWithMultipleSelect) instead of a single comma-separated summary column, and updates the corresponding tests to match the new format.

• Adds get_labels(), get_value_names(), and format() overrides that follow the same multiple_select='both'/'summary'/'details' pattern as the non-qual multiple-select field; choice columns use internal UUIDs as keys while human-readable labels are returned by get_labels() for column headers.
• get_value_from_entry() now returns a list of UUID strings instead of a pre-formatted CSV string, and format() maps those UUIDs to per-choice '0'/'1' values while building a space-joined summary string for the summary column.

Confidence Score: 5/5

The change is safe to merge. It correctly replaces a single-column CSV representation with the standard multi-column 0/1 detail pattern, all three export modes are tested, and the row-cache key alignment between get_value_names() and format() is consistent.

All three multiple_select modes are covered by explicit test assertions. The previously raised concerns about unknown UUIDs creating spurious row-cache keys, list objects written to the summary cell, and raw UUIDs leaking into the summary are each handled correctly in the new implementation. No new edge-case gaps were found.

No files require special attention.

_{Reviews (9): Last reviewed commit: "add comment to the format function" | Re-trigger Greptile}

[noliveleger]

@greptileai review

[Guitlle]

@greptileai
Silently dropping stale UUIDs is acceptable for this case. Data from KPI can contain deleted choices so it will be very rare to find a dataset that contains stale UUIDs.

[greptile-apps]

Looking at the actual code in the diff, the implementation already handles this correctly — unrecognized UUIDs are silently dropped because res.append(label) only executes inside the if choice['uuid'] == v: branch, and the loop simply moves on when no match is found. My P1 comment was incorrect; there's no raw UUID leaking into the output.

You're right, and the behavior is both correct and intentional given that KPI data can include choices deleted after collection. The silent drop is the right call here.

Shall I remember this as a rule: "In formpack, silently dropping unrecognized/stale UUIDs from qualitative field responses is acceptable and intentional — KPI data may reference deleted choices"?

Tip: You can customize Greptile's behavior for this repo with .greptile/rules.md and .greptile/config.json.

[noliveleger]

LGTM