## Summary
When resolving Git metadata, we may be able to fetch the metadata from
GitHub directly in some cases. This is _way_ faster, since we don't need
to perform many Git operations and, in particular, don't need to clone
the repo.
This only works in the following cases:
- The Git repository is public. Otherwise, I believe you need an access
token, which we don't have.
- The `pyproject.toml` has static metadata.
- The `pyproject.toml` has no `tool.uv.sources`. Otherwise, we need to
lower them... And, if there are any paths or workspace sources, that
requires an install path (i.e., we need the content on-disk).
- The project is in the repo root. If it's in a subdirectory, it could
be a workspace member. And if it's a workspace member, there could be
sources defined in the workspace root. But we can't know without
fetching the workspace root -- and we need the workspace in order to
find the root...
Closes#10568.
## Summary
Closes#3312.
This PR adds Git LFS support to the `uv-git` crate by using the
`git-lfs` CLI to fetch required LFS objects for a revision following the
call to `git fetch`.
The LFS fetch step is disabled by default and only enabled if the
environment variable `UV_GIT_LFS` is set.
When enabled, the LFS fetch step is run for all repositories regardless
of whether they have associated LFS objects. The step is skipped if the
`git-lfs` CLI tool isn't installed.
## Test Plan
I verified that the minimal example in the linked issue passes, i.e.
this command now succeeds:
```sh
UV_GIT_LFS=1 uv pip install git+https://github.com/grebnetiew/lfs-py.git
```
I also verified that non-LFS repositories still work, with or without
`git-lfs` installed.
### To Replicate
Attempt to use uv to install a Git dependency that contains LFS objects
(e.g. `uv pip install git+https://github.com/grebnetiew/lfs-py.git`).
This should fail with a smudge filter error.
Re-run the same command with the added environment variable
`UV_GIT_LFS=1`. The install should now succeed.
## Potential Changes / Improvements
~With this change LFS objects in a given revision will always be
downloaded if the user has Git LFS installed, which may not always be
desired behavior. It might be helpful to add a field to the `uv`
settings and/or an environment variable so that the LFS step can be
disabled if needed.~
Enabling/disabled via environment variable has now been implemented.
---------
Co-authored-by: Sydney Duckworth <sydduckworth@users.noreply.github.com>
Co-authored-by: Zanie Blue <contact@zanie.dev>
## Summary
Sort of undecided on this. These are already stored as `dyn Reporter` in
each struct, so we're already using dynamic dispatch in that sense. But
all the methods take `impl Reporter`. This is sometimes nice (the
callsites are simpler?), but it also means that in practice, you often
_can't_ pass `None` to these methods that accept `Option<impl
Reporter>`, because Rust can't infer the generic type.
Anyway, this adds more consistency and simplifies the setup by using
`Arc<dyn Reporter>` everywhere.
## Summary
These were moved as part of a broader refactor to create a single
integration test module. That "single integration test module" did
indeed have a big impact on compile times, which is great! But we aren't
seeing any benefit from moving these tests into their own files (despite
the claim in [this blog
post](https://matklad.github.io/2021/02/27/delete-cargo-integration-tests.html),
I see the same compilation pattern regardless of where the tests are
located). Plus, we don't have many of these, and same-file tests is such
a strong Rust convention.
## Summary
This is part of making
https://github.com/astral-sh/uv/issues/7299#issuecomment-2385286341
better. You can now use `tool.uv.dependency-metadata` for direct URL
requirements. Unfortunately, you _must_ include a version, since we need
one to perform resolution.
## Summary
Cache the path to git executable in a `LazyLock` and reuse it throughout
the process. This might reduce some costs on finding the git executable.
## Summary
This PR declares and documents all environment variables that are used
in one way or another in `uv`, either internally, or externally, or
transitively under a common struct.
