[TypeChecker] Implement Double <-> CGFloat implicit conversion #34401

xedin · 2020-10-23T00:53:07Z

Implement an implicit conversion between Double and CGFloat types via
an implicit initializer call.

Resolves: rdar://problem/70592377

xedin · 2020-10-23T01:02:51Z

@swift-ci please test source compatibility

xedin · 2020-10-23T07:10:18Z

Debug build was a UPASS

include/swift/AST/Types.h

test/Misc/misc_diagnostics.swift

xwu · 2020-10-26T19:54:13Z

This looks fantastic in terms of improving the usability of CGFloat.

To the extent that it touches upon the Swift language itself, and in particular type relationships and one of the original tentpole design decisions not to have implicit numeric conversions, I hope we'll have a chance to discuss and review this through the Swift Evolution process.

If presented as a sort of built-in ABI-stable retroactive typealias-equivalent, it would be pretty much a no-brainer, but if presented as an "implicit conversion," what's implied (to some, at least) would be a much more radical departure from Swift's design.

xedin · 2020-10-26T20:00:38Z

Thanks, @xwu! At this point I'm mostly trying to figure out whether it's feasible and what the performance impact is going to be. This would definitely be pitched and discussed on Swift Evolution if this experiment is a success.

xedin · 2020-10-27T20:31:07Z

@swift-ci please build toolchain macOS

xedin · 2020-10-28T03:06:57Z

@swift-ci please build toolchain macOS

swift-ci · 2020-10-28T08:25:05Z

macOS Toolchain
Download Toolchain
Git Sha - c47feeacc3a5408c63ab4fc39dc9d50d00fccd42

Install command
tar -zxf swift-PR-34401-738-osx.tar.gz --directory ~/

xedin · 2020-10-31T07:35:35Z

@swift-ci please build toolchain

swift-ci · 2020-10-31T08:54:35Z

Linux Toolchain (Ubuntu 16.04)
Download Toolchain
Git Sha - 2573c7d6dc075fa12e3bd349745efcb3c19dce4e

Install command
tar zxf swift-PR-34401-477-ubuntu16.04.tar.gz
More info

swift-ci · 2020-10-31T10:59:34Z

macOS Toolchain
Download Toolchain
Git Sha - 2573c7d6dc075fa12e3bd349745efcb3c19dce4e

Install command
tar -zxf swift-PR-34401-745-osx.tar.gz --directory ~/

xedin · 2020-11-12T01:13:29Z

@swift-ci please test source compatibility

xedin · 2020-11-14T06:19:28Z

@swift-ci please test source compatibility

xedin · 2020-11-16T04:47:09Z

@swift-ci please test source compatibility

xedin · 2020-11-17T00:38:18Z

@swift-ci please test source compatibility

xedin · 2020-11-17T06:59:28Z

@swift-ci please build toolchain

swift-ci · 2020-11-17T10:28:06Z

macOS Toolchain
Download Toolchain
Git Sha - b54012d38118cb84e3a496ccf39794d6b0eae3d3

Install command
tar -zxf swift-PR-34401-774-osx.tar.gz --directory ~/

xedin · 2021-01-07T23:17:27Z

@swift-ci please test source compatibility

…w a `SetVector`

This makes it possible to pick `CGFloat` function/operator overloads even in the presence of literals, otherwise non-default literal score gets in a way and solver ends up producing a lot more solutions with implicit Double<->CGFloat conversion than it can disambiguate, so it's better to just preserve the old behavior and pick `CGFloat` concrete overloads when appropriate.

… arguments If argument is a floating-point literal, with newly introduced implicit Double<->CGFloat conversion, sometimes it's better to choose a concrete function/operator overload on `CGFloat` even if it's not a default literal type e.g. `let _: CGFloat = min(1.0, log(<CGFloat value>))` shouldn't form solutions with `Double` arguments since it would result in multiple implicit conversions, it's better to use a `CGFloat` type for the arguments.

…ce of Double<->CGFloat conversion Not all of the unary operators have `CGFloat` overloads, so in order to preserve previous behavior (and overall best solution) with implicit Double<->CGFloat conversion we need to allow attempting generic operators for such cases. ```swift let _: CGFloat = -.pi / 2 ``` `-` doesn't have `CGFloat` overload (which might be an oversight), so in order to preserve type-checking behavior solver can't be allowed to pick `-(Double) -> Double` based on overload of `/`, the best possible solution would be with `/` as `(CGFloat, CGFloat) -> CGFloat` and `-` on a `FloatingPoint` protocol.

