Updates our Python interpreter discovery to conform to the rules
described in #2386, please see that issue for a full description of the
behavior. Briefly, we now will search for interpreters that satisfy a
requested version without stopping at the first Python executable.
Additionally, if retrieving information about an interpreter fails we
will continue to search for a working interpreter. We also add the
plumbing necessary to request Python implementations other than CPython,
though we do not add support for other implementations at this time.
A major internal goal of this work is to prepare for user-facing managed
toolchains i.e. fetching a requested version during `uv run`. These APIs
are not introduced, but there is some managed toolchain handling as
required for our test suite.
Some noteworthy implementation changes:
- The `uv_interpreter::find_python` module has been removed in favor of
a `uv_interpreter::discovery` module.
- There are new types to help structure interpreter requests and track
sources
- Executable discovery is implemented as a big lazy iterator and is a
central authority for source precedence
- `uv_interpreter::Error` variants were split into scoped types in each
module
- There's much more unit test coverage, but not for Windows yet
Remaining work:
- [x] Write new test cases
- [x] Determine correct behavior around executables in the current
directory
- _Future_: Combine `PythonVersion` and `VersionRequest`
- _Future_: Consider splitting `ManagedToolchain` into local and remote
variants
- _Future_: Add Windows unit test coverage
- _Future_: Explore behavior around implementation precedence (i.e.
CPython over PyPy)
Refactors split into:
- #3329
- #3330
- #3331
- #3332Closes#2386
Instead of saying
> we can conclude that you require==0a0.dev0 and
pandas-stubs==2.0.3.230814 are incompatible.
we'll say
> we can conclude that your requirements and pandas-stubs==2.0.3.230814
are incompatible.
Closes#3710
I'm not sure how to get unit test coverage for this, might look into
that. Ideally we'd skip this branch entirely?
Pubgrub stores incompatibilities as (package name, version range)
tuples, meaning it needs to clone the package name for each
incompatibility, and each non-borrowed operation on incompatibilities.
https://github.com/astral-sh/uv/pull/3673 made me realize that
`PubGrubPackage` has gotten large (expensive to copy), so like `Version`
and other structs, i've added an `Arc` wrapper around it.
It's a pity clippy forbids `.deref()`, it's less opaque than `&**` and
has IDE support (clicking on `.deref()` jumps to the right impl).
## Benchmarks
It looks like this matters most for complex resolutions which, i assume
because they carry larger `PubGrubPackageInner::Package` and
`PubGrubPackageInner::Extra` types.
```bash
hyperfine --warmup 5 "./uv-main pip compile -q ./scripts/requirements/jupyter.in" "./uv-branch pip compile -q ./scripts/requirements/jupyter.in"
hyperfine --warmup 5 "./uv-main pip compile -q ./scripts/requirements/airflow.in" "./uv-branch pip compile -q ./scripts/requirements/airflow.in"
hyperfine --warmup 5 "./uv-main pip compile -q ./scripts/requirements/boto3.in" "./uv-branch pip compile -q ./scripts/requirements/boto3.in"
```
```
Benchmark 1: ./uv-main pip compile -q ./scripts/requirements/jupyter.in
Time (mean ± σ): 18.2 ms ± 1.6 ms [User: 14.4 ms, System: 26.0 ms]
Range (min … max): 15.8 ms … 22.5 ms 181 runs
Benchmark 2: ./uv-branch pip compile -q ./scripts/requirements/jupyter.in
Time (mean ± σ): 17.8 ms ± 1.4 ms [User: 14.4 ms, System: 25.3 ms]
Range (min … max): 15.4 ms … 23.1 ms 159 runs
Summary
./uv-branch pip compile -q ./scripts/requirements/jupyter.in ran
1.02 ± 0.12 times faster than ./uv-main pip compile -q ./scripts/requirements/jupyter.in
```
```
Benchmark 1: ./uv-main pip compile -q ./scripts/requirements/airflow.in
Time (mean ± σ): 153.7 ms ± 3.5 ms [User: 165.2 ms, System: 157.6 ms]
Range (min … max): 150.4 ms … 163.0 ms 19 runs
Benchmark 2: ./uv-branch pip compile -q ./scripts/requirements/airflow.in
Time (mean ± σ): 123.9 ms ± 4.6 ms [User: 152.4 ms, System: 133.8 ms]
Range (min … max): 118.4 ms … 138.1 ms 24 runs
Summary
./uv-branch pip compile -q ./scripts/requirements/airflow.in ran
1.24 ± 0.05 times faster than ./uv-main pip compile -q ./scripts/requirements/airflow.in
```
```
Benchmark 1: ./uv-main pip compile -q ./scripts/requirements/boto3.in
Time (mean ± σ): 327.0 ms ± 3.8 ms [User: 344.5 ms, System: 71.6 ms]
Range (min … max): 322.7 ms … 334.6 ms 10 runs
Benchmark 2: ./uv-branch pip compile -q ./scripts/requirements/boto3.in
Time (mean ± σ): 311.2 ms ± 3.1 ms [User: 339.3 ms, System: 63.1 ms]
Range (min … max): 307.8 ms … 317.0 ms 10 runs
Summary
./uv-branch pip compile -q ./scripts/requirements/boto3.in ran
1.05 ± 0.02 times faster than ./uv-main pip compile -q ./scripts/requirements/boto3.in
```
<!--
Thank you for contributing to uv! To help us out with reviewing, please
consider the following:
- Does this pull request include a summary of the change? (See below.)
