When using go.work, the go command packs module-specific errors into
synthetic results corresponding to the module query ("modulepath/...").
Extract these for use in diagnostics, by packing them into a new
moduleErrorMap error type.
Fixesgolang/go#50862
Change-Id: I68bf9cf4e9ebf4595acdc4da0347ed97171d637f
Reviewed-on: https://go-review.googlesource.com/c/tools/+/405259
Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Run-TryBot: Robert Findley <rfindley@google.com>
This reverts commit 5ca4cc8b9a.
Reason for revert: appears insufficient to land CL 410955.
Change-Id: If7029c9a43568e210e89d473b738400f5155715e
Reviewed-on: https://go-review.googlesource.com/c/tools/+/410995
Reviewed-by: Robert Griesemer <gri@google.com>
Reviewed-by: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Griesemer <gri@google.com>
Auto-Submit: Robert Griesemer <gri@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
1. Avoid unnecessary intermediate map updates.
2. Avoid accumulating defers in a loop when the control is simple.
Yield: -10% CPU, -37% allocs.
Typical results:
$ go test -v ./gopls/internal/regtest/bench -run=TestBenchmarkDidChange -didchange_dir=$HOME/w/kubernetes -didchange_file=pkg/util/hash/hash.go
Before:
BenchmarkStatistics 100 25932206 ns/op 11684109 B/op 75458 allocs/op
After:
BenchmarkStatistics 100 23294195 ns/op 11293472 B/op 47299 allocs/op
Also, move profiling logic outside the loop so that later runs
don't overwrite earlier runs. (This doesn't appear to be a problem
in practice, presumably because the last run is the big one.)
Updates golang/go#45686
Change-Id: I538ca6bb88cc18f1eefe35d2db29a62e5190280e
Reviewed-on: https://go-review.googlesource.com/c/tools/+/410697
Run-TryBot: Robert Findley <rfindley@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
Reviewed-by: Robert Findley <rfindley@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
This change replaces various uses of FileSet with either
nothing (when the parameter wasn't really needed) or token.File.
Notably, astutil.Imports was being used to extract the imports
of a file (available at ast.File.Imports), forcing a number
of wrappers to have a FileSet parameter.
Also, simplify various expressions file.Position().Line to file.Line().
Change-Id: I19fe86a18aba50352275f77ed737513744d3930c
Reviewed-on: https://go-review.googlesource.com/c/tools/+/410366
Run-TryBot: Alan Donovan <adonovan@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
Use a detached Context for all progress notifications. In particular,
using a detached Context for the window/workDoneProgress/create
notification avoids races where the $/cancelRequest notification and
create response cross paths, such that the client has created a progress
dialog but the server thinks that starting progress failed.
Also, as a matter of best practice don't store a context on the WorkDone
type, despite the fact that this Context is detached. Instead, only
close over a Context in the WorkDoneWriter, which requires a Context in
order to function but which implements the io.Writer interface.
The TestProgressBarErrors test should now pass reliably.
Fixesgolang/go#46930
Change-Id: I0d115ed3a62de97fe545c8dc0403e7bb55f6e481
Reviewed-on: https://go-review.googlesource.com/c/tools/+/409936
Run-TryBot: Robert Findley <rfindley@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Bryan Mills <bcmills@google.com>
Logic in mod.Hover relied upon ColumnMapper.TokFile being the only file
in its FileSet. Fix this to compare offsets with offsets.
Also use safetoken.InRange in semantic.go, to fix an incorrect bounds
check.
Change-Id: Ia18a0799fea1120674e3772efdf0bdb724e554e7
Reviewed-on: https://go-review.googlesource.com/c/tools/+/409934
TryBot-Result: Gopher Robot <gobot@golang.org>
Run-TryBot: Robert Findley <rfindley@google.com>
Reviewed-by: Alan Donovan <adonovan@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
It turns out PointSpan was only ever used as part of an oft repeated
pattern to get a token.Pos from a protocol position. Cut out the
middle.
