aboutsummaryrefslogtreecommitdiff
path: root/gopls/internal/lsp/source/format.go
diff options
context:
space:
mode:
Diffstat (limited to 'gopls/internal/lsp/source/format.go')
-rw-r--r--gopls/internal/lsp/source/format.go391
1 files changed, 391 insertions, 0 deletions
diff --git a/gopls/internal/lsp/source/format.go b/gopls/internal/lsp/source/format.go
new file mode 100644
index 000000000..b8206edc9
--- /dev/null
+++ b/gopls/internal/lsp/source/format.go
@@ -0,0 +1,391 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Package source provides core features for use by Go editors and tools.
+package source
+
+import (
+ "bytes"
+ "context"
+ "fmt"
+ "go/ast"
+ "go/format"
+ "go/parser"
+ "go/token"
+ "strings"
+ "text/scanner"
+
+ "golang.org/x/tools/gopls/internal/lsp/protocol"
+ "golang.org/x/tools/gopls/internal/lsp/safetoken"
+ "golang.org/x/tools/internal/diff"
+ "golang.org/x/tools/internal/event"
+ "golang.org/x/tools/internal/imports"
+)
+
+// Format formats a file with a given range.
+func Format(ctx context.Context, snapshot Snapshot, fh FileHandle) ([]protocol.TextEdit, error) {
+ ctx, done := event.Start(ctx, "source.Format")
+ defer done()
+
+ // Generated files shouldn't be edited. So, don't format them
+ if IsGenerated(ctx, snapshot, fh.URI()) {
+ return nil, fmt.Errorf("can't format %q: file is generated", fh.URI().Filename())
+ }
+
+ pgf, err := snapshot.ParseGo(ctx, fh, ParseFull)
+ if err != nil {
+ return nil, err
+ }
+ // Even if this file has parse errors, it might still be possible to format it.
+ // Using format.Node on an AST with errors may result in code being modified.
+ // Attempt to format the source of this file instead.
+ if pgf.ParseErr != nil {
+ formatted, err := formatSource(ctx, fh)
+ if err != nil {
+ return nil, err
+ }
+ return computeTextEdits(ctx, snapshot, pgf, string(formatted))
+ }
+
+ // format.Node changes slightly from one release to another, so the version
+ // of Go used to build the LSP server will determine how it formats code.
+ // This should be acceptable for all users, who likely be prompted to rebuild
+ // the LSP server on each Go release.
+ buf := &bytes.Buffer{}
+ fset := FileSetFor(pgf.Tok)
+ if err := format.Node(buf, fset, pgf.File); err != nil {
+ return nil, err
+ }
+ formatted := buf.String()
+
+ // Apply additional formatting, if any is supported. Currently, the only
+ // supported additional formatter is gofumpt.
+ if format := snapshot.View().Options().GofumptFormat; snapshot.View().Options().Gofumpt && format != nil {
+ // gofumpt can customize formatting based on language version and module
+ // path, if available.
+ //
+ // Try to derive this information, but fall-back on the default behavior.
+ //
+ // TODO: under which circumstances can we fail to find module information?
+ // Can this, for example, result in inconsistent formatting across saves,
+ // due to pending calls to packages.Load?
+ var langVersion, modulePath string
+ mds, err := snapshot.MetadataForFile(ctx, fh.URI())
+ if err == nil && len(mds) > 0 {
+ if mi := mds[0].Module; mi != nil {
+ langVersion = mi.GoVersion
+ modulePath = mi.Path
+ }
+ }
+ b, err := format(ctx, langVersion, modulePath, buf.Bytes())
+ if err != nil {
+ return nil, err
+ }
+ formatted = string(b)
+ }
+ return computeTextEdits(ctx, snapshot, pgf, formatted)
+}
+
+func formatSource(ctx context.Context, fh FileHandle) ([]byte, error) {
+ _, done := event.Start(ctx, "source.formatSource")
+ defer done()
+
+ data, err := fh.Read()
+ if err != nil {
+ return nil, err
+ }
+ return format.Source(data)
+}
+
+type ImportFix struct {
+ Fix *imports.ImportFix
+ Edits []protocol.TextEdit
+}
+
+// AllImportsFixes formats f for each possible fix to the imports.
+// In addition to returning the result of applying all edits,
+// it returns a list of fixes that could be applied to the file, with the
+// corresponding TextEdits that would be needed to apply that fix.