- Does this pull request include a descriptive title?
- Does this pull request include references to any relevant issues?
-->
## Summary
This PR falls back to writing an unnamed requirement if it appears to be
a relative URL. pip is way more flexible when providing an unnamed
requirement than when providing a PEP 508 requirement. For example,
_only_ this works:
```
black @ file:///Users/crmarsh/workspace/uv/scripts/packages/black_editable
```
Any other form will fail.
Meanwhile, _all_ of these work:
```
file:///Users/crmarsh/workspace/uv/scripts/packages/black_editable
scripts/packages/black_editable
./scripts/packages/black_editable
file:./scripts/packages/black_editable
file:scripts/packages/black_editable
```
Closes https://github.com/astral-sh/uv/issues/3180.
This makes use of the newly added `Ord` impl on `PubGrubPackage` to make
the output of `format_terms` independent of hashmap iteration order.
This was already collecting the terms into an intermediate `Vec`, so
sorting probably isn't going to add any significant overhead here.
(Plus, this is only running when formatting an error message after a
solution could not be found, so an extra sort doesn't seem like a big
deal here.)
Note that some tests are updated in this commit as a result of this
change. As far as I can tell, the semantic meaning of the output remains
the same. But the order of the listed packages does not.
Specific thing motivating this change is, in a subsequent, I added
`Option<MarkerTree>` to `PubGrubPackage::Package`, and this caused
similar changes in test output. So I backtracked and isolated this
change from the addition of `Option<MarkerTree>`.
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.
It looks like the last vestiges of `Derivative` were removed in commit
7eaed07f6c, but the then rendered
superfluous `derive(Derivative)` wasn't removed.
This is split out from workspaces support, which needs editables in the
bluejay commands. It consists mainly of refactorings:
* Move the `editable` module one level up.
* Introduce a `BuiltEditableMetadata` type for `(LocalEditable,
Metadata23, Requirements)`.
* Add editables to `InstalledPackagesProvider` so we can use
`EmptyInstalledPackages` for them.
## Summary
The main motivation here is that the `.filename()` method that we
implement on `Url` will do URL decoding for the last segment, which we
were missing here.
The errors are a bit awkward, because in
`crates/uv-resolver/src/lock.rs`, we wrap in `failed to extract filename
from URL: {url}`, so in theory we want the underlying errors to _omit_
the URL? But sometimes they use `#[error(transparent)]`?
## Summary
Uncertain about this, but we don't actually need the full
`SourceDistFilename`, only the name and version -- and we often have
that information already (as in the lockfile routines). So by flattening
the fields onto `RegistrySourceDist`, we can avoid re-parsing for
information we already have.
## Summary
This PR adds lossless deserialization for `GitSourceDist` distributions
in the lockfile. Specifically, we now properly preserve the requested
revision, the subdirectory, and the precise Git commit SHA.
## Test Plan
`cargo test`
## Summary
This PR introduces parallelism to the resolver. Specifically, we can
perform PubGrub resolution on a separate thread, while keeping all I/O
on the tokio thread. We already have the infrastructure set up for this
with the channel and `OnceMap`, which makes this change relatively
simple. The big change needed to make this possible is removing the
lifetimes on some of the types that need to be shared between the
resolver and pubgrub thread.
