Add unmeasurable variance kind for marking types whose variance result is unreliable #30416

weswigham · 2019-03-14T23:40:56Z

Those types being mapped types which are generic on their keys - in these cases our variance results are inaccurate, but not consistently on the positive or negative side. This is simply because the rules for relating object types are more permissive then mapped types when it comes to values, but are less permissive when it come to modifiers and keys. Simply disabling variance probing on mapped types isn't sufficient - that just causes us to kick the can down the line to a type containing the generic mapped type, which then exhibits the same issue.

Fixes #29991

…t is unreliable

jack-williams · 2019-03-14T23:57:20Z

This also fixes #29698.

jack-williams · 2019-03-14T23:57:59Z

src/compiler/checker.ts

@@ -13014,7 +13025,9 @@ namespace ts {
                        getCombinedMappedTypeOptionality(source) <= getCombinedMappedTypeOptionality(target));
                if (modifiersRelated) {
                    let result: Ternary;
-                    if (result = isRelatedTo(getConstraintTypeFromMappedType(target), getConstraintTypeFromMappedType(source), reportErrors)) {


Is it possible to only report as unmeasurable when the mapped types have modifiers?

Isn't the Record example exactly the case of when there isn't a modifier?

Yep, adding the modifier check means that this will not fix #29698. But the two issues aren't exactly the same so I was wondering if there exists two distinct fixes. #29698 only really breaks because when using string as a key type.

This was the kind of (hack) I had in mind: https://pastebin.com/n431Puhu. I have no idea if this is even feasible, or correct. I thought I'd at least play around with something rather than solely firing off random questions.

weswigham · 2019-04-19T00:13:20Z

@typescript-bot pack this

typescript-bot · 2019-04-19T00:13:24Z

Heya @weswigham, I've started to run the tarball bundle task on this PR at 33b6e42. You can monitor the build here. It should now contribute to this PR's status checks.

sandersn · 2019-04-19T15:40:16Z

[Pasting my reply from #30973]

Fixes lodash — 485k → 85k types. (vs 83k on 3.3)
Fixes reactstrap 8.0.0, basically — 1538k → 591k types (vs 493k on 3.3)
So it fixes #30973

Reactstrap 8.0.1 does better than this fix does relative to 8.0.0 on master. In fact, 8.0.1 no longer benefits from this fix. On master, it creates 107k types, and on this branch it creates 108k types (80k on 3.3).

weswigham · 2019-04-24T21:44:20Z

@typescript-bot run dt & @typescript-bot user test this

typescript-bot · 2019-04-24T21:44:22Z

Heya @weswigham, I've started to run the parallelized Definitely Typed test suite on this PR at bbfb25f. You can monitor the build here. It should now contribute to this PR's status checks.

typescript-bot · 2019-04-24T21:44:22Z

Heya @weswigham, I've started to run the community code test suite on this PR at bbfb25f. You can monitor the build here. It should now contribute to this PR's status checks.

weswigham · 2019-04-24T22:00:02Z

@typescript-bot test this 😃

typescript-bot · 2019-04-24T22:00:07Z

Heya @weswigham, I've started to run the extended test suite on this PR at bbfb25f. You can monitor the build here. It should now contribute to this PR's status checks.

weswigham · 2019-04-24T23:13:06Z

@typescript-bot run dt so I can get a full error log on each shard thanks

typescript-bot · 2019-04-24T23:13:11Z

Heya @weswigham, I've started to run the parallelized Definitely Typed test suite on this PR at bbfb25f. You can monitor the build here. It should now contribute to this PR's status checks.

…index: string]: any}`

weswigham · 2019-04-25T20:40:18Z

@typescript-bot run dt now that I've patched Readonly<any> to be any instead of the absolutely-not-as-permissive-as-desired {readonly [index: string]: any}.

typescript-bot · 2019-04-25T20:40:21Z

Heya @weswigham, I've started to run the parallelized Definitely Typed test suite on this PR at 15f8c10. You can monitor the build here. It should now contribute to this PR's status checks.

…n some cases

weswigham · 2019-04-25T23:08:12Z

@typescript-bot run dt once more - I've amended Unmeasurable variances to still allow a non-structural out for identical types and comparisons with any. While the second isn't quite correct, in the cases where it's not, the structural errors are essentially useless in figuring out what nonlinear relationship made your any not be related enough to whatever else (spoiler: usually mapped type keys, since keyof any is just string | number | symbol which is just a string indexer, which isn't nearly as strong a type as the mapped type is being related in higher order as). Since going to structural checks there seems to break some people in DT, imma just defer that to when we fix the higher-order mapped type relationship to actually capture the nuance of index signature behavior.

typescript-bot · 2019-04-25T23:08:13Z

Heya @weswigham, I've started to run the parallelized Definitely Typed test suite on this PR at dc40484. You can monitor the build here. It should now contribute to this PR's status checks.