+func AllImportsFixes(ctx context.Context, snapshot Snapshot, fh FileHandle) (allFixEdits []protocol.TextEdit, editsPerFix []*ImportFix, err error) {
+ ctx, done := event.Start(ctx, "source.AllImportsFixes")
+ defer done()
+
+ pgf, err := snapshot.ParseGo(ctx, fh, ParseFull)
+ if err != nil {
+ return nil, nil, err
+ }
+ if err := snapshot.RunProcessEnvFunc(ctx, func(opts *imports.Options) error {
+ allFixEdits, editsPerFix, err = computeImportEdits(snapshot, pgf, opts)
+ return err
+ }); err != nil {
+ return nil, nil, fmt.Errorf("AllImportsFixes: %v", err)
+ }
+ return allFixEdits, editsPerFix, nil
+}
+
+// computeImportEdits computes a set of edits that perform one or all of the
+// necessary import fixes.
+func computeImportEdits(snapshot Snapshot, pgf *ParsedGoFile, options *imports.Options) (allFixEdits []protocol.TextEdit, editsPerFix []*ImportFix, err error) {
+ filename := pgf.URI.Filename()
+
+ // Build up basic information about the original file.
+ allFixes, err := imports.FixImports(filename, pgf.Src, options)
+ if err != nil {
+ return nil, nil, err
+ }
+
+ allFixEdits, err = computeFixEdits(snapshot, pgf, options, allFixes)
+ if err != nil {
+ return nil, nil, err
+ }
+
+ // Apply all of the import fixes to the file.
+ // Add the edits for each fix to the result.
+ for _, fix := range allFixes {
+ edits, err := computeFixEdits(snapshot, pgf, options, []*imports.ImportFix{fix})
+ if err != nil {
+ return nil, nil, err
+ }
+ editsPerFix = append(editsPerFix, &ImportFix{
+ Fix: fix,
+ Edits: edits,
+ })
+ }
+ return allFixEdits, editsPerFix, nil
+}
+
+// ComputeOneImportFixEdits returns text edits for a single import fix.
+func ComputeOneImportFixEdits(snapshot Snapshot, pgf *ParsedGoFile, fix *imports.ImportFix) ([]protocol.TextEdit, error) {
+ options := &imports.Options{
+ LocalPrefix: snapshot.View().Options().Local,
+ // Defaults.
+ AllErrors: true,
+ Comments: true,
+ Fragment: true,
+ FormatOnly: false,
+ TabIndent: true,
+ TabWidth: 8,
+ }
+ return computeFixEdits(snapshot, pgf, options, []*imports.ImportFix{fix})
+}
+
+func computeFixEdits(snapshot Snapshot, pgf *ParsedGoFile, options *imports.Options, fixes []*imports.ImportFix) ([]protocol.TextEdit, error) {
+ // trim the original data to match fixedData
+ left, err := importPrefix(pgf.Src)
+ if err != nil {
+ return nil, err
+ }
+ extra := !strings.Contains(left, "\n") // one line may have more than imports
+ if extra {
+ left = string(pgf.Src)
+ }
+ if len(left) > 0 && left[len(left)-1] != '\n' {
+ left += "\n"
+ }
+ // Apply the fixes and re-parse the file so that we can locate the
+ // new imports.
+ flags := parser.ImportsOnly
+ if extra {
+ // used all of origData above, use all of it here too
+ flags = 0
+ }
+ fixedData, err := imports.ApplyFixes(fixes, "", pgf.Src, options, flags)
+ if err != nil {
+ return nil, err
+ }
+ if fixedData == nil || fixedData[len(fixedData)-1] != '\n' {
+ fixedData = append(fixedData, '\n') // ApplyFixes may miss the newline, go figure.
+ }
+ edits := snapshot.View().Options().ComputeEdits(left, string(fixedData))
+ return protocolEditsFromSource([]byte(left), edits)
+}
+
+// importPrefix returns the prefix of the given file content through the final
+// import statement. If there are no imports, the prefix is the package
+// statement and any comment groups below it.
