Add rocksdb_slicetransform_create_capped_prefix() to the C API

[dfa1]

Adds rocksdb_slicetransform_create_capped_prefix() to the C API, mirroring the existing fixed-prefix function. Verified via a new c_test phase.

[github-actions]

✅ clang-tidy: No findings on changed lines

Completed in 66.3s.

[github-actions]

✅ Claude Code Review

Auto-triggered after CI passed — reviewing commit c9c7e7c

---

Summary

Clean, well-structured PR that adds rocksdb_slicetransform_create_capped_prefix() to the C API. The implementation exactly follows the established pattern used by rocksdb_slicetransform_create_fixed_prefix() and rocksdb_slicetransform_create_noop(). The test is thorough and correctly exercises the key behavioral difference (short keys being InDomain). All four required locations are updated consistently (c.h, c.cc, c_base.h, c_base.cc).

High-severity findings (0):

No high-severity findings.

Full review (click to expand)

Findings

🔴 HIGH

None.

🟡 MEDIUM

None.

🟢 LOW / NIT

L1. Test iterates entirely in-memtable without flushing — db/c_test.c

• Issue: The test only exercises the in-memtable path (no rocksdb_flush call). The existing "prefix" phase test also doesn't flush, so this is consistent with the existing pattern, but a flush + SST-read path would provide additional coverage for the capped prefix transform interacting with block-based table prefix bloom filters.
• Suggested fix: Optional enhancement for future work; not blocking since the underlying CappedPrefixTransform is extensively tested with SST files in C++ tests (db_bloom_filter_test.cc, db_etc2_test.cc, slice_transform_test.cc, etc.).

L2. Test comment is verbose — db/c_test.c:3503-3513

• Issue: The 8-line block comment explaining the test rationale is more detailed than typical c_test.c comments. While helpful for understanding intent, it could be condensed.
• Suggested fix: Consider trimming to 2-3 lines, e.g.: // Capped-prefix transform: all keys are InDomain, including those shorter than the cap. Exercise with a key shorter than cap ("ba", 2 bytes, cap 3).

Cross-Component Analysis

Context	Applicable?	Assessment
WritePreparedTxnDB	No	C API wrapper only; underlying transform is context-agnostic
ReadOnly DB	No	Factory function; no DB state dependency
User-defined timestamps	No	Prefix transform is orthogonal
All other contexts	No	Thin wrapper delegates entirely to existing, well-tested C++ code

Assumption stress-test:

• Claim: "follows exact same pattern as fixed_prefix" — Verified. The 6 lines of the new function body are structurally identical to rocksdb_slicetransform_create_fixed_prefix, differing only in the NewCappedPrefixTransform call. Memory ownership via SliceTransformWrapper destructor is correct.
• Claim: "cap_len=0 is safe" — Verified. CappedPrefixTransform(0) returns empty prefix for all keys, InDomain returns true. Existing unit test at util/slice_transform_test.cc:41-45 confirms this.

Positive Observations

• Correct pattern adherence: The implementation exactly mirrors the two existing factory functions — no deviation, no missed fields.
• Thoughtful test design: The test correctly uses fresh options (not the shared options variable modified by the "prefix" phase) to avoid hash-based memtable partitioning, and specifically tests a key shorter than the cap length to exercise the key behavioral difference from fixed-prefix.
• Complete coverage of code locations: All four required files (c.h, c.cc, c_base.h, c_base.cc) plus the release note are updated.
• Appropriate release note: Short, one-line, references both the new and existing function — follows CLAUDE.md guidance.

---

ℹ️ About this response

Generated by Claude Code.
Review methodology: claude_md/ci_review_prompt.md

Limitations:

• Claude may miss context from files not in the diff
• Large PRs may be truncated
• Always apply human judgment to AI suggestions

Commands:

• /claude-review [context] — Request a code review
• /claude-query <question> — Ask about the PR or codebase

[dfa1]

@pdillinger hello! can you please have a look? thanks!