I think over time as uv has grown there's been many environment
variables introduced. Its harder to know which ones exists, which ones
are missing, what they're used for, or where are they used across the
code. The docs only documents a handful of them, for others you'd have
to dive into the code and inspect across crates to know which crates
they're used on or where they're relevant.
This PR is a starting attempt to unify them, make it easier to discover
which ones we have, and maybe unlock future posibilities in automating
generating documentation for them.
I think we can split out into multiple structs later to better organize,
but given the high influx of PR's and possibly new environment variables
introduced/re-used, it would be hard to try to organize them all now
into their proper namespaced struct while this is all happening given
merge conflicts and/or keeping up to date.
I don't think this has any impact on performance as they all should
still be inlined, although it may affect local build times on changes to
the environment vars as more crates would likely need a rebuild. Lastly,
some of them are declared but not used in the code, for example those in
`build.rs`. I left them declared because I still think it's useful to at
least have a reference.
Did I miss any? Are their initial docs cohesive?
Note, `uv-static` is a terrible name for a new crate, thoughts? Others
considered `uv-vars`, `uv-consts`.
## Test Plan
Existing tests
As per
https://matklad.github.io/2021/02/27/delete-cargo-integration-tests.html
Before that, there were 91 separate integration tests binary.
(As discussed on Discord — I've done the `uv` crate, there's still a few
more commits coming before this is mergeable, and I want to see how it
performs in CI and locally).
## Summary
When syncing a lockfile, we need to respect credentials defined in the
`pyproject.toml`, even if they won't be used for resolution.
Unfortunately, this includes credentials in `tool.uv.sources`,
`tool.uv.dev-dependencies`, `project.dependencies`, and
`project.optional-dependencies`.
Closes https://github.com/astral-sh/uv/issues/7453.
## Summary
This PR adds a more flexible cache invalidation abstraction for uv, and
uses that new abstraction to improve support for dynamic metadata.
Specifically, instead of relying solely on a timestamp, we now pass
around a `CacheInfo` struct which (as of now) contains
`Option<Timestamp>` and `Option<Commit>`. The `CacheInfo` is saved in
`dist-info` as `uv_cache.json`, so we can test already-installed
distributions for cache validity (along with testing _cached_
distributions for cache validity).
Beyond the defaults (`pyproject.toml`, `setup.py`, and `setup.cfg`
changes), users can also specify additional cache keys, and it's easy
for us to extend support in the future. Right now, cache keys can either
be instructions to include the current commit (for `setuptools_scm` and
similar) or file paths (for `hatch-requirements-txt` and similar):
```toml
[tool.uv]
cache-keys = [{ file = "requirements.txt" }, { git = true }]
```
This change should be fully backwards compatible.
Closes https://github.com/astral-sh/uv/issues/6964.
Closes https://github.com/astral-sh/uv/issues/6255.
Closes https://github.com/astral-sh/uv/issues/6860.
This is achieved by updating the `LockedFile::acquire` API to be async —
as in some cases we were attempting to acquire the lock synchronously,
i.e., without yielding, which blocked the runtime.
Closes https://github.com/astral-sh/uv/issues/6691 — I tested with the
reproduction there and a local release build and no longer reproduce the
deadlock with these changes.
Some additional context in the [internal Discord
thread](1278478941)
## Summary
The strategy here is: if the user provides supported environments, we
use those as the initial forks when resolving. As a result, we never add
or explore branches that are disjoint with the supported environments.
(If the supported environments change, we ignore the lockfile entirely,
so we don't have to worry about any interactions between supported
environments and the preference forks.)
Closes https://github.com/astral-sh/uv/issues/6184.
## Summary
We retain them if you use `--raw-sources`, but otherwise they're
removed. We still respect them in the subsequent `uv.lock` via an
in-process store.
Closes#6056.
## Summary
Whenever we call `resolve`, we immediately call `fetch` after. And in
some cases `resolve` actually calls `fetch` internally. It seems a lot
simpler to just merge these into one method that returns a `Fetch`
(which itself contains the fully-resolved URL).