A related PR, https://github.com/astral-sh/uv/pull/1163, found that
adding `yield_now` calls improved throughput. With optimal scheduling we
might be able to get away with everything on the same thread here.
However, in the ideal pipeline with perfect prefetching, the resolution
and prefetching can run completely in parallel without depending on one
another. While this would be very difficult to achieve, even with our
current prefetching pattern we see a consistent performance improvement
from parallelism.
This does also require reverting a few of the changes from
https://github.com/astral-sh/uv/pull/3413, but not all of them. The
sharing is isolated to the resolver task.
## Test Plan
On smaller tasks performance is mixed with ~2% improvements/regressions
on both sides. However, on medium-large resolution tasks we see the
benefits of parallelism, with improvements anywhere from 10-50%.
```
./scripts/requirements/jupyter.in
Benchmark 1: ./target/profiling/baseline (resolve-warm)
Time (mean ± σ): 29.2 ms ± 1.8 ms [User: 20.3 ms, System: 29.8 ms]
Range (min … max): 26.4 ms … 36.0 ms 91 runs
Benchmark 2: ./target/profiling/parallel (resolve-warm)
Time (mean ± σ): 25.5 ms ± 1.0 ms [User: 19.5 ms, System: 25.5 ms]
Range (min … max): 23.6 ms … 27.8 ms 99 runs
Summary
./target/profiling/parallel (resolve-warm) ran
1.15 ± 0.08 times faster than ./target/profiling/baseline (resolve-warm)
```
```
./scripts/requirements/boto3.in
Benchmark 1: ./target/profiling/baseline (resolve-warm)
Time (mean ± σ): 487.1 ms ± 6.2 ms [User: 464.6 ms, System: 61.6 ms]
Range (min … max): 480.0 ms … 497.3 ms 10 runs
Benchmark 2: ./target/profiling/parallel (resolve-warm)
Time (mean ± σ): 430.8 ms ± 9.3 ms [User: 529.0 ms, System: 77.2 ms]
Range (min … max): 417.1 ms … 442.5 ms 10 runs
Summary
./target/profiling/parallel (resolve-warm) ran
1.13 ± 0.03 times faster than ./target/profiling/baseline (resolve-warm)
```
```
./scripts/requirements/airflow.in
Benchmark 1: ./target/profiling/baseline (resolve-warm)
Time (mean ± σ): 478.1 ms ± 18.8 ms [User: 482.6 ms, System: 205.0 ms]
Range (min … max): 454.7 ms … 508.9 ms 10 runs
Benchmark 2: ./target/profiling/parallel (resolve-warm)
Time (mean ± σ): 308.7 ms ± 11.7 ms [User: 428.5 ms, System: 209.5 ms]
Range (min … max): 287.8 ms … 323.1 ms 10 runs
Summary
./target/profiling/parallel (resolve-warm) ran
1.55 ± 0.08 times faster than ./target/profiling/baseline (resolve-warm)
```
## Summary
Fixes a small discrepancy between the pip compile outputs for
`annotation-style=split` and `annotation-style=line` commands.
### Problem
Consider the following `pyproject.toml` file.
```sh
$ cat pyproject.toml
[project]
name = "uv_test"
dynamic = ["version"]
dependencies = ["click"]
```
Running uv pip compile with annotation-style=split on uv 0.1.44 version
yields the following:
```sh
❯ uv pip compile pyproject.toml --annotation-style=split
Resolved 1 package in 2ms
# This file was autogenerated by uv via the following command:
# uv pip compile pyproject.toml --annotation-style=split
click==8.1.7
# via uv-test (pyproject.toml)
```
However, running uv pip compile with annotation-style=line doesn't
include source info for root level dependencies.
