refactor(parquet): introduce string_size_scan_state for string size kernels

[vyasr]

Description

This PR builds on #23479 to create a minimal shared state for the Parquet string-size scan kernels. It extracts page_decode_output_state from page_state_s, grouping the output conversion fields (dtype_len, dtype_len_in, and ts_scale) that are used while computing string output sizes.

The primary addition is string_size_scan_state, composed from page_decode_setup_state, page_decode_stream_state, and page_decode_output_state. The three string-size scan kernels (compute_page_string_sizes_kernel, compute_delta_page_string_sizes_kernel, and compute_delta_length_page_string_sizes_kernel) now use this smaller composed state instead of the full page_state_s. That removes unrelated nesting and progress fields from these kernels' shared-memory state while preserving the output-conversion metadata they need.

A fresh cuobjdump -res-usage comparison against upstream/main shows shared-memory reductions in all three migrated kernels. Register usage does not increase anywhere; it drops on a few newer architectures:

Kernel	Arch summary	shmem delta	register changes
compute_page_string_sizes_kernel	all generated arch variants	-536 bytes	unchanged except sm_100f: 32 -> 31
compute_delta_page_string_sizes_kernel	all generated arch variants	-536 bytes	unchanged except sm_90a: 40 -> 32, sm_120a: 48 -> 40, sm_120: 48 -> 40
compute_delta_length_page_string_sizes_kernel	all generated arch variants	-544 bytes	unchanged

Representative sm80 measurements:

Kernel	shmem before	shmem after	delta	regs before	regs after
compute_page_string_sizes_kernel	1160	624	-536	40	40
compute_delta_page_string_sizes_kernel	3528	2992	-536	32	32
compute_delta_length_page_string_sizes_kernel	2144	1600	-544	62	62

This is the same incremental pattern as #23479: extract one coherent sub-state, migrate the kernels that can use the narrower composition, and keep the remaining full decode state for the follow-up integration PR.

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: c229b34d-cfbd-4f9f-9518-8997334b9915

📥 Commits

Reviewing files that changed from the base of the PR and between 2baed40 and 17b1a06.

📒 Files selected for processing (6)

• cpp/src/io/parquet/decode_fixed.cu
• cpp/src/io/parquet/page_data.cu
• cpp/src/io/parquet/page_decode.cuh
• cpp/src/io/parquet/page_delta_decode.cu
• cpp/src/io/parquet/page_state_composed.cuh
• cpp/src/io/parquet/page_string_decode.cu

🚧 Files skipped from review as they are similar to previous changes (6)

• cpp/src/io/parquet/page_state_composed.cuh
• cpp/src/io/parquet/page_delta_decode.cu
• cpp/src/io/parquet/page_string_decode.cu
• cpp/src/io/parquet/decode_fixed.cu
• cpp/src/io/parquet/page_data.cu
• cpp/src/io/parquet/page_decode.cuh

---

📝 Walkthrough

Summary by CodeRabbit

• Bug Fixes
  • Improved Parquet decoding reliability for fixed-width, byte-stream-split, delta, and string data.
  • Corrected conversions for decimals, timestamps, fixed-length strings, and other output formats.
  • Fixed null-value handling, including zero filling and destination sizing.
  • Strengthened validation of malformed byte-stream data.
  • Improved consistency when decoding nested data and converted outputs.

Walkthrough

Parquet page decoding now stores conversion metadata in output_cvt. Fixed-width, general, delta, timestamp, decimal, byte-stream-split, and string-size decoding paths use converted widths and timestamp scaling.

Changes

Parquet output conversion state

Layer / File(s)	Summary
Conversion state and initialization cpp/src/io/parquet/page_decode.cuh	Adds page_decode_output_state and moves output widths and timestamp scaling into page_state_s::output_cvt.
Fixed-width and page decoding cpp/src/io/parquet/decode_fixed.cu, cpp/src/io/parquet/page_data.cu, cpp/src/io/parquet/page_data.cuh	Uses output_cvt for destination sizing, input widths, decimal and timestamp conversion, byte-stream-split validation, and null filling.
Delta binary output sizing cpp/src/io/parquet/page_delta_decode.cu	Uses the converted output width for value writes and null filling.
String-size scan state and widths cpp/src/io/parquet/page_state_composed.cuh, cpp/src/io/parquet/page_string_decode.cu, cpp/src/io/parquet/page_string_utils.cuh	Adds shared string-scan state and uses converted input widths for fixed-length string calculations.

