cmd/compile: add new range behind GOEXPERIMENT=range

[rsc]

It would be good to get code for #61405 into the tree for easier use.
I propose we check in the current prototype behind GOEXPERIMENT=range.

[rsc]

This proposal has been added to the active column of the proposals project
and will now be reviewed at the weekly proposal review meetings.
— rsc for the proposal review group

[earthboundkid]

Would it make sense to have a experimental coro package users could try too?

[rsc]

No change in consensus, so accepted. 🎉
This issue now tracks the work of implementing the proposal.
— rsc for the proposal review group

[aclements]

Would it make sense to have a experimental coro package users could try too?

@carlmjohnson , I think it's pretty unclear what we want to do with coro. Note that @rsc's new proposal for the iter package (#61897) includes iter.Pull, which will probably use the coro techniques under the hood, though coro may not become a package in its own right.

[gopherbot]

Change https://go.dev/cl/518376 mentions this issue: go/ssa/interp: disable test for GOROOT/test/range.go

[gopherbot]

Change https://go.dev/cl/510541 mentions this issue: cmd/compile: implement range over func

[gopherbot]

Change https://go.dev/cl/510537 mentions this issue: cmd/compile, go/types: typechecking of range over int, func

[gopherbot]

Change https://go.dev/cl/510538 mentions this issue: cmd/compile: implement range over integer

[meling]

Will these changes make its way into gopls in some way, so that the "experiment" can be used from IDEs without compile errors?

[DeedleFake]

Is this version of the implementation going to have unused bool returns from the outer function removed?

Also, I wanted to just turn this on via gotip env -w GOEXPERIMENT=range so that it would be easier to play around with, but doing so breaks my usual Go installation. Every command just gives an unknown experiment error and fails. I tried setting it as an environment variable when running gotip download 510541, but that just failed with a claim that there was no valid Go toolchain available to build with. Is this expected behavior? That latter one in particular feels like a bug to me.

Edit: I tried building the toolchain with the environment variable already set again in Linux this time and it worked this time, except that the defaults are still what they already were.

[meling]

I get this:

% GOEXPERIMENT=range gotip version
go: unknown GOEXPERIMENT range
% gotip version
go version devel go1.22-51ed3cb702 Wed Aug 30 21:18:04 2023 +0000 darwin/arm64
% GOEXPERIMENT=range gotip download
Updating the go development tree...
From https://go.googlesource.com/go
 * branch                  master     -> FETCH_HEAD
HEAD is now at 51ed3cb702 net: retry TestDialTimeout subtests with progressively shorter timeouts
Building Go cmd/dist using /usr/local/go. (go1.21.0 darwin/arm64)
Building Go toolchain1 using /usr/local/go.
Building Go bootstrap cmd/go (go_bootstrap) using Go toolchain1.
Building Go toolchain2 using go_bootstrap and Go toolchain1.
go_bootstrap: unknown GOEXPERIMENT range
go tool dist: FAILED: /Users/meling/sdk/gotip/pkg/tool/darwin_arm64/go_bootstrap install -pgo=off cmd/asm cmd/cgo cmd/compile cmd/link: exit status 2
gotip: failed to build go: exit status 2

Anyone got this working on macOS?

[meling]

Ok. I had to download gotip again:

% go install golang.org/dl/gotip@latest

Now I can compile with GOEXPERIMENT=range.

[rsc]

It's not all checked in yet. When it is, I'll close this issue.

[aclements]

This is all checked in now (and has been for a while). Closing.