Change-Id: I0b2c0fc3d335e6bbd3c1ac72c6f75e2c40c60ca5
Reviewed-on: https://go-review.googlesource.com/c/tools/+/408717
TryBot-Result: Gopher Robot <gobot@golang.org>
Run-TryBot: Robert Findley <rfindley@google.com>
Reviewed-by: Alan Donovan <adonovan@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
This eliminates some duplication, and lays the groundwork for removing
the use of token.File within ColumnMapper.
Change-Id: I54fe570bfc4f7bca0da643b8727e890dc6343208
Reviewed-on: https://go-review.googlesource.com/c/tools/+/406135
Run-TryBot: Robert Findley <rfindley@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Alan Donovan <adonovan@google.com>
The TokenConverter has been trimmed down to a thin wrapper around
token.File, and can now be removed.
Change-Id: I9985492274c88e6a13e6d62dadab5595c75c7840
Reviewed-on: https://go-review.googlesource.com/c/tools/+/406134
TryBot-Result: Gopher Robot <gobot@golang.org>
Run-TryBot: Robert Findley <rfindley@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
Reviewed-by: Alan Donovan <adonovan@google.com>
Eliminate the need to work around newline terminated files in
completion, by storing selection ranges as token.Pos and using an
lsppos.TokenMapper derived from the file content, which does not have
problems with newline termination.
This simplifies the completion logic, and removes the last use of
MappedRange.SpanRange, which is an inconisitent API in that it returns
positions in the compiled source, rather than edited source.
Change-Id: I65232787956325172b48fd42d85cbb598039ee5a
Reviewed-on: https://go-review.googlesource.com/c/tools/+/407889
Reviewed-by: Alan Donovan <adonovan@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Run-TryBot: Robert Findley <rfindley@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
Due to not having a pointer receiver, memoization of the computed
MappedRange.protocolRange had no effect.
Rather than fix the memoization, just remove it since it has apparently
not affected performance significantly.
Change-Id: Ib34becef44dca4074dcc7c8af93883d9a1060f8d
Reviewed-on: https://go-review.googlesource.com/c/tools/+/408716
Reviewed-by: Alan Donovan <adonovan@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Run-TryBot: Robert Findley <rfindley@google.com>
CL 405546 introduced a latent bug in MappedRange, because it naively
used the wrong TokenConverter to convert mapped positions to offsets.
This was detected via related clean-up work in another CL. Fix this by
passing the correct converter from MappedRange.Range. Add a test that
would have demonstrated the breakage.
More cleanup is needed here. It is subtle that MappedRange.Converter
maps the adjusted position for its start and end, and there may be some
places where this invariant has been broken over the years.
Add additional documentation and bug reports.
Change-Id: If7f177894bac1242ddcc1786e79c7559455e9291
Reviewed-on: https://go-review.googlesource.com/c/tools/+/407887
Run-TryBot: Robert Findley <rfindley@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Alan Donovan <adonovan@google.com>
Add additional bug reporting related to position invariants, and update
marker test runners to panic on encountered bugs.
This revealed a panic on older Go versions, where we try to create a
range for a missing package name span. Lift the check into the caller
for this case, and leave a big comment explaining that
checkForOrphanedFile probably shouldn't be so tolerant of invalid calls.
Change-Id: Ie16a07afba0f8a5682cca50ad8f04450bfa2da65
Reviewed-on: https://go-review.googlesource.com/c/tools/+/407885
Reviewed-by: Alan Donovan <adonovan@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
gopls-CI: kokoro <noreply+kokoro@google.com>
Package loading (at least via go list) fails when the import header
doesn't parse, so we need to invalidate metadata upon a state change
from non parsing->parsing. Refactor metadata invalidation to implement
this feature, add additional documentation, and avoid unnecessary work.
This change revealed a latent bug (via TestDeleteDirectory): when
statting a deleted directory failed, we could fail to invalidate any
package IDs, even those we already knew about. This bug was masked by
the somewhat complicated semantics of pkgNameChanged. The semantics of
pkgNameChanged are simplified to encapsulate any change to a
package->file association.