Estimated code review effort: 3 (Moderate) | ~25 minutes

Possibly related PRs

• rapidsai/cudf#23241: Both PRs modify Parquet page decoding and overlapping string-decoding logic.
• rapidsai/cudf#23314: Both PRs modify Parquet DELTA decoding kernels.
• rapidsai/cudf#23495: Both PRs refactor Parquet page-decoding state and composed decode states.

Suggested labels: cuIO

Suggested reviewers: pointkernel, vuule, bdice

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly identifies the introduction of string_size_scan_state for Parquet string-size kernels, which is the primary objective of the pull request.
Description check	✅ Passed	The description is directly related to the changes and explains the new composed state, migrated kernels, and shared-memory improvements.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Caution

Some comments are outside the diff and can’t be posted inline due to platform limitations.

⚠️ Outside diff range comments (1)

cpp/src/io/parquet/page_data.cu (1)

84-95: 🩺 Stability & Availability | 🔴 Critical | ⚡ Quick win

Move the malformed-page check before the division.

Line 85 computes num_values by dividing data_len by s->output_cvt.dtype_len_in before the guard at line 88 checks s->output_cvt.dtype_len_in <= 0. If dtype_len_in is 0 for a malformed BYTE_STREAM_SPLIT page, this division executes first and causes an integer division by zero, which is undefined behavior in device code. The guard below is meant to catch exactly this case, but it runs too late.

The sibling function decode_fixed_width_split_values in decode_fixed.cu performs the check before the division. Apply the same order here.

🐛 Proposed fix: check before dividing

   auto const data_len   = cuda::std::distance(s->stream.data_start, s->stream.data_end);
-  auto const num_values = data_len / s->output_cvt.dtype_len_in;
-
-  // Check malformed BYTE_STREAM_SPLIT pages
-  if (s->output_cvt.dtype_len_in <= 0 or data_len <= 0) {
+
+  // Check malformed BYTE_STREAM_SPLIT pages
+  if (s->output_cvt.dtype_len_in <= 0 or data_len <= 0) {
     cg::invoke_one(block, [&]() {
       set_error(static_cast<kernel_error::value_type>(decode_error::INVALID_BYTE_STREAM_SPLIT_SIZE),
                 error_code);
     });
     return;
   }
+
+  auto const num_values = data_len / s->output_cvt.dtype_len_in;

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@cpp/src/io/parquet/page_data.cu` around lines 84 - 95, Move the malformed
BYTE_STREAM_SPLIT validation in the surrounding decode function before
calculating num_values, checking dtype_len_in and data_len before dividing
data_len by dtype_len_in. Preserve the existing set_error and return behavior,
matching the ordering used by decode_fixed_width_split_values.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Outside diff comments:
In `@cpp/src/io/parquet/page_data.cu`:
- Around line 84-95: Move the malformed BYTE_STREAM_SPLIT validation in the
surrounding decode function before calculating num_values, checking dtype_len_in
and data_len before dividing data_len by dtype_len_in. Preserve the existing
set_error and return behavior, matching the ordering used by
decode_fixed_width_split_values.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 5843b27c-4bfd-4769-ae37-5ae10d8035ef

📥 Commits

Reviewing files that changed from the base of the PR and between 7f58752 and ad71ecf.

📒 Files selected for processing (8)

• cpp/src/io/parquet/decode_fixed.cu
• cpp/src/io/parquet/page_data.cu
• cpp/src/io/parquet/page_data.cuh
• cpp/src/io/parquet/page_decode.cuh
• cpp/src/io/parquet/page_delta_decode.cu
• cpp/src/io/parquet/page_state_composed.cuh
• cpp/src/io/parquet/page_string_decode.cu
• cpp/src/io/parquet/page_string_utils.cuh

[bdice]

Everything seems fine except this comment block, which is a little strange.

[vyasr]

/merge