[regression] Error with adaptive predicate pushdown: "Invalid offset in sparse column chunk data: 754, no matching page found."

[erratic-pattern]

Describe the bug

Error:

Parquet error: Invalid offset in sparse column chunk data: 754, no matching page found.
If you are using a `SelectionStrategyPolicy::Mask`, ensure that the OffsetIndex is provided when creating the InMemoryRowGroup.

Occurs when:

1. A predicate uses RowSelectionStrategy::Selectors with a RowSelector list that skips an entire page.
2. Another predicate uses RowSelectionStrategy::Mask by triggering the mask run-length threshold
3. The column with RowSelectionStrategy::Mask is not in the output projection, so should_force_selectors does not force it to use RowSelectionStrategy::Selectors
4. The mask strategy attempts to fetch pages that were skipped, resulting in an error

To Reproduce

A minimal reproducer is available at: https://github.com/erratic-pattern/parquet_mask_strategy_missing_pages

git clone https://github.com/erratic-pattern/parquet_mask_strategy_missing_pages
cd parquet_mask_strategy_missing_pages
cargo test

The test uses a parquet file with:

• 2 row groups, 300 rows each
• Tag column with values 'a', 'b', 'c' sorted (100 rows each)
• Time column with alternating in-range/out-of-range values
• Page size set so tag='b' section contains at least one full page

The test simulates a query like SELECT tag WHERE tag IN ('a', 'c') AND time >= X AND time < Y with three predicates:

1. tag IN ('a', 'c') - creates initial selection [select 100, skip 100, select 100]
2. time >= X - creates sparse selection, pages fetched as Sparse
3. time < Y - triggers Mask strategy due to sparse selection from predicate 2

Additional context

• Introduced in parquet 57.1.0 via [Parquet] Adaptive Parquet Predicate Pushdown #8733
• Related to [Parquet] Support skipping pages with mask based evaluation #8845

[alamb]

For some more context, we encountered this error in our production systems in InfluxData after rolling out 57.2.0

[alamb]

@erratic-pattern could you please make a PR to this repository to add the reproducer from https://github.com/erratic-pattern/parquet_mask_strategy_missing_pages as a test?

That way we can more easily verify if the fix from @sdf-jkl in #9118 will also fix this bug?

[erratic-pattern]

Here is a regression test #9243

[sdf-jkl]

I ran the test locally and it fails. We'll need more work on #9118

Invalid offset in sparse column chunk data: 934, no matching page found.If you are using a `SelectionStrategyPolicy::Mask`, ensure that the OffsetIndex is provided when creating the InMemoryRowGroup."))

[alamb]

Thanks @erratic-pattern and @sdf-jkl -- I plan to work on this issue today

[alamb]

I looked at the test in #9243 (review) and it is looking good. I added a little more coverage and then plan to merge it in and turn my focus to #9118