Also refactor the parsing API somewhat, and add a bug report if we don't
get a ParseGoFile while inspecting for metadata changes.
Update most regtests to panic upon receiving bug reports.
Fixesgolang/go#52981
Change-Id: I1838963ecc9c01e316f887aa9d8f1260662281ab
Reviewed-on: https://go-review.googlesource.com/c/tools/+/407501
Reviewed-by: Alan Donovan <adonovan@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Run-TryBot: Robert Findley <rfindley@google.com>
Our mechanism for reporting internal bugs doesn't work unless we
actually notice them. Add an undocumented option to receive showMessage
dialogs on the first bug occurring server-side.
Change-Id: I259a4c13161271c350fae06dc6ab0e1621725c92
Reviewed-on: https://go-review.googlesource.com/c/tools/+/399624
TryBot-Result: Gopher Robot <gobot@golang.org>
Run-TryBot: Robert Findley <rfindley@google.com>
Reviewed-by: Alan Donovan <adonovan@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
Adds a function for computing the core type of a type for use within x/tools.
Updates golang/go#52940
Change-Id: I91645c0b27031506be51a5f9d53b3e125e6fd1c2
Reviewed-on: https://go-review.googlesource.com/c/tools/+/406838
Run-TryBot: Tim King <taking@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
Reviewed-by: Robert Findley <rfindley@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Sometimes one wants to run the regtests with verbose output without
seeing the wall of RPC logs. Add a new flag to control the printing of
logs, rather than overloading testing.Verbose().
Change-Id: Iea5727e3079005d229319cbf86f67067f7c1f85d
Reviewed-on: https://go-review.googlesource.com/c/tools/+/405894
gopls-CI: kokoro <noreply+kokoro@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
Reviewed-by: Suzy Mueller <suzmue@golang.org>
TryBot-Result: Gopher Robot <gobot@golang.org>
This change makes it so that the stub methods analysis can recognize
errors happening to method and function call expressions that are being
passed a concrete type to an interface parameter. This way, a method stub CodeAction will appear at the call site.
Updates golang/go#37537
Change-Id: I886d53f06a85b9e5160d882aa742bb2b7fcea139
Reviewed-on: https://go-review.googlesource.com/c/tools/+/404655
gopls-CI: kokoro <noreply+kokoro@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
Reviewed-by: Suzy Mueller <suzmue@golang.org>
Run-TryBot: Suzy Mueller <suzmue@golang.org>
The only real implementation of position conversion was via a
*token.File, so refactor the converter logic to eliminate the Converter
interface, and just use a single converter implementation that uses a
*token.File to convert between offsets and positions.
This change is meant to be a zero-impact refactoring for non-test code.
As such, I abstained from panicking in several places where it would
make sense. In later CLs, once the bug reporting API lands, we can
insert bug reports in these places.
Change-Id: Id2e503acd80d089bc5d73e983215784015471f04
Reviewed-on: https://go-review.googlesource.com/c/tools/+/405546
gopls-CI: kokoro <noreply+kokoro@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
Reviewed-by: Alan Donovan <adonovan@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
The existing debug.Bug mechanism for reporting internal bugs is
insufficient for several reasons:
- It requires a context, which is not always available.
- By being defined in the debug package, it is subject to import
cycles.
- It is too complicated. Listening for bugs requires understanding the
event package.
Replace this with a simpler 'bug' package with no dependencies, that
allows reporting, listing, and listening on internal bugs. Hopefully
this will fulfill the goal of debug.Bug, to help us track down rare
bugs.
Change-Id: I30cab58429b29bd2d944d62e94f5657e40a760fc
Reviewed-on: https://go-review.googlesource.com/c/tools/+/399623
Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
Reviewed-by: Alan Donovan <adonovan@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Run-TryBot: Robert Findley <rfindley@google.com>
In JSON output, the Go command was only setting the Error field for one
package involved in an import cycle. As a result, TestResolveImportCycle
was dependent on the chosen package, causing flakes when the chosen
package is not deterministic.
Arguably this behavior should be fixed, both in the go command and in
gopls, but for now make the test resilient to choice by asserting on any
of the possible errors.