```sh
❯ uv pip compile pyproject.toml --annotation-style=line
Resolved 1 package in 1ms
# This file was autogenerated by uv via the following command:
# uv pip compile pyproject.toml --annotation-style=line
click==8.1.7
```
With this PR:
```sh
❯ ../target/debug/uv pip compile --annotation-style=line pyproject.toml
Resolved 1 package in 6ms
# This file was autogenerated by uv via the following command:
# uv pip compile --annotation-style=line pyproject.toml
click==8.1.7 # via uv-test (pyproject.toml)
```
This also now matches `pip-tools` output:
```sh
❯ pip-compile --annotation-style=line pyproject.toml
#
# This file is autogenerated by pip-compile with Python 3.12
# by the following command:
#
# pip-compile --annotation-style=line pyproject.toml
#
click==8.1.7 # via uv_test (pyproject.toml)
```
## Test Plan
`cargo test`
Following from #3595, we'd like wheels to make their way into the lock
file even if the current environment selects an sdist. With #3595, this
didn't happen:
$ cargo run -p uv -- pip compile -p3.10 <(echo psycopg2)
--unstable-uv-lock-file
$ cat uv.lock
version = 1
[[distribution]]
name = "psycopg2"
version = "2.9.9"
source = "registry+https://pypi.org/simple"
[distribution.sdist]
url =
"dc6acaf46d/psycopg2-2.9.9.tar.gz"
hash =
"sha256:d1454bde93fb1e224166811694d600e746430c006fbb031ea06ecc2ea41bf156"
The above example uses `psycopg2`, which has an sdist and wheels only on
Windows. Since I ran the above on Linux, an sdist was selected. But no
wheels appeared in the lock file.
With this PR, wheels are now correctly plumbed through:
$ cargo run -p uv -- pip compile -p3.10 <(echo psycopg2)
--unstable-uv-lock-file
$ cat uv.lock
version = 1
[[distribution]]
name = "psycopg2"
version = "2.9.9"
source = "registry+https://pypi.org/simple"
[distribution.sdist]
url =
"dc6acaf46d/psycopg2-2.9.9.tar.gz"
hash =
"sha256:d1454bde93fb1e224166811694d600e746430c006fbb031ea06ecc2ea41bf156"
[[distribution.wheel]]
url =
"2767d96391/psycopg2-2.9.9-cp310-cp310-win32.whl"
hash =
"sha256:38a8dcc6856f569068b47de286b472b7c473ac7977243593a288ebce0dc89516"
[[distribution.wheel]]
url =
"6572dec683/psycopg2-2.9.9-cp310-cp310-win_amd64.whl"
hash =
"sha256:426f9f29bde126913a20a96ff8ce7d73fd8a216cfb323b1f04da402d452853c3"
[[distribution.wheel]]
url =
"1fc5b9d33c/psycopg2-2.9.9-cp311-cp311-win32.whl"
hash =
"sha256:ade01303ccf7ae12c356a5e10911c9e1c51136003a9a1d92f7aa9d010fb98372"
[[distribution.wheel]]
url =
"5133dd3183/psycopg2-2.9.9-cp311-cp311-win_amd64.whl"
hash =
"sha256:121081ea2e76729acfb0673ff33755e8703d45e926e416cb59bae3a86c6a4981"
[[distribution.wheel]]
url =
"f74ffe6b6f/psycopg2-2.9.9-cp312-cp312-win32.whl"
hash =
"sha256:d735786acc7dd25815e89cc4ad529a43af779db2e25aa7c626de864127e5a024"
[[distribution.wheel]]
url =
"c4a26e1918/psycopg2-2.9.9-cp312-cp312-win_amd64.whl"
hash =
"sha256:a7653d00b732afb6fc597e29c50ad28087dcb4fbfb28e86092277a559ae4e693"
[[distribution.wheel]]
url =
"ffeb9ac356/psycopg2-2.9.9-cp37-cp37m-win32.whl"
hash =
"sha256:5e0d98cade4f0e0304d7d6f25bbfbc5bd186e07b38eac65379309c4ca3193efa"
[[distribution.wheel]]
url =
"0a39176d36/psycopg2-2.9.9-cp37-cp37m-win_amd64.whl"
hash =
"sha256:7e2dacf8b009a1c1e843b5213a87f7c544b2b042476ed7755be813eaf4e8347a"
[[distribution.wheel]]
url =
"86b90d30c4/psycopg2-2.9.9-cp38-cp38-win32.whl"
hash =
"sha256:ff432630e510709564c01dafdbe996cb552e0b9f3f065eb89bdce5bd31fabf4c"
[[distribution.wheel]]
url =
"c439b378ef/psycopg2-2.9.9-cp38-cp38-win_amd64.whl"
hash =
"sha256:bac58c024c9922c23550af2a581998624d6e02350f4ae9c5f0bc642c633a2d5e"
[[distribution.wheel]]
url =
"5080c0e61a/psycopg2-2.9.9-cp39-cp39-win32.whl"
hash =
"sha256:c92811b2d4c9b6ea0285942b2e7cac98a59e166d59c588fe5cfe1eda58e72d59"
[[distribution.wheel]]
url =
"ec73fe66d4/psycopg2-2.9.9-cp39-cp39-win_amd64.whl"
hash =
"sha256:de80739447af31525feddeb8effd640782cf5998e1a4e9192ebdf829717e3913"
Ref #3351
Our current flow of data from "simple registry package" to "final
resolved distribution" goes through a number of types:
* `SimpleMetadata` is the API response from a registry that includes all
published versions for a package. Each version has an assortment of
metadata
associated with it.
