## Summary
This PR is a collaboration with @AlexWaygood from our pairing session
last Friday.
The main goal here is removing `ruff_linter::message::OldDiagnostic` in
favor of
using `ruff_db::diagnostic::Diagnostic` directly. This involved a few
major steps:
- Transferring the fields
- Transferring the methods and trait implementations, where possible
- Converting some constructor methods to free functions
- Moving the `SecondaryCode` struct
- Updating the method names
I'm hoping that some of the methods, especially those in the
`expect_ruff_*`
family, won't be necessary long-term, but I avoided trying to replace
them
entirely for now to keep the already-large diff a bit smaller.
### Related refactors
Alex and I noticed a few refactoring opportunities while looking at the
code,
specifically the very similar implementations for
`create_parse_diagnostic`,
`create_unsupported_syntax_diagnostic`, and
`create_semantic_syntax_diagnostic`.
We combined these into a single generic function, which I then copied
into
`ruff_linter::message` with some small changes and a TODO to combine
them in the
future.
I also deleted the `DisplayParseErrorType` and `TruncateAtNewline` types
for
reporting parse errors. These were added in #4124, I believe to work
around the
error messages from LALRPOP. Removing these didn't affect any tests, so
I think
they were unnecessary now that we fully control the error messages from
the
parser.
On a more minor note, I factored out some calls to the
`OldDiagnostic::filename`
(now `Diagnostic::expect_ruff_filename`) function to avoid repeatedly
allocating
`String`s in some places.
### Snapshot changes
The `show_statistics_syntax_errors` integration test changed because the
`OldDiagnostic::name` method used `syntax-error` instead of
`invalid-syntax`
like in ty. I think this (`--statistics`) is one of the only places we
actually
use this name for syntax errors, so I hope this is okay. An alternative
is to
use `syntax-error` in ty too.
The other snapshot changes are from removing this code, as discussed on
[Discord](https://discord.com/channels/1039017663004942429/1228460843033821285/1388252408848847069):
34052a1185/crates/ruff_linter/src/message/mod.rs (L128-L135)
I think both of these are technically breaking changes, but they only
affect
syntax errors and are very narrow in scope, while also pretty
substantially
simplifying the refactor, so I hope they're okay to include in a patch
release.
## Test plan
Existing tests, with the adjustments mentioned above
---------
Co-authored-by: Alex Waygood <Alex.Waygood@Gmail.com>
## Summary
This PR adds initial support for workspace diagnostics in the ty server.
Reference spec:
https://microsoft.github.io/language-server-protocol/specifications/lsp/3.17/specification/#workspace_diagnostic
This is currently implemented via the **pull diagnostics method** which
was added in the current version (3.17) and the server advertises it via
the `diagnosticProvider.workspaceDiagnostics` server capability.
**Note:** This might be a bit confusing but a workspace diagnostics is
not for a single workspace but for all the workspaces that the server
handles. These are the ones that the server received during
initialization. Currently, the ty server doesn't support multiple
workspaces so this capability is also limited to provide diagnostics
only for a single workspace (the first one if the client provided
multiple).
A new `ty.diagnosticMode` server setting is added which can be either
`workspace` (for workspace diagnostics) or `openFilesOnly` (for checking
only open files) (default). This is same as
`python.analysis.diagnosticMode` that Pyright / Pylance utilizes. In the
future, we could use the value under `python.*` namespace as fallback to
improve the experience on user side to avoid setting the value multiple
times.
Part of: astral-sh/ty#81
## Test Plan
This capability was introduced in the current LSP version (~3 years) and
the way it's implemented by various clients are a bit different. I've
provided notes on what I've noticed and what would need to be done on
our side to further improve the experience.
### VS Code
VS Code sends the `workspace/diagnostic` requests every ~2 second:
```
[Trace - 12:12:32 PM] Sending request 'workspace/diagnostic - (403)'.
[Trace - 12:12:32 PM] Received response 'workspace/diagnostic - (403)' in 2ms.
[Trace - 12:12:34 PM] Sending request 'workspace/diagnostic - (404)'.
[Trace - 12:12:34 PM] Received response 'workspace/diagnostic - (404)' in 2ms.
[Trace - 12:12:36 PM] Sending request 'workspace/diagnostic - (405)'.
[Trace - 12:12:36 PM] Received response 'workspace/diagnostic - (405)' in 2ms.
[Trace - 12:12:38 PM] Sending request 'workspace/diagnostic - (406)'.
[Trace - 12:12:38 PM] Received response 'workspace/diagnostic - (406)' in 3ms.
[Trace - 12:12:40 PM] Sending request 'workspace/diagnostic - (407)'.
[Trace - 12:12:40 PM] Received response 'workspace/diagnostic - (407)' in 2ms.
...
```
I couldn't really find any resource that explains this behavior. But,
this does mean that we'd need to implement the caching layer via the
previous result ids sooner. This will allow the server to avoid sending
all the diagnostics on every request and instead just send a response
stating that the diagnostics hasn't changed yet. This could possibly be
achieved by using the salsa ID.
