Loading completion suggestions can be slow, especially in GOPATH mode
where basically anything can change at any time. As a compromise, cache
everything for 30 seconds. Specifically, after a completion operation
finishes, if the cache is more than 30 seconds old, refresh it
asynchronously. That keeps user-facing latency consistent, without
chewing up CPU when the editor isn't in use. It does mean that if you
walk away for an hour and come back, the first completion may be stale.
In module mode this is relatively benign. The only things the
longer caching affects are the main module and replace targets, and
relevant packages in those will generally be loaded by gopls, so they'll
have full, up-to-date type information regardless.
In GOPATH mode this may be more troublesome, since it affects
everything. In particular, go get -u of a package that isn't imported
yet won't be reflected until the cache period expires. I think that's a
rare enough case not to worry about.
Change-Id: Iaadfd0ff647cda2b1dcdead9254b5492b397e86e
Reviewed-on: https://go-review.googlesource.com/c/tools/+/205163
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Packages that aren't imported in the current file will often have been
used elsewhere, which means that gopls will have their type information
available. Expose loaded packages in the Snapshot, and try to use that
information when possible for unimported packages.
Change-Id: Icb672618a9f9ec31b9796f0c5da56ed3d2b38aa7
Reviewed-on: https://go-review.googlesource.com/c/tools/+/204824
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
When a user completes rand.<>, propose rand.Seed (from math/rand) and
rand.Prime (from crypto/rand), etc.
Because we don't necessarily have type checking information for
unimported packages, I had to add shortcut cases to a number of
functions around the completion code. Better suggestions welcome.
Change-Id: I7822dc75c86b24156963e7bdd959443f4f2748b1
Reviewed-on: https://go-review.googlesource.com/c/tools/+/204819
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Reviewed-by: Muir Manders <muir@mnd.rs>
The metadata was being added to the cache before it was fully computed.
Change-Id: I6931476a715f0383f7739fa4e950dcaa6cbec4fe
Reviewed-on: https://go-review.googlesource.com/c/tools/+/204562
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Cottrell <iancottrell@google.com>
This way, we can tell if someone is building gopls at head, or if they
are using a tagged version. We should only change the version string on
a release branch.
Change-Id: I55eba534baa2e171315fe72242a400e2d2794314
Reviewed-on: https://go-review.googlesource.com/c/tools/+/205159
Reviewed-by: Ian Cottrell <iancottrell@google.com>
A lot has changed since golang/go#32794 was filed, and we now have many more
tests for the command line.
Fixesgolang/go#32794
Change-Id: Ib268865a2345fd6676b2679bd76197c2d8658a85
Reviewed-on: https://go-review.googlesource.com/c/tools/+/204818
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Cottrell <iancottrell@google.com>
Some analyses may panic, so we should be careful to recover.
Change-Id: Ie13df07aca1a21f4e6a66648cd4890b6a31966cc
Reviewed-on: https://go-review.googlesource.com/c/tools/+/204817
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
If the identifier doesn't have type info, don't try to access it.
Change-Id: I7e7c834c2863ec82a1269e60f7c50b0ce03f4692
Reviewed-on: https://go-review.googlesource.com/c/tools/+/204740
Run-TryBot: Michael Matloob <matloob@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
This change adds command line support for symbols.
Symbols are formatted as '{name} {type} {range}', with
children being preceded by a \t.
Example:
$ gopls symbols ~/tmp/foo/main.go
$
$ x Variable 7:5-7:6
$ y Constant 9:7-9:8
$ Quux Struct 29:6-29:10
$ Do Method 37:16-37:18
$ X Field 30:2-30:3
$ Y Field 30:5-30:6
Updates golang/go#32875
Change-Id: I1272fce733fb12b67e3d6fb948f5bf3de4ca2ca1
Reviewed-on: https://go-review.googlesource.com/c/tools/+/203609
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
This adds support for calling links from the gopls command line,
e.g.
$ gopls links ~/tmp/foo/main.go
Optional arguments are:
-json, which emits range and uri in JSON
With no arguments, a unique list of links are emitted.