* `VersionFiles` is the aforementioned metadata. It is split in two: a
group of files for source distributions and a group of files for wheels.
* `PrioritizedDist` collects a subset of the files from `VersionFiles`
to form a selection of the "best" sdist and the "best" wheel for the
current environment.
* `CompatibleDist` is created from a borrowed `PrioritizedDist` that,
perhaps among other things, encapsulates the decision of whether to pick
an sdist or a wheel. (This decision depends both on compatibility and
the action being performed. e.g., When doing installation, a
`CompatibleDist` will sometimes select an sdist over a wheel.)
* `ResolvedDistRef` is like a `ResolvedDist`, but borrows a `Dist`.
* `ResolvedDist` is the almost-final-form of a distribution in a
resolution and is created from a `ResolvedDistRef`.
* `AnnotatedResolvedDist` is a new data type that is the actual final
form of a distribution that a universal lock file cares about. It
bundles a `ResolvedDist` with some metadata needed to generate a lock
file.
One of the requirements of a universal lock file is that we include all
wheels (and maybe all source distributions? but at least one if it's
present) associated with a distribution. But the above flow of data (in
the step from `VersionFiles` to `PrioritizedDist`) drops all wheels
except for the best one.
To remedy this, in this PR, we rejigger `PrioritizedDist`,
`CompatibleDist` and `ResolvedDistRef` so that all wheel data is
preserved. And when a `ResolvedDistRef` is finally turned into a
`ResolvedDist`, we copy all of the wheel data. And finally, we adjust
the `Lock` constructor to read this new data and include it in the lock
file. To make this work, we also modify `RegistryBuiltDist` so that it
can contain one or more wheels instead of just one.
One shortcoming here (called out in the code as a FIXME) is that if a
source distribution is selected as the "best" thing to use (perhaps
there are no compatible wheels), then the wheels won't end up in the
lock file. I plan to fix this in a follow-up PR.
We also aren't totally consistent on source distribution naming.
Sometimes we use `sdist`. Sometimes `source`. Sometimes `source_dist`.
I think it'd be nice to just use `sdist` everywhere, but I do prefer
the type names to be `SourceDist`. And sometimes you want function
names to match the type names (i.e., `from_source_dist`), which in turn
leads to an appearance of inconsistency. I'm open to ideas.
Closes#3351
## Summary
In `ResolutionGraph::from_state`, we have mechanisms to grab the hashes
and metadata for all distributions -- but we then throw that information
away. This PR preserves it on a new `AnnotatedDist` (yikes, open to
suggestions) that wraps `ResolvedDist` and includes (1) the hashes
(computed or from the registry) and (2) the `Metadata23`, which lets us
extract the version.
Closes https://github.com/astral-sh/uv/issues/3356.
Closes https://github.com/astral-sh/uv/issues/3357.
## Summary
Splits this into two loops that each handle independent cases, to make
the code a little easier to reason about. No behavioral or logic changes
-- just splitting the `match` across two loops.
## Summary
It's confusing that we use `constraints` here because constraints mean
something else for us (e.g., `--constraint constraints.txt`). These are
really the dependencies of a given `PubGrubPackage` -- the type is even
called `PubGrubDependencies`.
## Summary
I think this is overall good change because it explicitly encodes (in
the type system) something that was previously implicit. I'm not a huge
fan of the names here, open to input.
It covers some of https://github.com/astral-sh/uv/issues/3506 but I
don't think it _closes_ it.