Add a new AnyOf expectation to support this type of assertion and tweak the
test output formatting. Also update the test to eagerly fail once the
didOpen notification has been fully processed, so that we don't have to
wait for the assertion timeout.
Updates golang/go#52904
Change-Id: Ic209d8fdcb7308c041b287a8f122c47e96d29a96
Reviewed-on: https://go-review.googlesource.com/c/tools/+/406274
Reviewed-by: Michael Matloob <matloob@golang.org>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
gopls-CI: kokoro <noreply+kokoro@google.com>
If -config=true, gopls vulncheck reads the package load
configuration JSON (build flags, env, tests, and later
maybe overlay info) from stdin.
And, add some logging to gopls/internal/vulncheck/command.go
that helps measuring the package loading overhead.
Update golang/go#50577
Change-Id: I5c1ce145b07f2bed03911613f42c09a3d6be6c28
Reviewed-on: https://go-review.googlesource.com/c/tools/+/404575
Reviewed-by: Robert Findley <rfindley@google.com>
In passing I noticed that this three-way comparison
is not (anti)symmetric. Such comparisons should consist
of a list of pairs of tests of this form:
if x.key1 < y.key1 { return -1 }
if x.key1 > y.key1 { return +1 }
...key2, etc...
return 0
Also in passing:
- simplify panic-proof debug string function.
- augment doc comment of (*Server).beginFileRequest
Change-Id: Idcd0844ea4e96fc2dee5bbc270f5a200b5c27aa0
Reviewed-on: https://go-review.googlesource.com/c/tools/+/405480
Reviewed-by: Robert Findley <rfindley@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Run-TryBot: Robert Findley <rfindley@google.com>
This CL adds a new analyzer for //go:embed directive, which
checks for the "embed" import.
Along with it, it improves doc for analysistest.Run for
comments of the the form "//...// want..." or
"//...// want..."
Updates #50262
Change-Id: I60ef0ab740feadd4fff3a4d758123b27ceda0bc6
Reviewed-on: https://go-review.googlesource.com/c/tools/+/400854
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Peter Weinberger <pjw@google.com>
Reviewed-by: Robert Findley <rfindley@google.com>
Run-TryBot: Nooras Saba <saba@golang.org>
gopls-CI: kokoro <noreply+kokoro@google.com>
And fix a minor issue in copytermlist.go --
copytermlist.go replaces type names from go/types with
qualified type names. Use of token.NoPos (filled with
ast.NewIdent) however confuses the go/format printer.
As a result, while transforming
func (x *term) includes(t Type) bool
to
func (x *term) includes(t types.Type) bool
go/format printer failed to compute the end position of
the parameter list, concluded RPAREN should be in a
different line from the parameter list, added a comma,
and printed
func (x *term) includes(t types.Type,) bool
Reuse the replaced node's position instead. (not 100%
correct, but better than NoPos)
Change-Id: Ia34e11562cc80c68dcf4b921ffffd926971c2215
Reviewed-on: https://go-review.googlesource.com/c/tools/+/405536
Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
Reviewed-by: Robert Findley <rfindley@google.com>
The internal/typeparams directory contains an initial draft of the
documentation later published under the golang.org/x/exp/typeparams
module. Remove this draft to avoid confusion.
Change-Id: I1f40468f99c0d2885e8fc380c75669f654ba971e
Reviewed-on: https://go-review.googlesource.com/c/tools/+/405260
Reviewed-by: Alan Donovan <adonovan@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Run-TryBot: Robert Findley <rfindley@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
Optimize building the symbol index for a file, in two ways:
- use the cached full parse tree, if it already exists
- if it doesn't exist, optimize parsing by skipping both comments and
object resolution, which aren't necessary for symbols
This results in around 3x faster initial indexing of symbols. In my
manual testing, indexing of Kubernetes went from 16s->5s, and indexing
of x/tools went from 2.4s->700ms.
Also fix a typo in gopls/internal/regtest/bench/bench_test.go.
