## Summary
Implement basic *Goto type definition* support for Red Knot's LSP.
This PR also builds the foundation for other LSP operations. E.g., Goto
definition, hover, etc., should be able to reuse some, if not most,
logic introduced in this PR.
The basic steps of resolving the type definitions are:
1. Find the closest token for the cursor offset. This is a bit more
subtle than I first anticipated because the cursor could be positioned
right between the callee and the `(` in `call(test)`, in which case we
want to resolve the type for `call`.
2. Find the node with the minimal range that fully encloses the token
found in 1. I somewhat suspect that 1 and 2 could be done at the same
time but it complicated things because we also need to compute the spine
(ancestor chain) for the node and there's no guarantee that the found
nodes have the same ancestors
3. Reduce the node found in 2. to a node that is a valid goto target.
This may require traversing upwards to e.g. find the closest expression.
4. Resolve the type for the goto target
5. Resolve the location for the type, return it to the LSP
## Design decisions
The current implementation navigates to the inferred type. I think this
is what we want because it means that it correctly accounts for
narrowing (in which case we want to go to the narrowed type because
that's the value's type at the given position). However, it does have
the downside that Goto type definition doesn't work whenever we infer `T
& Unknown` because intersection types aren't supported. I'm not sure
what to do about this specific case, other than maybe ignoring `Unkown`
in Goto type definition if the type is an intersection?
## Known limitations
* Types defined in the vendored typeshed aren't supported because the
client can't open files from the red knot binary (we can either
implement our own file protocol and handler OR extract the typeshed
files and point there). See
https://github.com/astral-sh/ruff/issues/17041
* Red Knot only exposes an API to get types for expressions and
definitions. However, there are many other nodes with identifiers that
can have a type (e.g. go to type of a globals statement, match patterns,
...). We can add support for those in separate PRs (after we figure out
how to query the types from the semantic model). See
https://github.com/astral-sh/ruff/issues/17113
* We should have a higher-level API for the LSP that doesn't directly
call semantic queries. I intentionally decided not to design that API
just yet.
## Test plan
https://github.com/user-attachments/assets/fa077297-a42d-4ec8-b71f-90c0802b4edb
Goto type definition on a union
<img width="1215" alt="Screenshot 2025-04-01 at 13 02 55"
src="https://github.com/user-attachments/assets/689cabcc-4a86-4a18-b14a-c56f56868085"
/>
Note: I recorded this using a custom typeshed path so that navigating to
builtins works.
## Summary
`std::time::now` isn't available on `wasm32-unknown-unknown` but it is
used by `FileTime::now`.
This PR replaces the usages of `FileTime::now` with a target specific
helper function that we already had in the memory file system.
Fixes https://github.com/astral-sh/ruff/issues/16966
## Test Plan
Tested that the playground no longer crash when adding an extra-path
... and switch to the new one.
We do this switch by converting the old diagnostics to a
`Diagnostic`, and then rendering that.
This does not quite emit identical output. There are some
changes. They *could* be fixed to remain the same, but the
changes aren't obviously worse to me and I think the right
way to *improve* them is to move Red Knot to the new `Diagnostic`
API.
The next commit will have the snapshot changes.
In our existing diagnostics, our message is just the diagnostic
ID, and the message goes to the annotation. In reality, the
diagnostic can have its own message distinct from the optional
messages associated with an annotation.
In order to make the outputs match, we do a small tweak here:
when the main diagnostic message is empty, we drop the colon
after the diagnostic ID.
I expect that we'll want to rejigger this output format more
in the future, but for now this was a very simple change to
preserve the status quo.
When moving over to the new renderer, I noticed that it
was emitting an extra line terminator compared to the status
quo. This removes it by turning the line terminator into a
line delimiter between diagnostics.
## Summary
Another salsa upgrade.
The main motivation is to stay on a recent salsa version because there
are still a lot of breaking changes happening.
The most significant changes in this update:
* Salsa no longer derives `Debug` by default. It now requires
`interned(debug)` (or similar)
* This version ships the foundation for garbage collecting interned
values. However, this comes at the cost that queries now track which
interned values they created (or read). The micro benchmarks in the
salsa repo showed a significant perf regression. Will see if this also
visible in our benchmarks.
