Benchmark

This page describes how to run performance benchmarks comparing Incremark with other Markdown streaming parsers.

Parsers Tested

Parser	Approach	Complexity	Description
Incremark	Incremental parsing	O(n)	Only parses new content
Streamdown	Full re-parsing	O(n²)	Re-parses entire content each time
markstream-vue	Full re-parsing	O(n²)	Re-parses entire content each time
ant-design-x	Full re-parsing	O(n²)	Based on Marked

Test Methodology

1. Read all .md files from test-data/ directory
2. Split content into small chunks (5 chars each, simulating AI token-by-token output)
3. For each parser:
   • Incremark: Uses append() method - only parses new content
   • Others: Accumulates string and re-parses entire content each time
4. Measure total time, per-chunk average, and memory usage

Run Benchmark

Clone Repository

git clone https://github.com/kingshuaishuai/incremark.git
cd incremark/benchmark-compare

Install Dependencies

pnpm install

Add Test Data

Place your Markdown files in the test-data/ directory:

# Example: copy some markdown files
cp /path/to/your/*.md test-data/

Run

# Standard run
pnpm benchmark

# With garbage collection (more accurate memory stats)
pnpm benchmark:gc

Output

The benchmark generates:

• Console output: Summary table with timing comparisons
• benchmark-results.json: Detailed results in JSON format

Sample Output

📊 Performance Summary (sorted by lines)

| filename          | lines | chars   | Incremark | Streamdown | vs Streamdown |
|-------------------|-------|---------|-----------|------------|---------------|
| short-doc.md      |    91 |   2,341 |      12.0 |       50.5 |          4.2x |
| medium-doc.md     |   147 |   4,123 |       9.0 |       58.8 |          6.6x |
| long-doc.md       |   391 |  12,456 |      19.1 |      208.4 |         10.9x |
| extra-long-doc.md |   916 |  28,912 |      87.7 |     1441.1 |         16.4x |

📈 Performance Summary:
   - Incremark total: 127.8 ms
   - Streamdown total: 1758.8 ms (13.8x slower)

Expected Results

Document Size	Lines	Incremark Advantage
Short	~100	4-6x faster
Medium	~400	10x+ faster
Long	~900	16-65x faster

Key Insight

The longer the document, the greater the advantage of incremental parsing.

This is because:

• Incremark: O(n) - only processes new content
• Others: O(n²) - re-processes entire accumulated content on each chunk

Benchmark Code

The benchmark script is located at benchmark-compare/benchmark.ts. Key implementation:

// Incremark: O(n) incremental parsing
const parser = createIncremarkParser()
for (const chunk of chunks) {
  parser.append(chunk)  // Only parses new content
}
parser.finalize()

// Others: O(n²) full re-parsing
let accumulated = ''
for (const chunk of chunks) {
  accumulated += chunk
  parseMarkdown(accumulated)  // Re-parses entire content
}

Contributing Test Data

We welcome contributions of real-world Markdown files for benchmarking. Please ensure:

1. Files are reasonably sized (100-1000+ lines)
2. Content is appropriate for public sharing
3. Files represent typical AI output scenarios

Submit via pull request to the benchmark-compare/test-data/ directory.