Fixes#52602
Change-Id: I0893e95410be96e94e5e9dee7a3aab30b59c19c5
Reviewed-on: https://go-review.googlesource.com/c/tools/+/403679
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
gopls-CI: kokoro <noreply+kokoro@google.com>
Reviewed-by: Alan Donovan <adonovan@google.com>
This avoids an import cycle that prevented these wrappers from being
used in the lsppos package.
Change-Id: I9eedd256db983dfcf962edba39e3d4f3a1aabdeb
Reviewed-on: https://go-review.googlesource.com/c/tools/+/403680
TryBot-Result: Gopher Robot <gobot@golang.org>
Run-TryBot: Robert Findley <rfindley@google.com>
Reviewed-by: Alan Donovan <adonovan@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
For use-cases that work only with token.Pos and protocol.Position, the
span package is unnecessarily indirect, and inefficient. It also loses
information about newline termination, and handles positions within CRLF
line endings incorrectly.
The lsppos package was written to bypass this complexity, but had
limited use and lacked tests.
Add tests, and an wrapper API that operates on token.Pos. Also fix
source.TestTokenOffset to not panic, and add a temporary exemption of
the new token.Offset usage.
This change also fixes position calculation in the case of empty file
content. The mapper now finds position (0, 0) at offset 0 of an empty
file.
Change-Id: I639bd3fac78a127b1c8eddad60b890449901c68c
Reviewed-on: https://go-review.googlesource.com/c/tools/+/403678
Reviewed-by: Alan Donovan <adonovan@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Instead of invoking the command through the LSP custom command,
call the vulncheck command hook directly. That reduces the extra
overhead of bringing up the full gopls server & package loading.
The vulncheck hook loads packages again any way, so the benefit
of running the check inside gopls's custom command framework
is not huge any more.
Still `gopls vulncheck` is useful - editors don't need to install
another binary for vulncheck feature, and it will output the
result in the format easier to handle than what `govulncheck`
currently offers.
Updates golang/go#50577
Change-Id: Ia21e6d7e0c37c4a1b02dc8bbca860143524c3d1b
Reviewed-on: https://go-review.googlesource.com/c/tools/+/404574
Reviewed-by: Robert Findley <rfindley@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
gopls-CI: kokoro <noreply+kokoro@google.com>
Run-TryBot: Hyang-Ah Hana Kim <hyangah@gmail.com>
While working on golang/go#52715, I discovered an infinite recursion in
gopls' completion logic: eachField assumes a finiteness of type pointers.
It is almost certainly a go/types bug that type-checked types expand
infinitely, but nevertheless we should use the more accurate
typeutil.Map for short-circuiting our search.
Change-Id: Ib1c7125e624f42882869acd4e0476e317d4da056
Reviewed-on: https://go-review.googlesource.com/c/tools/+/404335
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Alan Donovan <adonovan@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
Provide a default value for unsafe.Pointer in fillstruct.
Fixesgolang/go#52640
Change-Id: I10a1878fbf53b082f83f44e0ba2788ead14439d6
Reviewed-on: https://go-review.googlesource.com/c/tools/+/403535
TryBot-Result: Gopher Robot <gobot@golang.org>
Run-TryBot: Suzy Mueller <suzmue@golang.org>
gopls-CI: kokoro <noreply+kokoro@google.com>
Reviewed-by: Robert Findley <rfindley@google.com>
Reviewed-by: David Chase <drchase@google.com>
'gopls help cmd...' is redirected to 'gopls cmd... -h'.
The tool.Run operation for each subcommand already has
logic to show the usage message when it parses a -h flag.
Examples:
$ gopls help
$ gopls help remote
$ gopls help remote sessions
Fixesgolang/go#52598
Change-Id: I5135c6e7b130d839f880d0613534ac08de199bcf
Reviewed-on: https://go-review.googlesource.com/c/tools/+/403677
Run-TryBot: Alan Donovan <adonovan@google.com>
Reviewed-by: Robert Findley <rfindley@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
When a file is new or its package name has changed, we should invalidate
all packages for files in the current directory, not just packages
previously containing the file.