…rsion

…gs` API change

…urrently unsupported on Windows)

…al type This wasn't a problem before since locator wasn't really used by `ExprRewritter:coerceToType` but with Double<->CGFloat conversion it needs the locator to be anchored at a rewritten expression instead of the original one to form a correct implicit initializer call.

…s (argument or result)

…t conversions Change the conversion rule to favor any number of widenings (CGFloat -> Double) over even a single narrowing conversion and if there is no way to avoid narrowing (due to contextual requirements) attempt it as late as possible (the deeper in the AST that conversion is located the higher its score). This is a generally better rule when it comes to rounding and information loss that would result from narrowing conversions.

… narrowing conversions (Double -> CGFloat)

…onversion Doing so resulted in performance impact due to the number of partial solutions increase.

Just like generic overloads, `shrink` should always avoid any solutions with implicit conversions. Reducing disjunction domains becaused on solutions with implicit conversions could have negative performance impact due to the increase in total number of solutions that have to be examined as a result.

xedin · 2021-03-17T07:21:20Z

@swift-ci please test source compatibility

xwu

I'm excited about this! Just some nits...

include/swift/AST/Types.h

include/swift/Basic/LangOptions.h

include/swift/Sema/Constraint.h

include/swift/Sema/ConstraintSystem.h

lib/Sema/CSGen.cpp

lib/Sema/CSSimplify.cpp

lib/Sema/CSStep.cpp

Thanks Xiaodi Wu!

xedin · 2021-03-29T17:47:13Z

@swift-ci please test source compatibility

xedin · 2021-03-29T17:47:21Z

@swift-ci please test

xedin · 2021-03-29T21:24:10Z

Windows failure:

Failed Tests (1):
  Swift(windows-x86_64) :: Concurrency/Runtime/async_task_detached.swift

xedin · 2021-03-30T22:40:23Z

@swift-ci please test source compatibility release

xedin · 2021-03-31T17:20:24Z

Release suite failed with 05:38:56 Caused: hudson.remoting.ChannelClosedException: Channel "hudson.remoting.Channel@79e5e9aa:macOS-21": Remote call on macOS-21 failed. The channel is closing down or has closed down. Since SE-0307 got accepted I'm going to land this PR.

swift-ci · 2021-03-31T19:52:05Z

Build failed
Swift Test Linux Platform
Git Sha - a559ca8

swift-ci · 2021-03-31T20:25:49Z

Build failed
Swift Test OS X Platform
Git Sha - a559ca8

xedin requested a review from hborla October 23, 2020 00:59

benrimmington reviewed Oct 23, 2020

View reviewed changes

include/swift/AST/Types.h Outdated Show resolved Hide resolved

xwu reviewed Oct 24, 2020

View reviewed changes

test/Misc/misc_diagnostics.swift Outdated Show resolved Hide resolved

xedin force-pushed the implicit-cgfloat-conversion branch from ee8542a to 51aa211 Compare October 26, 2020 19:07

xedin force-pushed the implicit-cgfloat-conversion branch from 51aa211 to c47feea Compare October 28, 2020 03:06

xedin added the swift evolution pending discussion Flag → feature: A feature that has a Swift evolution proposal currently in review label Oct 30, 2020

xedin force-pushed the implicit-cgfloat-conversion branch from c47feea to 2573c7d Compare October 31, 2020 07:35

xedin force-pushed the implicit-cgfloat-conversion branch from 2573c7d to bbfb1d9 Compare November 12, 2020 01:13

xedin force-pushed the implicit-cgfloat-conversion branch from bbfb1d9 to 0d5de0e Compare November 14, 2020 06:18

xedin force-pushed the implicit-cgfloat-conversion branch 2 times, most recently from 0fddd92 to b54012d Compare November 17, 2020 00:38

xedin force-pushed the implicit-cgfloat-conversion branch from b54012d to ad047cd Compare January 7, 2021 17:41

xedin force-pushed the implicit-cgfloat-conversion branch from a556030 to ffc3e4f Compare January 16, 2021 00:48

xedin added 15 commits March 17, 2021 00:18

[CSBindings] Adjust Double<->CGFloat handling since Bindings are no…

ba2bf82

…w a `SetVector`

[TypeChecker] NFC: Add more tests for implicit Double<->CGFloat conve…

826c145

…rsion

[TypeChecker] Add a flag to disable Double<->CGFloat implicit conversion

3c731d8

[CSBindings] Adopt Double <-> CGFloat handling to `addPotentialBindin…

0321425

…gs` API change

[AST] NFC: Fix a typo in TypeBase::isDoubleType comment

4a053b6

[CSSimplify] Avoid using designated initializer for ArgumentInfo (c…

4f5a8d4

…urrently unsupported on Windows)