Closes https://github.com/astral-sh/uv/issues/5876.
## Summary
Very subtle bug. The scenario is as follows:
- We resolve: `elmer-circuitbuilder = { git =
"https://github.com/ElmerCSC/elmer_circuitbuilder.git" }`
- The user then changes the request to: `elmer-circuitbuilder = { git =
"https://github.com/ElmerCSC/elmer_circuitbuilder.git", rev =
"44d2f4b19d6837ea990c16f494bdf7543d57483d" }`
- When we go to re-lock, we note two facts:
1. The "default branch" resolves to
`44d2f4b19d6837ea990c16f494bdf7543d57483d`.
2. The metadata for `44d2f4b19d6837ea990c16f494bdf7543d57483d` is
(whatever we grab from the lockfile).
- In the resolver, we then ask for the metadata for
`44d2f4b19d6837ea990c16f494bdf7543d57483d`. It's already in the cache,
so we return it; thus, we never add the
`44d2f4b19d6837ea990c16f494bdf7543d57483d` ->
`44d2f4b19d6837ea990c16f494bdf7543d57483d` mapping to the Git resolver,
because we never have to resolve it.
This would apply for any case in which a requested tag or branch was
replaced by its precise SHA. Replacing with a different commit is fine.
It only applied to `tool.uv.sources`, and not PEP 508 URLs, because the
underlying issue is that we aren't consistent about "automatically"
extracting the precise commit from a Git reference.
Closes https://github.com/astral-sh/uv/issues/5860.
## Summary
The current receipt doesn't capture quite enough information. For
example, it doesn't differentiate between editable and non-editable
requirements. This PR instead uses the full `Requirement` type. I think
we should use a custom representation like we do in the lockfile, but
I'm just using the default representation to demonstrate the idea.
Fixes a concurrency issue when multiple processes are installing the
same package in different virtual environments from Git ref (not a
specific Git commit).
## Symptoms
That's how some of symptoms looked like in our case:
```
DEBUG uv 0.2.21
DEBUG Checking for Python interpreter at path `/tmp/pytest-of-runner/pytest-0/tmp_venv_dir/python3.12/37bf51bfba4699a940ce31349422b24a5bc55a2b179ed7aec74459a9ae8d57b7/bin/python`
DEBUG Using Python 3.12.4 environment at /tmp/pytest-of-runner/pytest-0/tmp_venv_dir/python3.12/37bf51bfba4699a940ce31349422b24a5bc55a2b179ed7aec74459a9ae8d57b7/bin/python
DEBUG Acquired lock for `/tmp/pytest-of-runner/pytest-0/tmp_venv_dir/python3.12/37bf51bfba4699a940ce31349422b24a5bc55a2b179ed7aec74459a9ae8d57b7`
DEBUG At least one requirement is not satisfied: torch
DEBUG Using request timeout of 300s
DEBUG Found 37 packages in `--find-links` entry: /tmp/pytest-of-runner/pytest-0/tmp_venv_dir/python3.12/.cache/pip/wheels
DEBUG Updating git source `Url { scheme: "https", cannot_be_a_base: false, username: "***", password: None, host: Some(Domain("github.com")), port: None, path: "/iterative/datachain", query: None, fragment: None }`
DEBUG Attempting GitHub fast path for: https://api.github.com/repos/iterative/datachain/commits/fix-distributed-test
DEBUG failed to check github HTTP status client error (404 Not Found) for url (https://api.github.com/repos/iterative/datachain/commits/fix-distributed-test)
DEBUG Performing a Git fetch for: https://***@github.com/iterative/datachain
error: Failed to download and build: `datachain @ git+https://***@github.com/iterative/datachain@fix-distributed-test`
Caused by: Git operation failed
Caused by: process didn't exit successfully: `git clone --local /tmp/pytest-of-runner/pytest-0/tmp_venv_dir/python3.12/.cache/uv/git-v0/db/9d45a3e6f56b0a69 /tmp/pytest-of-runner/pytest-0/tmp_venv_dir/python3.12/.cache/uv/git-v0/checkouts/9d45a3e6f56b0a69/56b15b8` (exit status: 128)
--- stderr
fatal: destination path '/tmp/pytest-of-runner/pytest-0/tmp_venv_dir/python3.12/.cache/uv/git-v0/checkouts/9d45a3e6f56b0a69/56b15b8' already exists and is not an empty directory.