Fixesgolang/go#52500
Change-Id: I9fc9857a7abcd4e730068871c899d274e1736967
Reviewed-on: https://go-review.googlesource.com/c/tools/+/401795
Run-TryBot: Robert Findley <rfindley@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Suzy Mueller <suzmue@golang.org>
Give placeholders for type params in func literal completions. For
example:
func foo[T any](func(T) T) {}
foo(<>)
Will now give "func(<T>) <T> {}" where <> denotes a placeholder.
Change-Id: Iadde73ed6b88e1410c28dfa33a20ab6a51235c93
Reviewed-on: https://go-review.googlesource.com/c/tools/+/400616
Run-TryBot: Muir Manders <muir@mnd.rs>
gopls-CI: kokoro <noreply+kokoro@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
In cases like:
type foo[T any] struct{}
func bar[T any](foo[T]) {}
bar[int](<A>)
bar(<B>)
At <A> we will now offer "foo[int]{}". At <B> we will now offer a
snippet "foo[<T>]{}" which lets the user fill in the type arg.
Note that we have no knowledge of type inference, so you can be left
with superfluous type args after completion.
Change-Id: Ia7d63284f3317d9367864fdae3e3f9ae68fdff1a
Reviewed-on: https://go-review.googlesource.com/c/tools/+/400615
Run-TryBot: Muir Manders <muir@mnd.rs>
gopls-CI: kokoro <noreply+kokoro@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
In cases like:
func foo[T int | string](T) {}
foo[int](<>)
Previously at <> we would favor int and string candidates. This is
because go/types doesn't instantiate foo in this case (for some
reason). Work around the issue by using types.CheckExpr to re-check
the *ast.CallExpr.Fun. CheckExpr seems to do a better than a full type
check in the face of errors.
Updates golang/go#52291
Updates golang/go#52503
Change-Id: Ide436428f3232db2e06ea3cc22ea250edbf28685
Reviewed-on: https://go-review.googlesource.com/c/tools/+/400614
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Reviewed-by: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
changes
In principle, we should only parse through the cache when in workspace
mode, else we risk pinning AST in memory that we'll never need.
Alternatively: when unsure we should default to NOT parsing through the
cache.
For now, just update the logic to check for metadata changes to not
memoize its result.
Change-Id: I200af9ffb3353ba8065e46100a588dce6239dc66
Reviewed-on: https://go-review.googlesource.com/c/tools/+/401794
Run-TryBot: Robert Findley <rfindley@google.com>
Reviewed-by: Suzy Mueller <suzmue@golang.org>
gopls-CI: kokoro <noreply+kokoro@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Update FindHoverContext to retrieve & render package documentation for a hovered package. Add a regtest for this hovering feature.
Updates golang/go#51848
Change-Id: If57396d59be9c4cf7e09b64e39832de6f996c7ca
Reviewed-on: https://go-review.googlesource.com/c/tools/+/400820
Reviewed-by: Robert Findley <rfindley@google.com>
Run-TryBot: Dylan Le <dungtuanle@google.com>
gopls-CI: kokoro <noreply+kokoro@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
In cases like:
func foo[A int|string](a A) {}
foo[_](<>)
We now prefer ints and strings at <> by matching against the type
constraint. Note that even if "_" is replaced with "int", we still
prefer strings since the type checker doesn't seem to want to
instantiate foo unless the params check out.
Change-Id: I0e7acfef0775752a96fcfe23e7e2e3d939820eee
Reviewed-on: https://go-review.googlesource.com/c/tools/+/394017
Run-TryBot: Muir Manders <muir@mnd.rs>
gopls-CI: kokoro <noreply+kokoro@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
Reviewed-by: Suzy Mueller <suzmue@golang.org>
Auto-Submit: Peter Weinberger <pjw@google.com>
Robert Findley
Robert Griesemer
Robert Griesemer
Alan Donovan
Hana (Hyang-Ah) Kim
Tim King
Marwan Sulaiman
Nooras Saba
Suzy Mueller
Muir Manders
Dylan Le