Updates golang/go#32875
Change-Id: I1e7cbf00a636c05ccf21bd544d9a5b7742d5d70b
GitHub-Last-Rev: 7ed1e4612186bce4077d3c73f2407cf6def211d9
GitHub-Pull-Request: golang/tools#181
Reviewed-on: https://go-review.googlesource.com/c/tools/+/203297
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
The approach of using ASTs to determine error ranges had more
complications than I anticipated. Revert it for now to go back to the
old user experience, while we consider a better approach.
Change-Id: I5b23f0147c26089330f8a4bbf7b6914ae66cf59a
Reviewed-on: https://go-review.googlesource.com/c/tools/+/204561
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Cottrell <iancottrell@google.com>
This change stops the memory leak from happening. It does not stop the
large number of allocations done by analysis, however, so these still
need to be prevented through correct cancellation.
Change-Id: I1cf7fbf6f85ed413af1f2ea55f91862a6d7f2db7
Reviewed-on: https://go-review.googlesource.com/c/tools/+/204558
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Cottrell <iancottrell@google.com>
This change adds command line support for signatureHelp.
If the location provided corresponds to a function, that
function signature is displayed. In case that function is
documented the related comment is shown as well.
Example:
$ gopls signature ~/tmp/foo/main.go:7:5
$
$ Next(n int) []byte
$
$ Next returns a slice containing the next n bytes from
$ the buffer, advancing the buffer as if the bytes had been
$ returned by Read.
Note that linebreaks shown in the comment are just to adhere
commit message guidelines. The command prints documentation
comments on one line.
Updates golang/go#32875
Change-Id: Ib0dcc3267c594f95d80b74f289c1235c2c0c5f64
Reviewed-on: https://go-review.googlesource.com/c/tools/+/204057
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
This adds support for the LSP implemention call, based
on the guru code for getting implementations. The guru code
did much more than we need, so some of the code has been
dropped, and other parts of it are ignored (for now).
Fixesgolang/go#32973
Change-Id: I1a24450e17d5364f25c4b4120be5320b13ac822b
Reviewed-on: https://go-review.googlesource.com/c/tools/+/203918
Run-TryBot: Michael Matloob <matloob@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
This adds support for calling suggestedfix from the gopls command line, e.g.
$ gopls suggestedfix ~/tmp/foo/main.go
Optional arguments are:
-w, which writes the changes back to the original file; and
-d, which prints a unified diff to stdout
With no arguments, the changed files are printed to stdout.
Wasn't sure if the command should be `suggestedfix` or just `fix` or `quickfix`?
Also this applies all changes to a file, does not allow for selective fixes.
Updates golang/go#32875
Change-Id: I8b75f9824be82974f6edb7c03383b4d56116943c
GitHub-Last-Rev: 070fcda33ac3494bfe8f19c2cd78c089c713ed98
GitHub-Pull-Request: golang/tools#174
Reviewed-on: https://go-review.googlesource.com/c/tools/+/202480
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
When proposing packages to import, we can propose more relevant packages
first. Introduce that concept to the pkg struct, and sort by it when
returning candidates.
In all cases we prefer stdlib packages first. Then, in module mode, we
prefer packages that are in the module's dependencies over those that
aren't. We could go further and prefer direct deps over indirect too,
but I didn't have the code for that handy.
I also changed the alphabetical sort from import path to package name,
because that's what the user sees first in the UI.
Updates golang/go#31906
Change-Id: Ia981ee9ffe3202e2a68eef3a36f65e81849a4ac2
Reviewed-on: https://go-review.googlesource.com/c/tools/+/204203
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
go/parser has switched from reporting no position for the end of a
broken file to reporting an invalid position. This broke on of our tests
that contains broken code. Change the test case as a result.
Change-Id: I4feb7790539994e593c56d5ae84929364c1eec1c
Reviewed-on: https://go-review.googlesource.com/c/tools/+/204202
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
Reviewed-by: Ian Cottrell <iancottrell@google.com>
When our expected type is a named type from another package, we now always
search that other package for completion candidates, even if it is not currently
imported.
