Share downloaded URL wheels across hash fragments

[charliermarsh]

Summary

We omit known hash fields from CanonicalUrl, preserving source selectors and unknown fragments, and index downloaded URL wheels by their canonical URL and computed hashes. This lets uv sync --offline reuse a wheel downloaded from a fragment-hashed URL when the lockfile stores its URL and hashes separately.

Keep HTTP metadata, source builds, and in-flight downloads scoped to their original hash declarations. Sharing checks the complete hash policy and reads metadata from the selected archive, so a newer revision cannot hide an older cached wheel or pair new wheel bytes with old metadata. Existing cache keys remain valid, and references to older revisions survive cache pruning.

Apply the same filename, hash, size, and archive-existence checks to every cached URL wheel candidate. When online, refresh cached contents that no longer match the requested hashes or size before resolving dependencies. Fresh integrity failures are rejected without an extra retry, including in the full-download fallback.

Follows #21279, which added URL hash reuse during resolution.

[konstin]

That's the same as #21210, though I don't mind which one we merge.

[codspeed]

Merging this PR will improve performance by 8.55%

⚡ 3 improved benchmarks
✅ 22 untouched benchmarks
⏩ 12 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	resolve_warm_jupyter	72.9 ms	66 ms	+10.47%
⚡	Simulation	resolve_warm_jupyter_universal	314.8 ms	289.6 ms	+8.71%
⚡	Simulation	resolve_warm_airflow	800.2 ms	751.4 ms	+6.51%

Tip

Curious why performance improved? Comment @codspeedbot explain why performance improved on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing charlie/normalize-wheel-cache-hashes (558b601) with main (a22198b)}

Footnotes

1. 12 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[astral-sh-bot]

uv test inventory changes

This PR changes the tests when compared with the main base revision.

• Added tests: 8
• Removed tests: 0
• Changed suites: 5

uv-cache: +1 / -0

Added:

• uv-cache::wheel::tests::archive_and_metadata_url_identity

Removed: none

uv-cache-key: +2 / -0

Added:

• uv-cache-key::canonical_url::tests::canonical_url_ignores_known_hash_fragments
• uv-cache-key::canonical_url::tests::canonical_url_preserves_other_fragments

Removed: none

uv::lock: +2 / -0

Added:

• uv::lock::lock::lock_wheel_url_fragment_download
• uv::lock::lock::lock_wheel_url_hash_cache

Removed: none

uv::pip: +2 / -0

Added:

• uv::pip::pip_sync::direct_url_invalid_cache
• uv::pip::pip_sync::direct_url_wheel_cache_refresh

Removed: none

uv::pip_install: +1 / -0

Added:

• uv::pip_install::pip_install::direct_url_hash_cache_metadata

Removed: none