mirror of
https://github.com/golang/go
synced 2024-11-05 15:46:11 -07:00
ecd3fc4348
We use file identities pervasively throughout gopls. Prior to this change, the identity is the modification date of an unopened file, or the hash of an opened file. That means that opening a file changes its identity, which causes unnecessary churn in the cache. Unfortunately, there isn't an easy way to fix this. Changing the cache key to something else, such as the modification time, means that we won't unify cache entries if a change is made and then undone. The approach here is to read files eagerly in GetFile, so that we know their hashes immediately. That resolves the churn, but means that we do a ton of file IO at startup. Incidental changes: Remove the FileSystem interface; there was only one implementation and it added a fair amount of cruft. We have many other places that assume os.Stat and such work. Add direct accessors to FileHandle for URI, Kind, and Version. Most uses of (FileHandle).Identity were for stuff that we derive solely from the URI, and this helped me disentangle them. It is a *ton* of churn, though. I can revert it if you want. Change-Id: Ia2133bc527f71daf81c9d674951726a232ca5bc9 Reviewed-on: https://go-review.googlesource.com/c/tools/+/237037 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
203 lines
6.0 KiB
Go
203 lines
6.0 KiB
Go
package source
|
|
|
|
import (
|
|
"context"
|
|
"go/ast"
|
|
"go/token"
|
|
"sort"
|
|
|
|
"golang.org/x/tools/internal/lsp/protocol"
|
|
)
|
|
|
|
type FoldingRangeInfo struct {
|
|
mappedRange
|
|
Kind protocol.FoldingRangeKind
|
|
}
|
|
|
|
// FoldingRange gets all of the folding range for f.
|
|
func FoldingRange(ctx context.Context, snapshot Snapshot, fh FileHandle, lineFoldingOnly bool) (ranges []*FoldingRangeInfo, err error) {
|
|
// TODO(suzmue): consider limiting the number of folding ranges returned, and
|
|
// implement a way to prioritize folding ranges in that case.
|
|
pgh := snapshot.View().Session().Cache().ParseGoHandle(ctx, fh, ParseFull)
|
|
file, _, m, _, err := pgh.Parse(ctx)
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
fset := snapshot.View().Session().Cache().FileSet()
|
|
|
|
// Get folding ranges for comments separately as they are not walked by ast.Inspect.
|
|
ranges = append(ranges, commentsFoldingRange(fset, m, file)...)
|
|
|
|
foldingFunc := foldingRange
|
|
if lineFoldingOnly {
|
|
foldingFunc = lineFoldingRange
|
|
}
|
|
|
|
visit := func(n ast.Node) bool {
|
|
rng := foldingFunc(fset, m, n)
|
|
if rng != nil {
|
|
ranges = append(ranges, rng)
|
|
}
|
|
return true
|
|
}
|
|
// Walk the ast and collect folding ranges.
|
|
ast.Inspect(file, visit)
|
|
|
|
sort.Slice(ranges, func(i, j int) bool {
|
|
irng, _ := ranges[i].Range()
|
|
jrng, _ := ranges[j].Range()
|
|
return protocol.CompareRange(irng, jrng) < 0
|
|
})
|
|
|
|
return ranges, nil
|
|
}
|
|
|
|
// foldingRange calculates the folding range for n.
|
|
func foldingRange(fset *token.FileSet, m *protocol.ColumnMapper, n ast.Node) *FoldingRangeInfo {
|
|
var kind protocol.FoldingRangeKind
|
|
var start, end token.Pos
|
|
switch n := n.(type) {
|
|
case *ast.BlockStmt:
|
|
// Fold from position of "{" to position of "}".
|
|
start, end = n.Lbrace+1, n.Rbrace
|
|
case *ast.CaseClause:
|
|
// Fold from position of ":" to end.
|
|
start, end = n.Colon+1, n.End()
|
|
case *ast.CallExpr:
|
|
// Fold from position of "(" to position of ")".
|
|
start, end = n.Lparen+1, n.Rparen
|
|
case *ast.FieldList:
|
|
// Fold from position of opening parenthesis/brace, to position of
|
|
// closing parenthesis/brace.
|
|
start, end = n.Opening+1, n.Closing
|
|
case *ast.GenDecl:
|
|
// If this is an import declaration, set the kind to be protocol.Imports.
|
|
if n.Tok == token.IMPORT {
|
|
kind = protocol.Imports
|
|
}
|
|
start, end = n.Lparen+1, n.Rparen
|
|
case *ast.CompositeLit:
|
|
// Fold from position of "{" to position of "}".
|
|
start, end = n.Lbrace+1, n.Rbrace
|
|
}
|
|
if !start.IsValid() || !end.IsValid() {
|
|
return nil
|
|
}
|
|
return &FoldingRangeInfo{
|
|
mappedRange: newMappedRange(fset, m, start, end),
|
|
Kind: kind,
|
|
}
|
|
}
|
|
|
|
// lineFoldingRange calculates the line folding range for n.
