[PERFORMANCE]: Optimize stream parser buffer management (O(n²) → O(n))

[crivetimihai]

Summary

The stream parsers in mcpgateway/wrapper.py use string concatenation (+=) and slicing in tight loops, creating O(n²) time complexity. For large streaming responses with many chunks, this causes significant performance degradation.

Impact

• CPU Overhead: Each buffer += text creates a new string object, copying all existing content
• Memory Churn: Frequent allocations and deallocations increase GC pressure
• Latency: Large SSE/NDJSON streams become progressively slower as buffer grows

Operation	Current	Optimized
Append chunk	O(n)	O(1) amortized
Slice buffer	O(n)	O(1)
Total per chunk	O(n)	O(1) amortized

Affected Code

File: mcpgateway/wrapper.py

ndjson_lines() (Lines 269-303)

buffer = ""
async for chunk in resp.aiter_bytes():
    buffer += text  # Line 291: O(n) - creates new string!
    while True:
        nl_idx = buffer.find("\n")  # O(n) search
        buffer = buffer[nl_idx + 1:]  # Line 297: O(n) - creates new string!

sse_events() (Lines 306-364)

Same pattern at lines 324, 330, 348.

Proposed Fix

Use io.StringIO or bytearray for O(1) amortized append:

import io

async def ndjson_lines(resp: httpx.Response) -> AsyncIterator[str]:
    buffer = io.StringIO()
    async for chunk in resp.aiter_bytes():
        buffer.write(decoder.decode(chunk))  # O(1) amortized
        # Process complete lines...

Or use split-based approach:

async def ndjson_lines(resp: httpx.Response) -> AsyncIterator[str]:
    partial_line = ""
    async for chunk in resp.aiter_bytes():
        lines = (partial_line + decoder.decode(chunk)).split('\n')
        partial_line = lines.pop()  # Keep incomplete line
        for line in lines:
            if line.strip():
                yield line.strip()

Acceptance Criteria

• No string concatenation (+=) in buffer loops
• No string slicing for buffer truncation
• Memory usage bounded regardless of stream size
• Existing SSE/NDJSON tests pass
• Performance improvement measurable for large streams

References

• Detailed analysis: todo/newperf/optimize-stream-parser-buffers.md
• Python io.StringIO