## Summary
Use `Type::Divergent` to avoid "too many iterations" panic on an
infinitely-nested tuple in an implicit instance attribute.
The regression here is from checking all tuple elements to see if they
contain a Divergent type. It's 5% on one project, 1% on another, and
zero on the rest. I spent some time looking into eliminating this
regression by tracking a flag on inference results to note if they could
possibly contain any Divergent type, but this doesn't really work --
there are too many different ways a type containing a Divergent type
could enter an inference result. Still thinking about whether there are
other ways to reduce this. One option is if we see certain kinds of
non-atomic types that are commonly expensive to check for Divergent, we
could make `has_divergent_type` a Salsa query on those types.
## Test Plan
Added mdtest.
Co-authored-by: Alex Waygood <Alex.Waygood@Gmail.com>
## Summary
This PR adds a new `Type::TypedDict` variant. Before this PR, we treated
`TypedDict`-based types as dynamic Todo-types, and I originally planned
to make this change a no-op. And we do in fact still treat that new
variant similar to a dynamic type when it comes to type properties such
as assignability and subtyping. But then I somehow tricked myself into
implementing some of the things correctly, so here we are. The two main
behavioral changes are: (1) we now also detect generic `TypedDict`s,
which removes a few false positives in the ecosystem, and (2) we now
support *attribute* access (not key-based indexing!) on these types,
i.e. we infer proper types for something like
`MyTypedDict.__required_keys__`. Nothing exciting yet, but gets the
infrastructure into place.
Note that with this PR, the type of (the type) `MyTypedDict` itself is
still represented as a `Type::ClassLiteral` or `Type::GenericAlias` (in
case `MyTypedDict` is generic). Only inhabitants of `MyTypedDict`
(instances of `dict` at runtime) are represented by `Type::TypedDict`.
We may want to revisit this decision in the future, if this turns out to
be too error-prone. Right now, we need to use `.is_typed_dict(db)` in
all the right places to distinguish between actual (generic) classes and
`TypedDict`s. But so far, it seemed unnecessary to add additional `Type`
variants for these as well.
part of https://github.com/astral-sh/ty/issues/154
## Ecosystem impact
The new diagnostics on `cloud-init` look like true positives to me.
## Test Plan
Updated and new Markdown tests
## Summary
This PR updates Salsa to pull in Ibraheem's multithreading improvements (https://github.com/salsa-rs/salsa/pull/921).
## Performance
A small regression for single-threaded benchmarks is expected because
papaya is slightly slower than a `Mutex<FxHashMap>` in the uncontested
case (~10%). However, this shouldn't matter as much in practice because:
1. Salsa has a fast-path when only using 1 DB instance which is the
common case in production. This fast-path is not impacted by the changes
but we measure the slow paths in our benchmarks (because we use multiple
db instances)
2. Fixing the 10x slowdown for the congested case (multi threading)
outweights the downsides of a 10% perf regression for single threaded
use cases, especially considering that ty is heavily multi threaded.
## Test Plan
`cargo test`
## 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.
This causes spurious query cycles.
This PR also includes an update to Salsa, which gives us db events on
cycle iteration, so we can write tests asserting the absence of a cycle.
## Summary
Update Salsa to pull in https://github.com/salsa-rs/salsa/pull/788 which
fixes the, by now, famous *access to field whilst the value is being
initialized*.
This PR also re-enables all tests that previously triggered the panic.
## Test Plan
`cargo test`
Update to latest Salsa main branch, so as to get a baseline for
measuring the perf effect of https://github.com/salsa-rs/salsa/pull/786
on red-knot in isolation from other recent changes in Salsa main branch.
## 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`
Pulls in the latest Salsa main branch, which supports fixpoint
iteration, and uses it to handle all query cycles.
With this, we no longer need to skip any corpus files to avoid panics.
Latest perf results show a 6% incremental and 1% cold-check regression.
This is not a "no cycles" regression, as tomllib and typeshed do trigger
some definition cycles (previously handled by our old
`infer_definition_types` fallback to `Unknown`). We don't currently have
a benchmark we can use to measure the pure no-cycles regression, though
I expect there would still be some regression; the fixpoint iteration
feature in Salsa does add some overhead even for non-cyclic queries.
I think this regression is within the reasonable range for this feature.
We can do further optimization work later, but I don't think it's the
top priority right now. So going ahead and acknowledging the regression
on CodSpeed.
Mypy primer is happy, so this doesn't regress anything on our
currently-checked projects. I expect it probably unlocks adding a number
of new projects to our ecosystem check that previously would have
panicked.
Fixes#13792Fixes#14672
Update to latest Salsa main branch. This provides a point of comparison
for the perf impact of fixpoint iteration, which is based on latest
Salsa main.
This requires an update to the locked version of our boxcar dep, since
Salsa now depends on a newer version of boxcar.
## 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>
Jack O'Connor
renovate[bot]
Carl Meyer
Micha Reiser
Shunsuke Shibayama
Ibraheem Ahmed
David Peter
Alex Waygood