## Test Plan
`cargo test`
## Summary
This PR implements the first part of
https://github.com/astral-sh/ruff/discussions/16440. It ensures that Red
Knot's module resolver is case sensitive on all systems.
This PR combines a few approaches:
1. It uses `canonicalize` on non-case-sensitive systems to get the real
casing of a path. This works for as long as no symlinks or mapped
network drives (the windows `E:\` is mapped to `\\server\share` thingy).
This is the same as what Pyright does
2. If 1. fails, fall back to recursively list the parent directory and
test if the path's file name matches the casing exactly as listed in by
list dir. This is the same approach as CPython takes in its module
resolver. The main downside is that it requires more syscalls because,
unlike CPython, we Red Knot needs to invalidate its caches if a file
name gets renamed (CPython assumes that the folders are immutable).
It's worth noting that the file watching test that I added that renames
`lib.py` to `Lib.py` currently doesn't pass on case-insensitive systems.
Making it pass requires some more involved changes to `Files`. I plan to
work on this next. There's the argument that landing this PR on its own
isn't worth it without this issue being addressed. I think it's still a
good step in the right direction even when some of the details on how
and where the path case sensitive comparison is implemented.
## Test plan
I added multiple integration tests (including a failing one). I tested
that the `case-sensitivity` detection works as expected on Windows,
MacOS and Linux and that the fast-paths are taken accordingly.
We don't actually hook this up to anything in this PR, but we do
go to some trouble to granularly unit test it. The unit tests caught
plenty of bugs after I initially wrote down the implementation, so they
were very much worth it.
Closes#16506
Instead of hard-coding a specific context window,
it seemed prudent to make this configurable. That
makes it easier to test different context window
sizes as well.
I am not totally convinced that this is the right
place for this configuration. I could see the context
window size being a property of `Diagnostic` instead,
since we might want to change the context window
size based not just on some end user configuration,
but perhaps also the specific diagnostic.
But for now, I think it's fine for it to live here,
and all of the rendering logic doesn't care where
it lives. So it should be relatively easy to change
in the future.
This adds a new configuration knob to diagnostic rendering that, when
enabled, will make diagnostic rendering much more terse. Specifically,
it will guarantee that each diagnostic will only use one line.
This doesn't actually hook the concise output option up to anything.
We'll do that plumbing in the next commit.
## Summary
This PR introduces a new mdtest option `system` that can either be
`in-memory` or `os`
where `in-memory` is the default.
The motivation for supporting `os` is so that we can write OS/system
specific tests
with mdtests. Specifically, I want to write mdtests for the module
resolver,
testing that module resolution is case sensitive.
## Test Plan
I tested that the case-sensitive module resolver test start failing when
setting `system = "os"`
... with supporting types. This is meant to give us a base to work with
in terms of our new diagnostic data model. I expect the representations
to be tweaked over time, but I think this is a decent start.
I would also like to add doctest examples, but I think it's better if we
wait until an initial version of the renderer is done for that.
This puts them out of the way so that they can hopefully be removed more
easily in the (near) future, and so that they don't get in the way of
the new types. This also makes the intent of the migration a bit clearer
in the code and hopefully results in less confusion.
This trait should eventually go away, so we rename it (and supporting
types) to make room for a new concrete `Diagnostic` type.
This commit is just the rename. In the next commit, we'll move it to a
different module.
Summary
--
This is a follow up addressing the comments on #16425. As @dhruvmanila
pointed out, the naming is a bit tricky. I went with `has_no_errors` to
try to differentiate it from `is_valid`. It actually ends up negated in
most uses, so it would be more convenient to have `has_any_errors` or
`has_errors`, but I thought it would sound too much like the opposite of
`is_valid` in that case. I'm definitely open to suggestions here.
Test Plan
--
Existing tests.
This is a small little hack to make the `Diagnostic` trait
capable of supporting attaching multiple spans.
This design should be considered transient. This was just the
quickest way that I could see to pass multiple spans through from
the type checker to the diagnostic renderer.
This commit has no behavioral changes.
This refactor moves the logic for turning a `D: Diagnostic` into
an `annotate_snippets::Message` into its own types. This would
ideally just be a function or something, but the `annotate-snippets`
types want borrowed data, and sometimes we need to produce owned
data. So we gather everything we need into our own types and then
spit it back out in the format that `annotate-snippets` wants.