|
|
func lineFoldingRange(fset *token.FileSet, m *protocol.ColumnMapper, n ast.Node) *FoldingRangeInfo {
|
|
|
|
// TODO(suzmue): include trailing empty lines before the closing
|
|
// parenthesis/brace.
|
|
var kind protocol.FoldingRangeKind
|
|
var start, end token.Pos
|
|
switch n := n.(type) {
|
|
case *ast.BlockStmt:
|
|
// Fold lines between "{" and "}".
|
|
if !n.Lbrace.IsValid() || !n.Rbrace.IsValid() {
|
|
break
|
|
}
|
|
nStmts := len(n.List)
|
|
if nStmts == 0 {
|
|
break
|
|
}
|
|
// Don't want to fold if the start is on the same line as the brace.
|
|
if fset.Position(n.Lbrace).Line == fset.Position(n.List[0].Pos()).Line {
|
|
break
|
|
}
|
|
// Don't want to fold if the end is on the same line as the brace.
|
|
if fset.Position(n.Rbrace).Line == fset.Position(n.List[nStmts-1].End()).Line {
|
|
break
|
|
}
|
|
start, end = n.Lbrace+1, n.List[nStmts-1].End()
|
|
case *ast.CaseClause:
|
|
// Fold from position of ":" to end.
|
|
start, end = n.Colon+1, n.End()
|
|
case *ast.FieldList:
|
|
// Fold lines between opening parenthesis/brace and closing parenthesis/brace.
|
|
if !n.Opening.IsValid() || !n.Closing.IsValid() {
|
|
break
|
|
}
|
|
nFields := len(n.List)
|
|
if nFields == 0 {
|
|
break
|
|
}
|
|
// Don't want to fold if the start is on the same line as the parenthesis/brace.
|
|
if fset.Position(n.Opening).Line == fset.Position(n.List[nFields-1].End()).Line {
|
|
break
|
|
}
|
|
// Don't want to fold if the end is on the same line as the parenthesis/brace.
|
|
if fset.Position(n.Closing).Line == fset.Position(n.List[nFields-1].End()).Line {
|
|
break
|
|
}
|
|
start, end = n.Opening+1, n.List[nFields-1].End()
|
|
case *ast.GenDecl:
|
|
// If this is an import declaration, set the kind to be protocol.Imports.
|
|
if n.Tok == token.IMPORT {
|
|
kind = protocol.Imports
|
|
}
|
|
// Fold from position of "(" to position of ")".
|
|
if !n.Lparen.IsValid() || !n.Rparen.IsValid() {
|
|
break
|
|
}
|
|
nSpecs := len(n.Specs)
|
|
if nSpecs == 0 {
|
|
break
|
|
}
|
|
// Don't want to fold if the end is on the same line as the parenthesis/brace.
|
|
if fset.Position(n.Lparen).Line == fset.Position(n.Specs[0].Pos()).Line {
|
|
break
|
|
}
|
|
// Don't want to fold if the end is on the same line as the parenthesis/brace.
|
|
if fset.Position(n.Rparen).Line == fset.Position(n.Specs[nSpecs-1].End()).Line {
|
|
break
|
|
}
|
|
start, end = n.Lparen+1, n.Specs[nSpecs-1].End()
|
|
case *ast.CompositeLit:
|
|
// Fold lines between "{" and "}".
|
|
if !n.Lbrace.IsValid() || !n.Rbrace.IsValid() {
|
|
break
|
|
}
|
|
if len(n.Elts) == 0 {
|
|
break
|
|
}
|
|
start, end = n.Lbrace+1, n.Elts[len(n.Elts)-1].End()
|
|
}
|
|
|
|
// Check that folding positions are valid.
|
|
if !start.IsValid() || !end.IsValid() {
|
|
return nil
|
|
}
|
|
// Do not fold if the start and end lines are the same.
|
|
if fset.Position(start).Line == fset.Position(end).Line {
|
|
return nil
|
|
}
|
|
return &FoldingRangeInfo{
|
|
mappedRange: newMappedRange(fset, m, start, end),
|
|
Kind: kind,
|
|
}
|
|
}
|
|
|
|
// commentsFoldingRange returns the folding ranges for all comment blocks in file.
|
|
// The folding range starts at the end of the first comment, and ends at the end of the
|
|
// comment block and has kind protocol.Comment.
|
|
func commentsFoldingRange(fset *token.FileSet, m *protocol.ColumnMapper, file *ast.File) (comments []*FoldingRangeInfo) {
|
|
for _, commentGrp := range file.Comments {
|
|
// Don't fold single comments.
|
|
if len(commentGrp.List) <= 1 {
|
|
continue
|
|
}
|
|
comments = append(comments, &FoldingRangeInfo{
|
|
// Fold from the end of the first line comment to the end of the comment block.
|
|
mappedRange: newMappedRange(fset, m, commentGrp.List[0].End(), commentGrp.End()),
|
|
Kind: protocol.Comment,
|
|
})
|
|
}
|
|
return comments
|
|
}
|