Add Rust ecosystem comparison benchmark suite

[Nebu1eto]

Summary

Adds a comprehensive benchmark suite that compares SheetKit against other popular Rust Excel libraries (calamine, rust_xlsxwriter, and edit-xlsx) across multiple scenarios including read, write, and specialized operations.

Key Changes

• New benchmark binary (benchmarks/rust-comparison/src/main.rs): A 1800+ line benchmark harness that:

  • Compares SheetKit against calamine (read-only), rust_xlsxwriter (write-only), and edit-xlsx (read/modify/write)
  • Mirrors the scenario structure of existing Node.js and Rust benchmarks for consistency
  • Implements comprehensive benchmarking infrastructure with warmup runs, multiple iterations, and statistical analysis (min, max, median, p95)

• Benchmark scenarios:

  • Read benchmarks: File reading performance across different file sizes
  • Write benchmarks: Large data writes (50k rows × 20 cols), styled cells, multi-sheet workbooks, formulas, text-heavy data, and merged cells
  • Scaling benchmarks: Write performance at various row counts (100, 1k, 10k, 100k, 500k rows)
  • Advanced features: Buffer round-trip (write→read), streaming writes, and random-access reads

• Performance metrics: Captures execution time (with min/max/median/p95 statistics), peak memory usage, and output file size

• System information detection: Automatically detects and reports CPU model, RAM, and Rust compiler version for reproducible benchmarking

• Platform support: Includes platform-specific code for macOS and Linux to measure RSS memory usage; graceful fallback for other platforms

• Workspace integration: Added new benchmark crate to workspace members in root Cargo.toml

Implementation Details

• Uses closure-based benchmark harness to allow flexible test setup and cleanup
• Implements proper file cleanup between runs to avoid disk space issues
• Supports both file-based and buffer-based operations where applicable
• Deterministic random cell selection for random-access benchmarks using seeded PRNG
• Comprehensive error handling with informative skip messages when fixture files are missing

https://claude.ai/code/session_012jgPjB6pTXjQpWarssU9yp

[claude]

Code Review

Issue: Decorative Section Marker Comments

File: benchmarks/rust-comparison/src/main.rs

The file contains multiple decorative section marker comments using repeated dashes (// -----------) and equals signs (// ===). According to CLAUDE.md line 58, the project prohibits section markers and decorative comments:

No section markers, decorative comments, or emoji in code or output.

Examples of violations:

1. Lines 20-22: Decorative dashes around "Helpers" section
2. Lines 159-161: Decorative dashes around "BenchResult" section
3. Lines 1143-1145: Decorative equals signs around "READ BENCHMARKS" section
4. Similar patterns throughout the file (~20+ occurrences)

Suggested fix: Remove all decorative comment lines. Simple section comments without decoration are acceptable (e.g., // Helpers), but the decorative lines of dashes and equals signs should be removed.

---

No other issues found. Checked for bugs and CLAUDE.md compliance.

[claude]

Code review

Issue Found: Section marker comments violate CLAUDE.md

The new benchmarks/rust-comparison/src/main.rs file contains decorative section marker comments that violate the code style guidelines.

CLAUDE.md Rule (Line 58):

No section markers, decorative comments, or emoji in code or output.

Examples found in the code:

Pattern 1 (dashed lines):

// ---------------------------------------------------------------------------
// Helpers
// ---------------------------------------------------------------------------

Pattern 2 (equals lines):

// ===================================================================
// READ BENCHMARKS: SheetKit vs calamine
// ===================================================================

Locations: These decorative section markers appear throughout benchmarks/rust-comparison/src/main.rs at lines:

• 20-22 (Helpers)
• 159-161 (BenchResult)
• 204-206 (Bench harness)
• 283-285, 378-380 (READ BENCHMARKS)
• 1034-1036, 1163-1165 (WRITE BENCHMARKS)
• 1224-1226 (WRITE SCALING)
• 1298-1300 (READ SCALING)
• 1375-1377, 1451-1453, 1722-1724 (Advanced benchmarks)
• And several more throughout the file (24 total lines)

Recommended fix: Remove all decorative section marker comment lines. The existing benchmarks/rust/src/main.rs demonstrates organizing benchmark code without section markers - function names and spacing provide sufficient structure.