Consider the example:
-- foo.go --
import "context"
func doSomething(ctx context.Context) {}
-- bar.go--
doSomething(<>)
"bar.go" doesn't import "context" yet, so normally you need to first import
"context" through whatever means before you get completion items from "context".
Now we notice that the expected type's package hasn't been imported yet and give
deep completions from "context".
Another use case is with literal completions. Consider:
-- foo.go --
import "bytes"
func doSomething(buf *bytes.Buffer) {}
-- bar.go--
doSomething(<>)
Now you will get a literal completion for "&bytes.Buffer{}" in "bar.go" even
though it hasn't imported "bytes" yet.
I had to pipe the import info around a bunch of places so the import is added
automatically for deep completions and literal completions.
Change-Id: Ie86af2aa64ee235038957c1eecf042f7ec2b329b
Reviewed-on: https://go-review.googlesource.com/c/tools/+/201207
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Closing over the checkPackageHandle creates a cycle that forces the
checkPackageHandle not to be garbage collected until the value is
created. If a value is never created, the handle will not be collected.
Change-Id: I0f94557da917330ebe307a0e843b16ca7382e210
Reviewed-on: https://go-review.googlesource.com/c/tools/+/204079
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
Reviewed-by: Ian Cottrell <iancottrell@google.com>
This change encompasses the refactorings needed to correctly implement
CL 204079. The goal of this CL is to make the actual relevant diffs more
clear.
Change-Id: I38acfd436e2380be790910e01b6e37d8280e9100
Reviewed-on: https://go-review.googlesource.com/c/tools/+/204139
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
Reviewed-by: Ian Cottrell <iancottrell@google.com>
This adds support for calling import from the gopls command line,
e.g.
$ gopls imports -w ~/tmp/foo/main.go
Optional arguments are:
-w, which writes the changes back to the original file; and
-d, which prints a unified diff to stdout
With no arguments, the changed file is printed to stdout.
Updates golang/go#32875
Change-Id: I12f980d977fe12c16e51b024c9dd28c33ba6c002
GitHub-Last-Rev: c3fdd90e25204e7a12a94e9dfde389b7674e7e6d
GitHub-Pull-Request: golang/tools#176
Reviewed-on: https://go-review.googlesource.com/c/tools/+/202624
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
This change eliminates the need for the importer struct. We should no
longer need the "seen" map for cycle detection. This is because
go/packages will not return import maps with cycles, and we fail in the
Import function if we see an import we do not recognize.
Change-Id: I06922c74e07eb47ce63b56fa2ac2099e7fc8bd8a
Reviewed-on: https://go-review.googlesource.com/c/tools/+/202299
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Cottrell <iancottrell@google.com>
This adds (or makes exported) a convenience function for reporting diagnostics with a
node directly (which is what folks usually want).
Change-Id: Ieb7ef2703f99d3a24ba7e48a779be62a7761cd0c
Reviewed-on: https://go-review.googlesource.com/c/tools/+/180237
Run-TryBot: Michael Matloob <matloob@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Cottrell <iancottrell@google.com>
In cases like:
type myInt int
const (
a = 1
b myInt = 2
)
var foo myInt = <>
We now prefer "b" over "a" since b's type matches the expected type
exactly.
Change-Id: I675934761cc17f6b303b63b4715b31dd1af7cea1
Reviewed-on: https://go-review.googlesource.com/c/tools/+/202737
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Now a budget of 0 disables mean unlimited and tests no longer set the
budget. Hopefully the deep completion tests will stop flaking.
Updates golang/go#34617
Change-Id: Icdff5e78dcf1cc3d3fcbf0326716b39b00f0a8c1
Reviewed-on: https://go-review.googlesource.com/c/tools/+/203338
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Also, address minor comments that I ignored from Heschi because I am
obnoxious.
Change-Id: I99dcac38578585af2cdd951dd2b9755732ef945f
Reviewed-on: https://go-review.googlesource.com/c/tools/+/203281
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Now that we are using the memoize package to cache analysis results, we
can use that cache for suggested fixes.
Change-Id: I42905a6fe575f49d38979d53d58ea8ec59210ae0
Reviewed-on: https://go-review.googlesource.com/c/tools/+/203278
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>
There is no need to cache the pass on the actionHandle, as it does not
need to be reused and does not live outside the exec function.