This factor was motivated by wanting to render multiple snippets.
The logic for generating a code frame is complicated enough that
it's worth splitting out so that we can reuse it for other spans.
(Note that one should consider this prototype-level code. It is
unlikely to survive for long.)
For now, the only thing one can configure is whether color is enabled or
not. This avoids needing to ask the `colored` crate whether colors have
been globally enabled or disabled. And, more crucially, avoids the need
to _set_ this global flag for testing diagnostic output. Doing so can
have unintended consequences, as outlined in #16115.
Fixes#16115
This essentially makes it impossible to construct a `Diagnostic`
that has a `TextRange` but no `File`.
This is meant to be a precursor to multi-span support.
(Note that I consider this more of a prototyping-change and not
necessarily what this is going to look like longer term.)
Reviewers can probably review this PR as one big diff instead of
commit-by-commit.
## Summary
Transition to using coarse-grained tracked structs (depends on
https://github.com/salsa-rs/salsa/pull/657). For now, this PR doesn't
add any `#[tracked]` fields, meaning that any changes cause the entire
struct to be invalidated. It also changes `AstNodeRef` to be
compared/hashed by pointer address, instead of performing a deep AST
comparison.
## Test Plan
This yields a 10-15% improvement on my machine (though weirdly some runs
were 5-10% without being flagged as inconsistent by criterion, is there
some non-determinism involved?). It's possible that some of this is
unrelated, I'll try applying the patch to the current salsa version to
make sure.
---------
Co-authored-by: Micha Reiser <micha@reiser.io>
## Summary
This PR adds support for user-level configurations
(`~/.config/knot/knot.toml`) to Red Knot.
Red Knot will watch the user-level configuration file for changes but
only if it exists
when the process start. It doesn't watch for new configurations,
mainly to simplify things for now (it would require watching the entire
`.config` directory because the `knot` subfolder might not exist
either).
The new `ConfigurationFile` struct seems a bit overkill for now but I
plan to use it for
hierarchical configurations as well.
Red Knot uses the same strategy as uv and Ruff by using the etcetera
crate.
## Test Plan
Added CLI and file watching test
## Summary
This PR adds a new `user_configuration_directory` method to `System`. We
need it to resolve where to lookup a user-level `knot.toml`
configuration file.
The method belongs to `System` because not all platforms have a
convention of where to store such configuration files (e.g. wasm).
I refactored `TestSystem` to be a simple wrapper around an `Arc<dyn
System...>` and use the `System.as_any` method instead to cast it down
to an `InMemory` system. I also removed some `System` specific methods
from `InMemoryFileSystem`, they don't belong there.
This PR removes the `os` feature as a default feature from `ruff_db`.
Most crates depending on `ruff_db` don't need it because they only
depend on `System` or only depend on `os` for testing. This was
necessary to fix a compile error with `red_knot_wasm`
## Test Plan
I'll make use of the method in my next PR. So I guess we won't know if
it works before then but I copied the code from Ruff/uv, so I have high
confidence that it is correct.
`cargo test`
## Summary
Adds a JSON schema generation step for Red Knot. This PR doesn't yet add
a publishing step because it's still a bit early for that
## Test plan
I tested the schema in Zed, VS Code and PyCharm:
* PyCharm: You have to manually add a schema mapping (settings JSON
Schema Mappings)
* Zed and VS code support the inline schema specification
```toml
#:schema /Users/micha/astral/ruff/knot.schema.json
[environment]
extra-paths = []
[rules]
call-possibly-unbound-method = "error"
unknown-rule = "error"
# duplicate-base = "error"
```
```json
{
"$schema": "file:///Users/micha/astral/ruff/knot.schema.json",
"environment": {
"python-version": "3.13",
"python-platform": "linux2"
},
"rules": {
"unknown-rule": "error"
}
}
```
https://github.com/user-attachments/assets/a18fcd96-7cbe-4110-985b-9f1935584411
The Schema overall works but all editors have their own quirks:
* PyCharm: Hovering a name always shows the section description instead
of the description of the specific setting. But it's the same for other
settings in `pyproject.toml` files 🤷
* VS Code (JSON): Using the generated schema in a JSON file gives
exactly the experience I want
* VS Code (TOML):
* Properties with multiple possible values are repeated during
auto-completion without giving any hint how they're different. 