If we switch from workspace diagnostics to open-files diagnostics, the
server would send the diagnostics only via the `textDocument/diagnostic`
endpoint. Here, when a document containing the diagnostic is closed, the
server would send a publish diagnostics notification with an empty list
of diagnostics to clear the diagnostics from that document. The issue is
the VS Code doesn't seem to be clearing the diagnostics in this case
even though it receives the notification. (I'm going to open an issue on
VS Code side for this today.)
https://github.com/user-attachments/assets/b0c0833d-386c-49f5-8a15-0ac9133e15ed
### Zed
Zed's implementation works by refreshing the workspace diagnostics
whenever the content of the documents are changed. This seems like a
very reasonable behavior and I was a bit surprised that VS Code didn't
use this heuristic.
https://github.com/user-attachments/assets/71c7b546-7970-434a-9ba0-4fa620647f6c
### Neovim
Neovim only recently added support for workspace diagnostics
(https://github.com/neovim/neovim/pull/34262, merged ~3 weeks ago) so
it's only available on nightly versions.
The initial support is limited and requires fetching the workspace
diagnostics manually as demonstrated in the video. It doesn't support
refreshing the workspace diagnostics either, so that would need to be
done manually as well. I'm assuming that these are just a temporary
limitation and will be implemented before the stable release.
https://github.com/user-attachments/assets/25b4a0e5-9833-4877-88ad-279904fffaf9
## Summary
This PR makes the necessary changes to the server that it can request
configurations from the client using the `configuration` request.
This PR doesn't make use of the request yet. It only sets up the
foundation (mainly the coordination between client and server)
so that future PRs could pull specific settings.
I plan to use this for pulling the Python environment from the Python
extension.
Deno does something very similar to this.
## Test Plan
Tested that diagnostics are still shown.
## Summary
Print the [new salsa memory usage
dumps](https://github.com/astral-sh/ruff/pull/18928) in mypy primer CI
runs to help us catch memory regressions. The numbers are rounded to the
nearest power of 1.1 (about a 5% threshold between buckets) to avoid overly sensitive diffs.
## Summary
Setting `TY_MEMORY_REPORT=full` will generate and print a memory usage
report to the CLI after a `ty check` run:
```
=======SALSA STRUCTS=======
`Definition` metadata=7.24MB fields=17.38MB count=181062
`Expression` metadata=4.45MB fields=5.94MB count=92804
`member_lookup_with_policy_::interned_arguments` metadata=1.97MB fields=2.25MB count=35176
...
=======SALSA QUERIES=======
`File -> ty_python_semantic::semantic_index::SemanticIndex`
metadata=11.46MB fields=88.86MB count=1638
`Definition -> ty_python_semantic::types::infer::TypeInference`
metadata=24.52MB fields=86.68MB count=146018
`File -> ruff_db::parsed::ParsedModule`
metadata=0.12MB fields=69.06MB count=1642
...
=======SALSA SUMMARY=======
TOTAL MEMORY USAGE: 577.61MB
struct metadata = 29.00MB
struct fields = 35.68MB
memo metadata = 103.87MB
memo fields = 409.06MB
```
Eventually, we should integrate these numbers into CI in some form. The
one limitation currently is that heap allocations in salsa structs (e.g.
interned values) are not tracked, but memoized values should have full
coverage. We may also want a peak memory usage counter (that accounts
for non-salsa memory), but that is relatively simple to profile manually
(e.g. `time -v ty check`) and would require a compile-time option to
avoid runtime overhead.
## Summary
Garbage collect ASTs once we are done checking a given file. Queries
with a cross-file dependency on the AST will reparse the file on demand.
This reduces ty's peak memory usage by ~20-30%.
The primary change of this PR is adding a `node_index` field to every
AST node, that is assigned by the parser. `ParsedModule` can use this to
create a flat index of AST nodes any time the file is parsed (or
reparsed). This allows `AstNodeRef` to simply index into the current
instance of the `ParsedModule`, instead of storing a pointer directly.
The indices are somewhat hackily (using an atomic integer) assigned by
the `parsed_module` query instead of by the parser directly. Assigning
the indices in source-order in the (recursive) parser turns out to be
difficult, and collecting the nodes during semantic indexing is
impossible as `SemanticIndex` does not hold onto a specific
`ParsedModuleRef`, which the pointers in the flat AST are tied to. This
means that we have to do an extra AST traversal to assign and collect
the nodes into a flat index, but the small performance impact (~3% on
cold runs) seems worth it for the memory savings.
Part of https://github.com/astral-sh/ty/issues/214.
## Summary
https://github.com/astral-sh/ty/issues/214 will require a couple
invasive changes that I would like to get merged even before garbage
collection is fully implemented (to avoid rebasing):
- `ParsedModule` can no longer be dereferenced directly. Instead you
need to load a `ParsedModuleRef` to access the AST, which requires a
reference to the salsa database (as it may require re-parsing the AST if
it was collected).