Change-Id: I1737271383776b35718df3475b4f888232d57ae4
Reviewed-on: https://go-review.googlesource.com/c/tools/+/203177
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
We now expect a type name when in the key or value of a *ast.MapType.
I also added an extra filter to expect a comparable type for the key.
Change-Id: I647cf4d791b2c0960ad3b12702b91b9bc168599b
Reviewed-on: https://go-review.googlesource.com/c/tools/+/197439
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
We don't need to worry about a package's errors unless it is the
top-level package. Also fix some fallback logic in the type error range
computation.
Change-Id: Ib26b5e25bd70193ea24ec4a197811eedf69b0e2c
Reviewed-on: https://go-review.googlesource.com/c/tools/+/202622
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
When VSCode applies its own fuzzy matching/filtering/ranking logic to
completion candidates, it can end up reordering and even omitting some
of our candidates. It is mainly a problem with deep completions (i.e.
VSCode downranks or completely hides deep completion candidates that
should be ranked at the top).
We now trick VSCode into not reordering our candidates by setting each
candidate's "filterText" to the completion prefix. This makes every
candidate look like an identically perfect match, so VSCode just
maintains the order specified by "sortText".
Note that we don't do this trick if server side fuzzy matching and
deep completions are disabled. In this case unimpeded client side
candidate filtering is necessary.
Change-Id: I677047bca12b9ce05a953016d0d89182f1fe44d6
Reviewed-on: https://go-review.googlesource.com/c/tools/+/202717
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
This change effectively reverts CL 202039. This CL was a mistake, as it
creates a cycle. Snapshots hold CheckPackageHandles, which in turn hold
pkgs.
Change-Id: I944304cb365f0ef98b5e54ea38edea6cece40453
Reviewed-on: https://go-review.googlesource.com/c/tools/+/202740
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>
If there are no imports that need organizing, don't send the "Organize
Imports" code action.
Change-Id: Id01521edd1524fb3f7372fd787d6c90418740cf3
Reviewed-on: https://go-review.googlesource.com/c/tools/+/202825
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
*ast.ArrayTypes are type expressions like "[]foo" or "[2]int". They
show up as standalone types (e.g. "var foo []int") and as part of
composite literals (e.g. "[]int{}"). I made the following
improvements:
- Always expect a type name for array types.
- Add a "type modifier" for array types so completions can be smart
when we know the expected type. For example:
var foo []int
foo = []i<>
we know we want a type name, but we also know the expected type is
"[]int". When evaluating type names such as "int" we turn the type
into a slice type "[]int" to match against the expected type.
- Tweak the AST fixing to add a phantom selector "_" after a naked
"[]" so you can complete directly after the right bracket.
I split out the type name related type inference bits into a separate
typeNameInference struct. It had become confusing and complicated,
especially now that you can have an expected type and expect a type
name at the same time.
Change-Id: I00878532187ee5366ab8d681346532e36fa58e5f
Reviewed-on: https://go-review.googlesource.com/c/tools/+/197438
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
This change modifies the invalidContent function to take a file change
type. This allows us to eliminate the separate invalidateMetadata
function. The logic of watching changed files is then further pushed
into the caching layer.
Updates golang/go#34218
Change-Id: Id31b3931c45ec408b6e7b4a362e00f9091ba4f70
Reviewed-on: https://go-review.googlesource.com/c/tools/+/201221
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Cottrell <iancottrell@google.com>
Originally the fuzzy matcher required a match in the final candidate
segment. For example, to match the candidate "foo.bar", the input had
to have at least one character that matched "bar". I previously
removed this requirement as it is too restrictive for deep completions
to be useful.
However, there was still some lingering final-segment favoritism in
the matching algorithm. In particular, there were penalties for not
matching the final segment's first character and for not matching the
final segment's word initial characters. However, these penalties only
made sense when we also required a final segment match. Consider this
example:
User input: "U"
Candidate "ErrUnexpectedEOF" - with only a single segment, we got big
penalties for not matching the leading "E" (since it is the final
segment).