[CSGen] Favor overload choices that support Double/CGFloat conversion…

0567900

…s (argument or result)

[CSGen] Tweak favoring hacks to avoid favoring overloads that require…

dbf401e

… narrowing conversions (Double -> CGFloat)

[CSGen] Don't favor overload choices that could require an implicit c…

307caf2

…onversion Doing so resulted in performance impact due to the number of partial solutions increase.

xedin force-pushed the implicit-cgfloat-conversion branch from 52607fb to 0e6198d Compare March 17, 2021 07:19

xwu reviewed Mar 17, 2021

View reviewed changes

[TypeChecker/SE-0307] NFC: Fix typos and clarify a couple of comments

a559ca8

Thanks Xiaodi Wu!

xedin marked this pull request as ready for review March 29, 2021 17:46

xedin merged commit f00c578 into swiftlang:main Mar 31, 2021

xedin changed the title ~~[DNM][TypeChecker] Implement Double <-> CGFloat implicit conversion~~ [TypeChecker] Implement Double <-> CGFloat implicit conversion Mar 31, 2021

typesanitizer mentioned this pull request Feb 8, 2021

[SR-14095] Improve diagnostic on incorrect usage of ^ and ** operators #56481

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[TypeChecker] Implement Double <-> CGFloat implicit conversion #34401

[TypeChecker] Implement Double <-> CGFloat implicit conversion #34401

xedin commented Oct 23, 2020

xedin commented Oct 23, 2020

xedin commented Oct 23, 2020

xwu commented Oct 26, 2020 •

edited

Loading

xedin commented Oct 26, 2020

xedin commented Oct 27, 2020

xedin commented Oct 28, 2020

swift-ci commented Oct 28, 2020

xedin commented Oct 31, 2020

swift-ci commented Oct 31, 2020

swift-ci commented Oct 31, 2020

xedin commented Nov 12, 2020

xedin commented Nov 14, 2020

xedin commented Nov 16, 2020

xedin commented Nov 17, 2020

xedin commented Nov 17, 2020

swift-ci commented Nov 17, 2020

xedin commented Jan 7, 2021

xedin commented Mar 17, 2021

xwu left a comment

xedin commented Mar 29, 2021

xedin commented Mar 29, 2021

xedin commented Mar 29, 2021

xedin commented Mar 30, 2021

xedin commented Mar 31, 2021

swift-ci commented Mar 31, 2021

swift-ci commented Mar 31, 2021

[TypeChecker] Implement Double <-> CGFloat implicit conversion #34401

[TypeChecker] Implement Double <-> CGFloat implicit conversion #34401

Conversation

xedin commented Oct 23, 2020

xedin commented Oct 23, 2020

xedin commented Oct 23, 2020

xwu commented Oct 26, 2020 • edited Loading

xedin commented Oct 26, 2020

xedin commented Oct 27, 2020

xedin commented Oct 28, 2020

swift-ci commented Oct 28, 2020

xedin commented Oct 31, 2020

swift-ci commented Oct 31, 2020

swift-ci commented Oct 31, 2020

xedin commented Nov 12, 2020

xedin commented Nov 14, 2020

xedin commented Nov 16, 2020

xedin commented Nov 17, 2020

xedin commented Nov 17, 2020

swift-ci commented Nov 17, 2020

xedin commented Jan 7, 2021

xedin commented Mar 17, 2021

xwu left a comment

Choose a reason for hiding this comment

xedin commented Mar 29, 2021

xedin commented Mar 29, 2021

xedin commented Mar 29, 2021

xedin commented Mar 30, 2021

xedin commented Mar 31, 2021

swift-ci commented Mar 31, 2021

swift-ci commented Mar 31, 2021

xwu commented Oct 26, 2020 •

edited

Loading