* The property description mushes together the description of the
property and the value, which looks sort of ridiculous. 
* Autocompletion and documentation hovering works (except the
limitations mentioned above)
* Zed:
* Very similar to VS Code with the exception that it uses the
description attribute to distinguish settings with multiple possible
values 
I don't think there's much we can do here other than hope (or help)
editors improve their auto completion. The same short comings also apply
to ruff, so this isn't something new. For now, I think this is good
enough
I found it useful to have the `&dyn Diagnostic` trait impl
specifically. I added `Arc<dyn Diagnostic>` for completeness.
(I do kind of wonder if we should be preferring `Arc<dyn ...>`
over something like `Box<dyn ...>` more generally, especially
for things with immutable APIs. It would make cloning cheap.)
This change does a simple swap of the existing renderer for one that
uses our vendored copy of `annotate-snippets`. We don't change anything
about the diagnostic data model, but this alone already makes
diagnostics look a lot nicer!
More refinements to the panic messages for failing mdtests to mimic the
output of the default panic hook more closely:
- We now print out `Box<dyn Any>` if the panic payload is not a string
(which is typically the case for salsa panics).
- We now include the panic's backtrace if you set the `RUST_BACKTRACE`
environment variable.
This fixes#15317. Our `catch_unwind` wrapper installs a panic hook that
captures (the rendered contents of) the panic info when a panic occurs.
Since the intent is that the caller will render the panic info in some
custom way, the hook silences the default stderr panic output.
However, the panic hook is a global resource, so if any one thread was
in the middle of a `catch_unwind` call, we would silence the default
panic output for _all_ threads.
The solution is to also keep a thread local that indicates whether the
current thread is in the middle of our `catch_unwind`, and to fall back
on the default panic hook if not.
## Test Plan
Artificially added an mdtest parse error, ran tests via `cargo test -p
red_knot_python_semantic` to run a large number of tests in parallel.
Before this patch, the panic message was swallowed as reported in
#15317. After, the panic message was shown.
This updates the mdtest harness to catch any panics that occur during
type checking, and to display the panic message as an mdtest failure.
(We don't know which specific line causes the failure, so we attribute
panics to the first line of the test case.)
## Summary
Fixes https://github.com/astral-sh/ruff/issues/15027
The `MemoryFileSystem::write_file` API automatically creates
non-existing ancestor directoryes
but we failed to update the status of the now created ancestor
directories in the `Files` data structure.
## Test Plan
Tested that the case in https://github.com/astral-sh/ruff/issues/15027
now passes regardless of whether the *Simple* case is commented out or
not
## Summary
This PR extends the mdtest configuration with a `log` setting that can
be any of:
* `true`: Enables tracing
* `false`: Disables tracing (default)
* String: An ENV_FILTER similar to `RED_KNOT_LOG`
```toml
log = true
```
Closes https://github.com/astral-sh/ruff/issues/13865
## Test Plan
I changed a test and tried `log=true`, `log=false`, and `log=INFO`
## Summary
Fixes a small scoping issue in `DiagnosticId::matches`
Note: I don't think we should use `lint:id` in mdtests just yet. I worry
that it could lead to many unnecessary churns if we decide **not** to
use `lint:<id>` as the format (e.g., `lint/id`).
The reason why users even see `lint:<rule>` is because the mdtest
framework uses the diagnostic infrastructure
Closes#14910
## Test Plan
Added tests
## Summary
This PR introduces a structured `DiagnosticId` instead of using a plain
`&'static str`. It is the first of three in a stack that implements a
basic rules infrastructure for Red Knot.
`DiagnosticId` is an enum over all known diagnostic codes. A closed enum
reduces the risk of accidentally introducing two identical diagnostic
codes. It also opens the possibility of generating reference
documentation from the enum in the future (not part of this PR).
The enum isn't *fully closed* because it uses a `&'static str` for lint
names. This is because we want the flexibility to define lints in
different crates, and all names are only known in `red_knot_linter` or
above. Still, lower-level crates must already reference the lint names
to emit diagnostics. We could define all lint-names in `DiagnosticId`
but I decided against it because:
* We probably want to share the `DiagnosticId` type between Ruff and Red
Knot to avoid extra complexity in the diagnostic crate, and both tools
use different lint names.
