fix(bigquery): handle API-wrapped null values in NULLABLE fields

[Srujan rai (Srujan-rai)]

BigQuery returns nulls as {'v': None}, not bare None. The old check value is None never matched, causing convert(None) to raise TypeError (eg. int(None)) for NULLABLE fields. Use flatten(value) before the check.

Adds regression test for the {'v': None} API response shape.

Summary

• BigQuery API returns null values wrapped as {'v': None}, not bare None
  • Old guard value is None never matched → fell through to convert(None) → TypeError (eg. int(None)) for any NULLABLE INTEGER/BOOLEAN/FLOAT/etc field
    with a null value
  • Fix: use flatten(value) is None before the null check, consistent with how the rest of parse() already uses flatten()
  • Adds regression test covering the {'v': None} API response shape for all NULLABLE types

Fixes #577

[gemini-code-assist]

Code Review

This pull request updates the parse utility to correctly handle wrapped null values from the BigQuery API by applying the flatten function to nullable fields. A unit test was added to verify this behavior. The reviewer suggested optimizing the implementation by flattening the value once at the beginning of the function to avoid redundant recursive calls and improve performance.

[gemini-code-assist]

The flatten(value) function is called multiple times throughout the parse function (see lines 113, 115, 120, and 123). To improve efficiency and avoid redundant recursive traversals, consider flattening the value once and reusing the result. Since flatten is idempotent for already-flattened values, this change is safe and improves performance, especially when processing large query results with many nullable fields.

Suggested change

	if field['mode'] == 'NULLABLE' and flatten(value) is None:
	return None
	value = flatten(value)
	if field['mode'] == 'NULLABLE' and value is None:
	return None

[Srujan rai (Srujan-rai)]

Applied value is now flattened once before all checks. All unit tests pass.