[BUG] Chunk Overlap Calculation Uses Incorrect Magic Number Division

[Crimsonyx412]

Project

vgrep

Description

The chunk overlap calculation in both Indexer and ServerIndexer divides chunk_overlap by 40, resulting in minimal effective overlap regardless of the configured value. With the default chunk_overlap = 64, the calculation 64 / 40 = 1 produces only 1 line of overlap between chunks.

Error Message

N/A - Logic error, no runtime error produced.

Debug Logs

N/A - Can be observed by indexing a file and examining chunk boundaries in the database:
sqlite3 ~/.vgrep/projects/*.db "SELECT start_line, end_line FROM chunks ORDER BY start_line;"

System Information

vgrep version: 0.1.0
Default chunk_overlap config: 64

Screenshots

No response

Steps to Reproduce

1. Set chunk_overlap = 64 (default) in config
2. Create a test file with 100+ lines of code
3. Run vgrep index
4. Query the database: SELECT start_line, end_line FROM chunks ORDER BY chunk_index;
5. Observe that consecutive chunks only overlap by 1 line

Expected Behavior

Chunks should overlap by a meaningful amount based on the configured chunk_overlap value (64 characters by default), preserving context between chunk boundaries for better semantic search quality.

Actual Behavior

Chunks overlap by only 1 line regardless of the chunk_overlap configuration, due to integer division chunk_overlap / 40 = 64 / 40 = 1.

Additional Context

Files affected:

• src/core/indexer.rs → chunk_content() (lines 354-358)
• src/watcher.rs → chunk_content() (lines 418-422)

Problematic code:

let overlap_start = if line_idx > 0 {
    line_idx.saturating_sub(self.chunk_overlap / 40)  // 64 / 40 = 1
} else {
    0
};

Fix direction: Remove the arbitrary division or use a proper line-count calculation based on average line length.

[echobt]

Marking as invalid: This issue is already addressed by open Pull Requests in the target repository (CortexLM/vgrep#6 and CortexLM/vgrep#25). Please check for existing PRs before submitting new issues.