* Lints require a lot of extra metadata beyond just the name. That's why
I think defining them close to their implementation is important.
In the long term, we may also want to support plugins, which would make
it impossible to know all lint names at compile time. The next PR in the
stack introduces extra syntax for defining lints.
A closed enum does have a few disadvantages:
* rustc can't help us detect unused diagnostic codes because the enum is
public
* Adding a new diagnostic in the workspace crate now requires changes to
at least two crates: It requires changing the workspace crate to add the
diagnostic and the `ruff_db` crate to define the diagnostic ID. I
consider this an acceptable trade. We may want to move `DiagnosticId` to
its own crate or into a shared `red_knot_diagnostic` crate.
## Preventing duplicate diagnostic identifiers
One goal of this PR is to make it harder to introduce ambiguous
diagnostic IDs, which is achieved by defining a closed enum. However,
the enum isn't fully "closed" because it doesn't explicitly list the IDs
for all lint rules. That leaves the possibility that a lint rule and a
diagnostic ID share the same name.
I made the names unambiguous in this PR by separating them into
different namespaces by using `lint/<rule>` for lint rule codes. I don't
mind the `lint` prefix in a *Ruff next* context, but it is a bit weird
for a standalone type checker. I'd like to not overfocus on this for now
because I see a few different options:
* We remove the `lint` prefix and add a unit test in a top-level crate
that iterates over all known lint rules and diagnostic IDs to ensure the
names are non-overlapping.
* We only render `[lint]` as the error code and add a note to the
diagnostic mentioning the lint rule. This is similar to clippy and has
the advantage that the header line remains short
(`lint/some-long-rule-name` is very long ;))
* Any other form of adjusting the diagnostic rendering to make the
distinction clear
I think we can defer this decision for now because the `DiagnosticId`
contains all the relevant information to change the rendering
accordingly.
## Why `Lint` and not `LintRule`
I see three kinds of diagnostics in Red Knot:
* Non-suppressable: Reveal type, IO errors, configuration errors, etc.
(any `DiagnosticId`)
* Lints: code-related diagnostics that are suppressable.
* Lint rules: The same as lints, but they can be enabled or disabled in
the configuration. The majority of lints in Red Knot and the Ruff
linter.
Our current implementation doesn't distinguish between lints and Lint
rules because we aren't aware of a suppressible code-related lint that
can't be configured in the configuration. The only lint that comes to my
mind is maybe `division-by-zero` if we're 99.99% sure that it is always
right. However, I want to keep the door open to making this distinction
in the future if it proves useful.
Another reason why I chose lint over lint rule (or just rule) is that I
want to leave room for a future lint rule and lint phase concept:
* lint is the *what*: a specific code smell, pattern, or violation
* the lint rule is the *how*: I could see a future `LintRule` trait in
`red_knot_python_linter` that provides the necessary hooks to run as
part of the linter. A lint rule produces diagnostics for exactly one
lint. A lint rule differs from all lints in `red_knot_python_semantic`
because they don't run as "rules" in the Ruff sense. Instead, they're a
side-product of type inference.
* the lint phase is a different form of *how*: A lint phase can produce
many different lints in a single pass. This is a somewhat common pattern
in Ruff where running one analysis collects the necessary information
for finding many different lints
* diagnostic is the *presentation*: Unlike a lint, the diagnostic isn't
the what, but how a specific lint gets presented. I expect that many
lints can use one generic `LintDiagnostic`, but a few lints might need
more flexibility and implement their custom diagnostic rendering (at
least custom `Diagnostic` implementation).
## Test Plan
`cargo test`
## Summary
- Add 383 files from `crates/ruff_python_parser/resources` to the test
corpus
- Add 1296 files from `crates/ruff_linter/resources` to the test corpus
- Use in-memory file system for tests
- Improve test isolation by cleaning the test environment between checks
- Add a mechanism for "known failures". Mark ~80 files as known
failures.
- The corpus test is now a lot slower (6 seconds).
Note:
While `red_knot` as a command line tool can run over all of these
files without panicking, we still have a lot of test failures caused by
explicitly "pulling" all types.
## Test Plan
Run `cargo test -p red_knot_workspace` while making sure that
- Introducing code that is known to lead to a panic fails the test
- Removing code that is known to lead to a panic from
`KNOWN_FAILURES`-files also fails the test
