## Summary
Our current handling of `--find-links` merges the entries in each index.
As a result, we can end up with `AnnotatedDist` entries that reference
distributions across indexes.
I'd like to change `--find-links` such that each `--find-links` entry is
just treated as its own index (so, e.g., if `requests` exists in the
first `--find-links` entry, we don't even check the registry by
default), which would _also_ fix this problem automatically. But that's
a behavior change... So for now, in the lockfile, we filter
distributions that don't match the source index URL.
There are two cases to consider:
- There's a source distribution. Then, for the ID to reference the
`--find-links` registry, the source distribution _must_ have come from
the `--find-links` entry, so it's fine to discard any wheels from the
"wrong" registry without breaking any compatibility guarantees.
- There's no source distribution. Then the best wheel must come from the
`--find-links` registry. We might lose some platform coverage by
discarding the other wheels, but it shouldn't break any of the
"guarantees", since we have at least one wheel that fits in the version
range.
Closes https://github.com/astral-sh/uv/issues/6015.
## Summary
Added the actual error message to the warning when uv fails to parse
`pyproject.toml`.
Resolves https://github.com/astral-sh/uv/issues/5934
## Test Plan
Took the case from the issue:
- have `pyproject.toml` which contains
```
[tool.uv]
foobar = false
```
-
```
$ uv venv --preview -v
```
- Expect the message that contains the actual problem in the
`pyproject.toml` like:
```
warning: Failed to parse `pyproject.toml` during settings discovery: unknown field `foobar`; skipping...
```
## Summary
A lot of the existing tests were no-ops. For convenience, we now use the
trick of: install from Test PyPI (to get an outdated "latest"), then
upgrade from PyPI.
Right now, the URL gets out-of-sync with the install path, since the
install path is canonicalized. This leads to a subtle error on Windows
(in CI) in which we don't preserve caching across resolution and
installation.
Surprisingly, this is a lockfile schema change: We can't store relative
paths in urls, so we have to store a `filename` entry instead of the
whole url.
Fixes#4355
## Summary
We now persist the `ResolverInstallerOptions` when writing out a tool
receipt. When upgrading, we grab the saved options, and merge with the
command-line arguments and user-level filesystem settings (CLI > receipt
> filesystem).
The loose consensus is that "fetch" doesn't have much meaning and that a
boolean flag makes more sense from the command line.
1. Adds `--allow-python-downloads` (hidden, default) and
`--no-python-downloads` to the CLI to quickly enable or disable
downloads
2. Deprecates `--python-fetch` in favor of the options from (1)
3. Removes `python-fetch` in favor of a `python-downloads` setting
5. Adds a `never` variant to the enum, allowing even explicit installs
to be disabled via the configuration file
## Test plan
I tested this with various `pyproject.toml`-level settings and `uv venv
--preview --python 3.12.2` and `uv python install 3.12.2` with and
without the new CLI flags.
Warn when there are missing bounds on transitive dependencies with
`--resolution lowest`.
Implemented as a lazy resolution graph check. Dev deps are odd because
they are missing the edge from the root that extras have (they are
currently orphans in the resolution graph), but this is more complex to
solve properly because we can put dev dep information in a `Requirement`
so i special cased them here.
Closes#2797
Should help with #1718
---------
Co-authored-by: Ibraheem Ahmed <ibraheem@ibraheem.ca>
## Summary
This PR rewrites the `MarkerTree` type to use algebraic decision
diagrams (ADD). This has many benefits:
- The diagram is canonical for a given marker function. It is impossible
to create two functionally equivalent marker trees that don't refer to
the same underlying ADD. This also means that any trivially true or
unsatisfiable markers are represented by the same constants.
- The diagram can handle complex operations (conjunction/disjunction) in
polynomial time, as well as constant-time negation.
- The diagram can be converted to a simplified DNF form for user-facing
output.
The new representation gives us a lot more confidence in our marker
operations and simplification, which is proving to be very important
(see https://github.com/astral-sh/uv/pull/5733 and
https://github.com/astral-sh/uv/pull/5163).
Unfortunately, it is not easy to split this PR into multiple commits
because it is a large rewrite of the `marker` module. I'd suggest
reading through the `marker/algebra.rs`, `marker/simplify.rs`, and
`marker/tree.rs` files for the new implementation, as well as the
updated snapshots to verify how the new simplification rules work in
practice. However, a few other things were changed:
- [We now use release-only comparisons for `python_full_version`, where
we previously only did for
`python_version`](https://github.com/astral-sh/uv/blob/ibraheem/canonical-markers/crates/pep508-rs/src/marker/algebra.rs#L522).