…ange, for now

weswigham · 2019-04-29T20:42:26Z

@typescript-bot run dt once more - per an offline request by @ahejlsberg I've walked back a bit of the change (and added a test to show what's still broken) to make what's here a bit easier to swallow so we can iterate on it.

typescript-bot · 2019-04-29T20:42:28Z

Heya @weswigham, I've started to run the parallelized Definitely Typed test suite on this PR at b646f28. You can monitor the build here. It should now contribute to this PR's status checks.

weswigham · 2019-04-29T23:39:41Z

@typescript-bot run dt again

typescript-bot · 2019-04-29T23:39:44Z

Heya @weswigham, I've started to run the parallelized Definitely Typed test suite on this PR at 18e5656. You can monitor the build here. It should now contribute to this PR's status checks.

ahejlsberg

Assuming we're good with the two DT changes, I think we should get this merged.

weswigham · 2019-05-03T09:44:26Z

Assuming we're good with the two DT changes, I think we should get this merged.

Yeah, the one is from current master and the other is a new circularity caused by some newly structural checking. (We lookup the type of a call expression, looking at the signature's return, the signature has a conditional on this in its return, which causes us to then compare that member's signature with the conditional extends member of the same name, which requires relating the same signature)

… comparisons with `this` types

weswigham · 2019-05-03T20:51:06Z

@typescript-bot run dt to see if my fix for the circular structural check error breaks anything. We don't have to ship it, since the error is kinda justifiable based on the comparisons we need to perform, but I feel like the method(): this extends SpecificSubtype ? A : B pattern is probably common enough in the wild to want it to not break, since it already worked provided the SpecificSubtype was generic (despite it only occurring once in DT).

typescript-bot · 2019-05-03T20:51:13Z

Heya @weswigham, I've started to run the parallelized Definitely Typed test suite on this PR at a7fc378. You can monitor the build here. It should now contribute to this PR's status checks.

weswigham · 2019-05-03T21:41:25Z

Oho, clean DT run. Noice.

weswigham · 2019-05-03T21:43:28Z

@ahejlsberg we can talk about 6a9e337 seperately if you like, but I've merged it with this PR (and merged this PR) for now, since it fixed the remaining regression on DT so this PR is squeaky clean.

Add unmeasurable variance kind for marking types whose variance resul…

029d3db

…t is unreliable

weswigham requested review from sandersn, ahejlsberg and RyanCavanaugh March 14, 2019 23:40

jack-williams reviewed Mar 14, 2019

View reviewed changes

weswigham added 2 commits April 18, 2019 17:03

Merge branch 'master' into add-unmeasurable-variance-kind

d945aa5

Remove now-unneeded nongeneric checks

33b6e42

weswigham mentioned this pull request Apr 19, 2019

@types/reactstrap@8.0.0 runs out of memory on TS 3.4 #30973

Closed

Merge branch 'master' into add-unmeasurable-variance-kind

bbfb25f

weswigham added 2 commits April 25, 2019 11:20

Merge branch 'master' into add-unmeasurable-variance-kind

10183d9

Add rule allowing Readonly<any> to be any instead of `{readonly […

15f8c10

…index: string]: any}`

All Unmeasurable variances to still shortcut structural comparisons i…

dc40484

…n some cases

weswigham added 2 commits April 29, 2019 11:21

Merge branch 'master' into add-unmeasurable-variance-kind

093cc72

Separate unmeasurable from unreliable to reduce the impact of this ch…

4e96d98

…ange, for now

weswigham force-pushed the add-unmeasurable-variance-kind branch from 5e624e8 to 4e96d98 Compare April 29, 2019 20:38

Fix lint

b646f28

Remove Readonly<any> -> any callout

18e5656

weswigham mentioned this pull request Apr 29, 2019

Trust variance checking during identity checking #30903

Closed

RyanCavanaugh assigned weswigham Apr 30, 2019

ahejlsberg approved these changes May 1, 2019

View reviewed changes

RyanCavanaugh approved these changes May 2, 2019

View reviewed changes

weswigham added 2 commits May 3, 2019 13:43

Add fix for circularity error triggered by deep signature return type…

6a9e337

… comparisons with `this` types

Merge branch 'master' into add-unmeasurable-variance-kind

a7fc378

weswigham force-pushed the add-unmeasurable-variance-kind branch from 6e0ee3a to a7fc378 Compare May 3, 2019 20:48

weswigham merged commit b365e65 into microsoft:master May 3, 2019

weswigham deleted the add-unmeasurable-variance-kind branch May 3, 2019 21:42

ahejlsberg mentioned this pull request Jul 13, 2022

Conditional type inside function resolves differently than at top-level #49852

Open

ahejlsberg mentioned this pull request Dec 17, 2023

Recursive types can depend on presence and source location of unused type declaration #55758

