-
Notifications
You must be signed in to change notification settings - Fork 12.5k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Improve apparent type of mapped types #57122
Conversation
@typescript-bot test top100 |
Heya @ahejlsberg, I've started to run the parallelized Definitely Typed test suite on this PR at 14724a0. You can monitor the build here. Update: The results are in! |
Heya @ahejlsberg, I've started to run the diff-based user code test suite on this PR at 14724a0. You can monitor the build here. Update: The results are in! |
Heya @ahejlsberg, I've started to run the diff-based top-repos suite on this PR at 14724a0. You can monitor the build here. Update: The results are in! |
Heya @ahejlsberg, I've started to run the faster perf test suite on this PR at 14724a0. You can monitor the build here. Update: The results are in! |
@ahejlsberg Here are the results of running the user test suite comparing There were infrastructure failures potentially unrelated to your change:
Otherwise... Something interesting changed - please have a look. Details
|
@ahejlsberg Here they are:
tscComparison Report - baseline..pr
System info unknown
Hosts
Scenarios
Developer Information: |
Hey @ahejlsberg, the results of running the DT tests are ready. |
@ahejlsberg Here are the results of running the top-repos suite comparing Something interesting changed - please have a look. Details
|
src/compiler/checker.ts
Outdated
const target = (type.target ?? type) as MappedType; | ||
const typeVariable = getHomomorphicTypeVariable(target); | ||
if (typeVariable && !target.declaration.nameType) { | ||
const constraint = getConstraintTypeFromMappedType(type); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
All of this makes sense now... it is tricky that mapped types don't have constraints resolvable by getBaseConstraintOfType
and such but we can reach for getConstraintTypeFromMappedType
and use that.
src/compiler/checker.ts
Outdated
const typeVariable = getHomomorphicTypeVariable(target); | ||
if (typeVariable && !target.declaration.nameType) { | ||
const constraint = getConstraintTypeFromMappedType(type); | ||
if (constraint.flags & TypeFlags.Index) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I wonder if this check shouldn't be replaced with an assert - it's quite a strong invariant that this branch can only be entered after getHomomorphicTypeVariable
returns true and that requires the constraint to be an Index
type.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
An assert here wouldn't be right. There many cases where an instantiation of a homomorphic mapped type doesn't have a keyof XXX
constraint anymore. For example, the homomorphic Readonly<T>
has the constraint type keyof T
, but the instantiation Readonly<{ a: string, b: string }>
has the constraint type "a" | "b"
. What we're looking for here are instantiations where the constraint type is still a keyof XXX
. That's what wasn't handled before.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thank you for the explanation - it makes sense, getHomomorphicTypeVariable
is called now with the target
and not the type
after all.
For what it's worth - I spent some sleepless nights on this last week so I built up some good context around this issue. LGTM (pending the resolution of #50034 ). |
@typescript-bot pack this |
Hey @gabritto, I've packed this into an installable tgz. You can install it for testing by referencing it in your
and then running There is also a playground for this build and an npm module you can use via |
|
||
type Curry<F extends ((...args: any) => any)> = | ||
<T extends any[]>(...args: Tools.Cast<Tools.Cast<T, Gaps<Parameters<F>>>, any[]>) => | ||
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
How is the change in the PR causing this error?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I believe the underlying issue is the same as it was in my PRs. This is now resolved through layers and thus getResolvedApparentTypeOfMappedType
sees the constraint of Parameters<F>
- that once gets computed as unknown[]
(by getDefaultConstraintOfConditionalType
when restrictive instantiation bails out from getConstraintOfDistributiveConditionalType
) and once as any
(by getConstraintOfDistributiveConditionalType
). So any
can get returned and getResolvedApparentTypeOfMappedType
can't recognize this as an array-like. So it ends up not d about it not being assignable to readonly any[]
(that is being checked to determine if the type can be used as rest).