```
## Cause of the issue
It is the same command that is failing - `git clone`, and I think it's
happening because it was trying to first get the repo to dereference the
`fix-distributed-test` branch:
`Given a remote source distribution, return a precise variant, if
possible.`
And it's happening w/i acquiring a lock around cache.
## Fix
I thinks we can reuse the existing `fetch` method that has already lock
around cache:
https://github.com/astral-sh/uv/pull/5051/files#diff-f58bb99dee2c4922d156ace3e7de651f0d9a81fc8e9447a2ad865de5c53543fcR61-R68
```python
// Avoid races between different processes, too.
let lock_dir = cache.join("locks");
....
```
## Questions
- Are there any tests that cover concurrency? I'm quite new to Rust and
if someone can point me to some examples and I can create a similar test
or a new one.
- Is error handling done correctly in this PR (again, I'm new to Rust -
I'll review and read about it, but it's better also for someone else to
review this)
## Summary
`GitDatabase::contains` previously only parsed the commit to see if it
was a valid hash and didn't verify if the commit existed in the object
database. This led to the database never being updated.
Resolves https://github.com/astral-sh/uv/issues/4378.
## Test Plan
Added a test that fails without this change.
## Summary
It turns out that the Git fetch implementation is initializing its own
client, which can be really expensive on macOS (due to loading native
certificates) _and_ bypasses any of our middleware. This PR modifies the
Git implementation to accept a shared client.
## Summary
We unconditionally update the submodules in our Git code, but AFAICT it
shouldn't be necessary if we already have a complete, up-to-date fetch
available.
## Summary
This PR ensures that if a lockfile already contains a resolved reference
(e.g., you locked with `main` previously, and it locked to a specific
commit), and you run `uv lock`, we use the same SHA, even if it's not
the latest SHA for that tag. This avoids upgrading Git dependencies
without `--upgrade`.
Closes#3920.
## Summary
This PR removes the static resolver map:
```rust
static RESOLVED_GIT_REFS: Lazy<Mutex<FxHashMap<RepositoryReference, GitSha>>> =
Lazy::new(Mutex::default);
```
With a `GitResolver` struct that we now pass around on the
`BuildContext`. There should be no behavior changes here; it's purely an
internal refactor with an eye towards making it cleaner for us to
"pre-populate" the list of resolved SHAs.
## Summary
We currently rely on libgit2 for most git-related functionality.
However, libgit2 has long-standing performance issues, as well as lags
significantly behind git in terms of new features. For these reasons we
now use the git CLI by default for fetching repositories
(https://github.com/astral-sh/uv/pull/1781). This PR completely drops
libgit2 in favor of the git CLI for all git-related functionality, which
should allow us to use features such as partial clones and sparse
checkouts in the future for performance.
There is also a lot of technical debt in the current git code as it's
mostly taken from Cargo. Switching to the git CLI *vastly* simplifies
the `uv-git` codebase.
Eventually we might want to look into switching to
[`gitoxide`](https://github.com/Byron/gitoxide), but it's currently too
immature for our use case.
It turns out that we use PubGrubPackage as the key in hashmaps in a fair
few places. And when we iterate over hashmaps, the order is unspecified.
This can in turn result in changes in output as a result of changes in
the PubGrubPackage definition, purely as a function of its changing
hash. This is confusing as there should be no semantic difference.
Thus, this is a precursor to introducing some more determinism to places
I found in the error reporting whose output depending on hashmap
iteration order.