Candidate "ErrUnexpectedEOF.Error" - "ErrUnexpectedEOF" is no longer
the final segment, so we didn't get penalties. And we didn't get
penalties for the final segment "Error" because we finished matching
after the first "U". As a result, this candidate slips through with a
higher score.
Fix by simplifying the skip penalty. Now we only penalize for skipping
the first character of the first or final segment (and the penalty is
lower). For deep completions, the first and final segment are both
"important" segments, so I think it makes sense to focus on both of
them. We don't want to penalize all segment starts because that makes
it harder to match deeper candidates where you often "ignore"
intermediate segments.
I had to adjust a few scores in the tests, but I don't think the
impact will be too big other than fixing the bug.
Fixesgolang/go#35062.
Change-Id: Id17a5c80bf0f80ce252fe990ccfbd51c1bac1c72
Reviewed-on: https://go-review.googlesource.com/c/tools/+/202638
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Remove the input type option. Now everything behaves as "symbol".
We don't use the "text" or "filename" input types, and I don't foresee
us using them. Removing them simplifies the code a bit, but simplifies
the tests a lot. It was tedious to make changes to the matcher logic
because you had to fret over test failure details that didn't actually
matter because we didn't use that functionality.
Change-Id: I651debde9e63ee283d7bc3ad718d22f4b9a127c0
Reviewed-on: https://go-review.googlesource.com/c/tools/+/202637
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
For *ast.Ident completion requests, this checks the parent node to
see if the token begins a statement and then based on the path adds
possible keyword completion candidates. The test lists some cases where
this approach cannot provide completion candidates.
The biggest thing missing is keywords for file level declarations
Updates golang/go#34009
Change-Id: I9d9c0c1eb88e362613feca66d0eea6b88705b9b0
Reviewed-on: https://go-review.googlesource.com/c/tools/+/196664
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
This change allows us to hanel cancel messages as they go into the queue, and
cancel messages that are ahead of them in the queue but not being processed yet.
This should reduce the amount of redundant work that we do when we are handling
a cancel storm.
Change-Id: Id1a58991407d75b68d65bacf96350a4dd69d4d2b
Reviewed-on: https://go-review.googlesource.com/c/tools/+/200766
Run-TryBot: Ian Cottrell <iancottrell@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
A continuation of CL 202298, only for analysis errors.
Change-Id: I957d52cef31938ef66be73463e92695a5b56869c
Reviewed-on: https://go-review.googlesource.com/c/tools/+/202540
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Cottrell <iancottrell@google.com>
Since a user's module cache is generally going to be much bigger than
their main module, one would expect that caching just information about
the module cache would be sufficient. It turns out that's not correct.
When we discover something in the module cache, we have to make sure
that a different version of it isn't already in scope. Doing that can
require information about the main module or replace targets, so that
needs to be cached too.
Concretely, when I'm working in x/tools, if a scan discovers a version
of x/tools in the module cache, it should usually ignore that version.
But that might not be true in more complicated cases, especially those
involving nested modules whose boundaries change.
So, cache everything except GOROOT. Since the new data is mutable,
we store it separately from the module cache data so that it can be
discarded easily between runs.
Change-Id: I47364f6c0270fee03af8898fec6c85d1b9c8d780
Reviewed-on: https://go-review.googlesource.com/c/tools/+/202045
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Scan most sources, including GOPATH, the module cache, the main module,
and replace targets as appropriate. Use the cached stdlib instead of
scanning GOROOT.
We heavily cache the contents of the module cache, so performance is
decent. But we have to look at all the modules not in the module cache
too to get the right versions of modules (see
(*ModuleResolver).canonicalize), which currently isn't cached at all,
even just for a single run. That ends up being pretty expensive.
The implementation changes are relatively small; add package name
loading to scan(), cache that result, and allow callers to control what
directories are scanned so that it can skip GOROOT.
I also cleared out most of the stdlib from the unimported completion
test and added a simple external completion to it for safety's sake.
Change-Id: Id50fd4703b1126be35a000fe90719e19c3ab84bf
Reviewed-on: https://go-review.googlesource.com/c/tools/+/199178
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