- `AstNodeRef` can only be dereferenced with the `node` method, which
takes a reference to the `ParsedModuleRef`. This allows us to encode the
fact that ASTs do not live as long as the database and may be collected
as soon a given instance of a `ParsedModuleRef` is dropped. There are a
number of places where we currently merge the `'db` and `'ast`
lifetimes, so this requires giving some types/functions two separate
lifetime parameters.
## Summary
Fixes https://github.com/astral-sh/ty/issues/556.
On Windows, system installations have different layouts to virtual
environments. In Windows virtual environments, the Python executable is
found at `<sys.prefix>/Scripts/python.exe`. But in Windows system
installations, the Python executable is found at
`<sys.prefix>/python.exe`. That means that Windows users were able to
point to Python executables inside virtual environments with the
`--python` flag, but they weren't able to point to Python executables
inside system installations.
This PR fixes that issue. It also makes a couple of other changes:
- Nearly all `sys.prefix` resolution is moved inside `site_packages.rs`.
That was the original design of the `site-packages` resolution logic,
but features implemented since the initial implementation have added
some resolution and validation to `resolver.rs` inside the module
resolver. That means that we've ended up with a somewhat confusing code
structure and a situation where several checks are unnecessarily
duplicated between the two modules.
- I noticed that we had quite bad error messages if you e.g. pointed to
a path that didn't exist on disk with `--python` (we just gave a
somewhat impenetrable message saying that we "failed to canonicalize"
the path). I improved the error messages here and added CLI tests for
`--python` and the `environment.python` configuration setting.
## Test Plan
- Existing tests pass
- Added new CLI tests
- I manually checked that virtual-environment discovery still works if
no configuration is given
- Micha did some manual testing to check that pointing `--python` to a
system-installation executable now works on Windows
## Summary
This PR partially solves https://github.com/astral-sh/ty/issues/164
(derived from #17643).
Currently, the definitions we manage are limited to those for simple
name (symbol) targets, but we expand this to track definitions for
attribute and subscript targets as well.
This was originally planned as part of the work in #17643, but the
changes are significant, so I made it a separate PR.
After merging this PR, I will reflect this changes in #17643.
There is still some incomplete work remaining, but the basic features
have been implemented, so I am publishing it as a draft PR.
Here is the TODO list (there may be more to come):
* [x] Complete rewrite and refactoring of documentation (removing
`Symbol` and replacing it with `Place`)
* [x] More thorough testing
* [x] Consolidation of duplicated code (maybe we can consolidate the
handling related to name, attribute, and subscript)
This PR replaces the current `Symbol` API with the `Place` API, which is
a concept that includes attributes and subscripts (the term is borrowed
from Rust).
## Test Plan
`mdtest/narrow/assignment.md` is added.
---------
Co-authored-by: David Peter <sharkdp@users.noreply.github.com>
Co-authored-by: Carl Meyer <carl@astral.sh>
## Summary
fixesastral-sh/ty#366
## Test Plan
* Added panic corpus regression tests
* I also wrote a hover regression test (see below), but decided not to
include it. The corpus tests are much more "effective" at finding these
types of errors, since they exhaustively check all expressions for
types.
<details>
```rs
#[test]
fn hover_regression_test_366() {
let test = cursor_test(
r#"
from ty_extensions import Intersection
class A: ...
class B: ...
def _(x: Intersection[A,<CURSOR> B]):
pass
"#,
);
assert_snapshot!(test.hover(), @r"
A & B
---------------------------------------------
```text
A & B
```
---------------------------------------------
info[hover]: Hovered content is
--> main.py:7:31
|
5 | class B: ...
6 |
7 | def _(x: Intersection[A, B]):
| ^^-^
| | |
| | Cursor offset
| source
8 | pass
|
");
}
```
</details>
## Summary
Support direct uses of `typing.TypeAliasType`, as in:
```py
from typing import TypeAliasType
IntOrStr = TypeAliasType("IntOrStr", int | str)
def f(x: IntOrStr) -> None:
reveal_type(x) # revealed: int | str
```
closes https://github.com/astral-sh/ty/issues/392
## Ecosystem
The new false positive here:
```diff
+ error[invalid-type-form] altair/utils/core.py:49:53: The first argument to `Callable` must be either a list of types, ParamSpec, Concatenate, or `...`
```
comes from the fact that we infer the second argument as a type
expression now. We silence false positives for PEP695 `ParamSpec`s, but
not for `P = ParamSpec("P")` inside `Callable[P, ...]`.
## Test Plan
New Markdown tests
Dhruv Manilawala
Ibraheem Ahmed
Brent Westbrook
Micha Reiser
Alex Waygood
DetachHead
Shunsuke Shibayama
Zanie Blue
justin
David Peter
Jo
lipefree
Adam Aaronson
Charlie Marsh