IIRC the other case (when unknown[]
) is returned makes any[]
(added by this Cast
) eliminated from the substitution type since, at that point, it's determined that the baseType
satisfies the added constraint - so the baseType
gets returned.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
So the issue is mainly in the fact that previously the substitution type wasn't simplified at all here and that & any[]
in it (coming from its added constraint) was preserved and made the comparison against readonly any[]
OK.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
What do you mean by "this is now resolved through layers"?
How do we go from a call to getResolvedApparentTypeOfMappedType
to calls to getDefaultConstraintOfConditionalType
and getConstraintOfDistributiveConditionalType
?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please note that I'm partially speaking based on what I remember about debugging this last week - so some details might be slightly off.
Gaps
is a mapped type - so we enter getResolvedApparentTypeOfMappedType
. With this PR we recognize that an instantiation of a homomorphic mapped type is received and we end up checking its constraint and if it's an index type. It turns out it is because it's something like keyof Parameters<...>
. So we grab the .type
and compute its base constraint. So now we are computing the constraint of a conditional type (through getConstraintFromConditionalType
) and, as mentioned above, this once resolves through getDefaultConstraintOfConditionalType
(for the restrictive instantiation case) and once through getConstraintOfDistributiveConditionalType
(for a regular case :P).
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
With latest commit this error is gone.
Hey @gabritto, I've packed this into an installable tgz. You can install it for testing by referencing it in your
and then running There is also a playground for this build and an npm module you can use via |
Hey @ahejlsberg, the results of running the DT tests are ready. Branch only errors:Package: lodash
|
@ahejlsberg Here are the results of running the top-repos suite comparing Something interesting changed - please have a look. Details
|
The test runs reveal inconsistencies related to indexing arrays with numeric string literals (as in #50034): function bar<T extends string[], K extends number>() {
type T00 = string[]["0"]; // Ok
type T01 = string[]["0.0"]; // Error, as expected
type T02 = string[][K | "0"]; // Error, but shouldn't be
type T10 = T["0"]; // Error, but shouldn't be
type T11 = T["0.0"]; // Error, as expected
type T12 = T[K | "0"]; // Error, but shouldn't be
} The errors on |
@typescript-bot test top100 |
Heya @ahejlsberg, I've started to run the diff-based top-repos suite on this PR at 633bfc3. You can monitor the build here. Update: The results are in! |
Heya @ahejlsberg, I've started to run the parallelized Definitely Typed test suite on this PR at 633bfc3. You can monitor the build here. Update: The results are in! |
Heya @ahejlsberg, I've started to run the diff-based user code test suite on this PR at 633bfc3. You can monitor the build here. Update: The results are in! |
Heya @ahejlsberg, I've started to run the faster perf test suite on this PR at 633bfc3. You can monitor the build here. Update: The results are in! |
@ahejlsberg Here are the results of running the user test suite comparing There were infrastructure failures potentially unrelated to your change:
Otherwise... Something interesting changed - please have a look. Details
|
@ahejlsberg Here they are:
tscComparison Report - baseline..pr
System info unknown
Hosts
Scenarios
Developer Information: |
Hey @ahejlsberg, the results of running the DT tests are ready. |
function bar<T extends string[], K extends number>() { | ||
type T00 = string[]["0"]; | ||
type T01 = string[]["0.0"]; // Error | ||
type T02 = string[][K | "0"]; | ||
type T10 = T["0"]; | ||
type T11 = T["0.0"]; // Error | ||
type T12 = T[K | "0"]; | ||
} |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
nit: those should likely be in a different test file since they don't have anything in common with this "test title" (assignmentToAnyArrayRestParameters
)
@ahejlsberg Here are the results of running the top-repos suite comparing Everything looks good! |
Test and performance suites are all clean. This is ready to merge. |
This PR supersedes #56727 (which was reverted). The improvements in this PR don't exhibit the issue that caused us to revert #56727.
EDIT:
However, the PR does break one test because the improved resolution of apparent types for mapped types causes the issue reported in #29919 to surface in that test.I'm going to recommend we merge #50034, which fixes #29919 and thus the broken test in this PR.Fixes #29919.
Fixes #56726.