Add data structure for memory efficient lists of std::variants

[weinig]

fdc91dd

Add data structure for memory efficient lists of std::variants
https://bugs.webkit.org/show_bug.cgi?id=283423

Reviewed by Darin Adler.

Adds a new data structure, VariantList<std::variant<Ts...>> which acts
like a Vector<std::variant<Ts...>> except that each element only takes up
its own size (+ tag size). The tradeoff is that the list is not random
access capable, and access to the elements is only available via iteration.
In practice, this tradeoff is a good one for the CSS and Style value types.

Rather than vending std::variants, VariantList vends VariantListProxy
objects which act like std::variants, but don't require the extra costs
of construction.

* Source/WTF/WTF.xcodeproj/project.pbxproj:
* Source/WTF/wtf/CMakeLists.txt:
* Source/WTF/wtf/StdLibExtras.h:
* Source/WTF/wtf/VariantList.h: Added.
* Source/WTF/wtf/VariantListOperations.h: Added.
* Tools/TestWebKitAPI/TestWebKitAPI.xcodeproj/project.pbxproj:
* Tools/TestWebKitAPI/Tests/WTF/VariantList.cpp: Added.

Canonical link: https://commits.webkit.org/287010@main

8ece8d1

Misc	iOS, visionOS, tvOS & watchOS	macOS	Linux	Windows
❌ 🧪 style	✅ 🛠 ios	✅ 🛠 mac	✅ 🛠 wpe	✅ 🛠 win
✅ 🧪 bindings	✅ 🛠 ios-sim	✅ 🛠 mac-AS-debug	✅ 🧪 wpe-wk2	✅ 🧪 win-tests
✅ 🧪 webkitperl	✅ 🧪 ios-wk2	✅ 🧪 api-mac	✅ 🧪 api-wpe
	✅ 🧪 ios-wk2-wpt	✅ 🧪 mac-wk1	✅ 🛠 wpe-cairo
❌ 🛠 🧪 jsc	✅ 🧪 api-ios	✅ 🧪 mac-wk2	✅ 🛠 gtk
❌ 🛠 🧪 jsc-arm64	✅ 🛠 vision	🧪 mac-AS-debug-wk2	❌ 🧪 gtk-wk2
	✅ 🛠 vision-sim	✅ 🧪 mac-wk2-stress	✅ 🧪 api-gtk
✅ 🛠 🧪 merge	✅ 🧪 vision-wk2	✅ 🧪 mac-intel-wk2	✅ 🛠 jsc-armv7
	✅ 🛠 tv		❌ 🧪 jsc-armv7-tests
	✅ 🛠 tv-sim
	✅ 🛠 watch
	✅ 🛠 watch-sim

[webkit-early-warning-system]

EWS run on previous version of this PR (hash 8b0b757)

Details

Misc	iOS, visionOS, tvOS & watchOS	macOS	Linux	Windows
❌ 🧪 style	✅ 🛠 ios	✅ 🛠 mac	✅ 🛠 wpe	✅ 🛠 win
✅ 🧪 bindings	✅ 🛠 ios-sim	✅ 🛠 mac-AS-debug	✅ 🧪 wpe-wk2	✅ 🧪 win-tests
✅ 🧪 webkitperl	✅ 🧪 ios-wk2	✅ 🧪 api-mac	✅ 🧪 api-wpe
	✅ 🧪 ios-wk2-wpt	✅ 🧪 mac-wk1	✅ 🛠 wpe-cairo
❌ 🛠 🧪 jsc	✅ 🧪 api-ios	✅ 🧪 mac-wk2	✅ 🛠 gtk
✅ 🛠 🧪 jsc-arm64	✅ 🛠 vision	✅ 🧪 mac-AS-debug-wk2	✅ 🧪 gtk-wk2
	✅ 🛠 vision-sim	✅ 🧪 mac-wk2-stress	✅ 🧪 api-gtk
	✅ 🧪 vision-wk2	✅ 🧪 mac-intel-wk2	✅ 🛠 jsc-armv7
	✅ 🛠 tv		✅ 🧪 jsc-armv7-tests
	✅ 🛠 tv-sim
	✅ 🛠 watch
	✅ 🛠 watch-sim

[weinig]

I have built these almost entirely using VectorBuffer and std::spans, but I can't quite figure out how to appease the unsafe buffers warnings in a few places where I use a pointer, not to derereference anything, just to do some math to figure out how much we need to grow.

I figure those same problems will have to be solved when updating Vector, so I could crib off that when the time comes.

[darinadler]

For the most part, @cdumez has convinced me to just use uint8_t for generic bytes and not start to use std::byte. Maybe you can convince me otherwise, after all uint8_t and LChar are the same type and one advantage of std::byte is that it’s not the same as any other type!

[weinig]

Other than some extra overloads, what's the argument?

[darinadler]

I guess maybe extra type conversions?

[weinig]

I would think that would be just a temporary problem while making the move from uint8_t to std::byte for actual opaque byte buffers.

But either way, this is entirely constrained to the data structure, so if we decide we really don't like std::byte, changing it will be very easy.

[darinadler]

This additional overloading is part of the cost of starting to use std::byte. I suppose byteCast sprinkling is less elegant.

[darinadler]

Consider putting these into Forward.h instead?

[darinadler]

I would normally omit the type argument name Variant from these forward declarations, but I can see including it if you think it adds clarity. I would also put forward declarations like these above a type definition like VariantListTag.

[darinadler]

An idiom like this seems like it needs the comment that explains what its intentions are.

[webkit-early-warning-system]

EWS run on current version of this PR (hash 8ece8d1)

Details

Misc	iOS, visionOS, tvOS & watchOS	macOS	Linux	Windows
❌ 🧪 style	✅ 🛠 ios	✅ 🛠 mac	✅ 🛠 wpe	✅ 🛠 win
✅ 🧪 bindings	✅ 🛠 ios-sim	✅ 🛠 mac-AS-debug	✅ 🧪 wpe-wk2	✅ 🧪 win-tests
✅ 🧪 webkitperl	✅ 🧪 ios-wk2	✅ 🧪 api-mac	✅ 🧪 api-wpe
	✅ 🧪 ios-wk2-wpt	✅ 🧪 mac-wk1	✅ 🛠 wpe-cairo
❌ 🛠 🧪 jsc	✅ 🧪 api-ios	✅ 🧪 mac-wk2	✅ 🛠 gtk
❌ 🛠 🧪 jsc-arm64	✅ 🛠 vision	🧪 mac-AS-debug-wk2	❌ 🧪 gtk-wk2
	✅ 🛠 vision-sim	✅ 🧪 mac-wk2-stress	✅ 🧪 api-gtk
✅ 🛠 🧪 merge	✅ 🧪 vision-wk2	✅ 🧪 mac-intel-wk2	✅ 🛠 jsc-armv7
	✅ 🛠 tv		❌ 🧪 jsc-armv7-tests
	✅ 🛠 tv-sim
	✅ 🛠 watch
	✅ 🛠 watch-sim

[weinig]

Doesn't look like any of those failures are related to this PR.

[webkit-commit-queue]

Committed 287010@main (fdc91dd): https://commits.webkit.org/287010@main

Reviewed commits have been landed. Closing PR #36900 and removing active labels.