Open

weswigham mentioned this pull request Feb 15, 2024

Add checker speculation helper, use in overload resolution #57421

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add unmeasurable variance kind for marking types whose variance result is unreliable #30416

Add unmeasurable variance kind for marking types whose variance result is unreliable #30416

weswigham commented Mar 14, 2019

jack-williams commented Mar 14, 2019

jack-williams Mar 14, 2019

weswigham Mar 15, 2019

jack-williams Mar 15, 2019

jack-williams Mar 15, 2019

weswigham commented Apr 19, 2019

typescript-bot commented Apr 19, 2019 •

edited

Loading

sandersn commented Apr 19, 2019 •

edited

Loading

weswigham commented Apr 24, 2019 •

edited

Loading

typescript-bot commented Apr 24, 2019 •

edited

Loading

typescript-bot commented Apr 24, 2019 •

edited

Loading

weswigham commented Apr 24, 2019

typescript-bot commented Apr 24, 2019 •

edited

Loading

weswigham commented Apr 24, 2019

typescript-bot commented Apr 24, 2019 •

edited

Loading

weswigham commented Apr 25, 2019

typescript-bot commented Apr 25, 2019 •

edited

Loading

weswigham commented Apr 25, 2019

typescript-bot commented Apr 25, 2019 •

edited

Loading

weswigham commented Apr 29, 2019

typescript-bot commented Apr 29, 2019 •

edited

Loading

weswigham commented Apr 29, 2019

typescript-bot commented Apr 29, 2019 •

edited

Loading

ahejlsberg left a comment

weswigham commented May 3, 2019 •

edited

Loading

weswigham commented May 3, 2019

typescript-bot commented May 3, 2019 •

edited

Loading

weswigham commented May 3, 2019

weswigham commented May 3, 2019

Add unmeasurable variance kind for marking types whose variance result is unreliable #30416

Add unmeasurable variance kind for marking types whose variance result is unreliable #30416

Conversation

weswigham commented Mar 14, 2019

jack-williams commented Mar 14, 2019

jack-williams Mar 14, 2019

Choose a reason for hiding this comment

weswigham Mar 15, 2019

Choose a reason for hiding this comment

jack-williams Mar 15, 2019

Choose a reason for hiding this comment

jack-williams Mar 15, 2019

Choose a reason for hiding this comment

weswigham commented Apr 19, 2019

typescript-bot commented Apr 19, 2019 • edited Loading

sandersn commented Apr 19, 2019 • edited Loading

weswigham commented Apr 24, 2019 • edited Loading

typescript-bot commented Apr 24, 2019 • edited Loading

typescript-bot commented Apr 24, 2019 • edited Loading

weswigham commented Apr 24, 2019

typescript-bot commented Apr 24, 2019 • edited Loading

weswigham commented Apr 24, 2019

typescript-bot commented Apr 24, 2019 • edited Loading

weswigham commented Apr 25, 2019

typescript-bot commented Apr 25, 2019 • edited Loading

weswigham commented Apr 25, 2019

typescript-bot commented Apr 25, 2019 • edited Loading

weswigham commented Apr 29, 2019

typescript-bot commented Apr 29, 2019 • edited Loading

weswigham commented Apr 29, 2019

typescript-bot commented Apr 29, 2019 • edited Loading

ahejlsberg left a comment

Choose a reason for hiding this comment

weswigham commented May 3, 2019 • edited Loading

weswigham commented May 3, 2019

typescript-bot commented May 3, 2019 • edited Loading

weswigham commented May 3, 2019

weswigham commented May 3, 2019

typescript-bot commented Apr 19, 2019 •

edited

Loading

sandersn commented Apr 19, 2019 •

edited

Loading

weswigham commented Apr 24, 2019 •

edited

Loading

typescript-bot commented Apr 24, 2019 •

edited

Loading

typescript-bot commented Apr 24, 2019 •

edited

Loading

typescript-bot commented Apr 24, 2019 •

edited

Loading

typescript-bot commented Apr 24, 2019 •

edited

Loading

typescript-bot commented Apr 25, 2019 •

edited

Loading

typescript-bot commented Apr 25, 2019 •

edited

Loading

typescript-bot commented Apr 29, 2019 •

edited

Loading

typescript-bot commented Apr 29, 2019 •

edited

Loading

weswigham commented May 3, 2019 •

edited

Loading

typescript-bot commented May 3, 2019 •

edited

Loading