I'm unsure how marker operations should work in the presence of
pre-release versions if we decide that this is incorrect.
- [Meaningless marker expressions are now
ignored](https://github.com/astral-sh/uv/blob/ibraheem/canonical-markers/crates/pep508-rs/src/marker/parse.rs#L502).
This means that a marker such as `'x' == 'x'` will always evaluate to
`true` (as if the expression did not exist), whereas we previously
treated this as always `false`. It's negation however, remains `false`.
- [Unsatisfiable markers are written as `python_version <
'0'`](https://github.com/astral-sh/uv/blob/ibraheem/canonical-markers/crates/pep508-rs/src/marker/tree.rs#L1329).
- The `PubGrubSpecifier` type has been moved to the new `uv-pubgrub`
crate, shared by `pep508-rs` and `uv-resolver`. `pep508-rs` also depends
on the `pubgrub` crate for the `Range` type, we probably want to move
`pubgrub::Range` into a separate crate to break this, but I don't think
that should block this PR (cc @konstin).
There is still some remaining work here that I decided to leave for now
for the sake of unblocking some of the related work on the resolver.
- We still use `Option<MarkerTree>` throughout uv, which is unnecessary
now that `MarkerTree::TRUE` is canonical.
- The `MarkerTree` type is now interned globally and can potentially
implement `Copy`. However, it's unclear if we want to add more
information to marker trees that would make it `!Copy`. For example, we
may wish to attach extra and requires-python environment information to
avoid simplifying after construction.
- We don't currently combine `python_full_version` and `python_version`
markers.
- I also have not spent too much time investigating performance and
there is probably some low-hanging fruit. Many of the test cases I did
run actually saw large performance improvements due to the markers being
simplified internally, reducing the stress on the old `normalize`
routine, especially for the extremely large markers seen in
`transformers` and other projects.
Resolves https://github.com/astral-sh/uv/issues/5660,
https://github.com/astral-sh/uv/issues/5179.
This PR tweaks the change made in #5904 so that the `profiling` Cargo
profile does _not_ have LTO enabled. With LTO enabled, compile times
even after just doing a `touch crates/uv/src/bin/uv.rs` are devastating:
$ cargo b --profile profiling -p uv
Compiling uv-cli v0.0.1 (/home/andrew/astral/uv/crates/uv-cli)
Compiling uv v0.2.34 (/home/andrew/astral/uv/crates/uv)
Finished `profiling` profile [optimized + debuginfo] target(s) in 3m 47s
Even with `lto = "thin"`, compile times are not great, but an
improvement:
$ cargo b --profile profiling -p uv
Compiling uv v0.2.34 (/home/andrew/astral/uv/crates/uv)
Finished `profiling` profile [optimized + debuginfo] target(s) in 53.98s
But our original configuration for `profiling`, prior to #5904, was with
LTO completely disabled:
$ cargo b --profile profiling -p uv
Compiling uv v0.2.34 (/home/andrew/astral/uv/crates/uv)
Finished `profiling` profile [optimized + debuginfo] target(s) in 30.09s
This gives reasonable-ish compile times, although I still want them to
be better.
This setup does risk that we are measuring something in benchmarks that
we are shipping, but in order to make those two the same, we'd either
need to make compile times way worse for development, or take a hit
to binary size and a slight hit to runtime performance in our release
builds. I would weakly prefer that we accept the hit to runtime
performance and binary size in order to bring our measurements in line
with what we ship, but I _strongly_ feel that we should not have compile
times exceeding minutes for development. When doing performance testing,
long compile times, for me anyway, break "flow" state.
A confounding factor here was that #5904 enabled LTO for the `release`
profile, but the `dist` profile (used by `cargo dist`) was still setting
it to `lto = "thin"`. However, because of shenanigans in our release
pipeline, we we actually using the `release` profile for binaries we
ship. This PR does not make any changes here other than to remove `lto =
"thin"` from the `dist` profile to make the fact that they are the same
a bit clearer.
cc @davfsa
## Summary
This PR adds a `DistExtension` field to some of our distribution types,
which requires that we validate that the file type is known and
supported when parsing (rather than when attempting to unzip). It
removes a bunch of extension parsing from the code too, in favor of
doing it once upfront.
Closes https://github.com/astral-sh/uv/issues/5858.
