coveragepy Performance Optimization

Upstream performance improvements to coveragepy/coveragepy, the standard Python code coverage measurement tool by Ned Batchelder.

Background

coverage.py instruments Python execution to measure which lines and branches are exercised by tests. It's used by virtually every Python project with CI coverage gates. Performance matters because coverage overhead directly increases test suite wall time — often 2-5x slower than uncovered execution.

Profiling reveals optimization surfaces in both the trace loop hot path and the data persistence layer.

Optimization Targets

Data Collection (Phase 1 — highest leverage)

Target	File	Approach
numbits encoding/union	`numbits.py`	Pre-allocate bytearray, replace `zip_longest` with explicit loop
`add_lines()` / `add_arcs()` batching	`sqldata.py`	Batch SQL INSERTs, reduce numbits round-trips
`should_trace()` sys.path check	`inorout.py`	Hash sys.path instead of full list comparison
`mapped_file_dict()` flush	`collector.py`	Snapshot strategy instead of retry loop

Parsing & Analysis (Phase 2)

Target	File	Approach
`PythonParser.parse_source()`	`parser.py`	Memoize tokenization, bulk newline indexing
`Analysis` set operations	`results.py`	Defer expensive calculations to lazy properties
SQLite query caching	`sqldata.py`	Cache `lines()`/`arcs()` results per context

Reporting (Phase 3)

Target	File	Approach
HTML report generation	`html.py`	Pre-compute analysis metadata, batch rendering
Path normalization	`files.py`	Verify cache hit rates, batch path ops

Results

No optimizations applied yet.

PRs

None yet.

PR	Branch	Status	Description

1.8 KiB Raw Blame History