+func importPrefix(src []byte) (string, error) {
+ fset := token.NewFileSet()
+ // do as little parsing as possible
+ f, err := parser.ParseFile(fset, "", src, parser.ImportsOnly|parser.ParseComments)
+ if err != nil { // This can happen if 'package' is misspelled
+ return "", fmt.Errorf("importPrefix: failed to parse: %s", err)
+ }
+ tok := fset.File(f.Pos())
+ var importEnd int
+ for _, d := range f.Decls {
+ if x, ok := d.(*ast.GenDecl); ok && x.Tok == token.IMPORT {
+ if e, err := safetoken.Offset(tok, d.End()); err != nil {
+ return "", fmt.Errorf("importPrefix: %s", err)
+ } else if e > importEnd {
+ importEnd = e
+ }
+ }
+ }
+
+ maybeAdjustToLineEnd := func(pos token.Pos, isCommentNode bool) int {
+ offset, err := safetoken.Offset(tok, pos)
+ if err != nil {
+ return -1
+ }
+
+ // Don't go past the end of the file.
+ if offset > len(src) {
+ offset = len(src)
+ }
+ // The go/ast package does not account for different line endings, and
+ // specifically, in the text of a comment, it will strip out \r\n line
+ // endings in favor of \n. To account for these differences, we try to
+ // return a position on the next line whenever possible.
+ switch line := tok.Line(tok.Pos(offset)); {
+ case line < tok.LineCount():
+ nextLineOffset, err := safetoken.Offset(tok, tok.LineStart(line+1))
+ if err != nil {
+ return -1
+ }
+ // If we found a position that is at the end of a line, move the
+ // offset to the start of the next line.
+ if offset+1 == nextLineOffset {
+ offset = nextLineOffset
+ }
+ case isCommentNode, offset+1 == tok.Size():
+ // If the last line of the file is a comment, or we are at the end
+ // of the file, the prefix is the entire file.
+ offset = len(src)
+ }
+ return offset
+ }
+ if importEnd == 0 {
+ pkgEnd := f.Name.End()
+ importEnd = maybeAdjustToLineEnd(pkgEnd, false)
+ }
+ for _, cgroup := range f.Comments {
+ for _, c := range cgroup.List {
+ if end, err := safetoken.Offset(tok, c.End()); err != nil {
+ return "", err
+ } else if end > importEnd {
+ startLine := safetoken.Position(tok, c.Pos()).Line
+ endLine := safetoken.Position(tok, c.End()).Line
+
+ // Work around golang/go#41197 by checking if the comment might
+ // contain "\r", and if so, find the actual end position of the
+ // comment by scanning the content of the file.
+ startOffset, err := safetoken.Offset(tok, c.Pos())
+ if err != nil {
+ return "", err
+ }
+ if startLine != endLine && bytes.Contains(src[startOffset:], []byte("\r")) {
+ if commentEnd := scanForCommentEnd(src[startOffset:]); commentEnd > 0 {
+ end = startOffset + commentEnd
+ }
+ }
+ importEnd = maybeAdjustToLineEnd(tok.Pos(end), true)
+ }
+ }
+ }
+ if importEnd > len(src) {
+ importEnd = len(src)
+ }
+ return string(src[:importEnd]), nil
+}
+
+// scanForCommentEnd returns the offset of the end of the multi-line comment
+// at the start of the given byte slice.
+func scanForCommentEnd(src []byte) int {
+ var s scanner.Scanner
+ s.Init(bytes.NewReader(src))
+ s.Mode ^= scanner.SkipComments
+
+ t := s.Scan()
+ if t == scanner.Comment {
+ return s.Pos().Offset
+ }
+ return 0
+}
+
+func computeTextEdits(ctx context.Context, snapshot Snapshot, pgf *ParsedGoFile, formatted string) ([]protocol.TextEdit, error) {
+ _, done := event.Start(ctx, "source.computeTextEdits")
+ defer done()
+
+ edits := snapshot.View().Options().ComputeEdits(string(pgf.Src), formatted)
+ return ToProtocolEdits(pgf.Mapper, edits)
+}
+
+// protocolEditsFromSource converts text edits to LSP edits using the original
+// source.
+func protocolEditsFromSource(src []byte, edits []diff.Edit) ([]protocol.TextEdit, error) {
+ m := protocol.NewMapper("", src)
+ var result []protocol.TextEdit
+ for _, edit := range edits {
+ rng, err := m.OffsetRange(edit.Start, edit.End)
+ if err != nil {
+ return nil, err
+ }
+
+ if rng.Start == rng.End && edit.New == "" {
+ // Degenerate case, which may result from a diff tool wanting to delete
+ // '\r' in line endings. Filter it out.
+ continue
+ }
+ result = append(result, protocol.TextEdit{
+ Range: rng,
+ NewText: edit.New,
+ })
+ }
+ return result, nil
+}
+
+// ToProtocolEdits converts diff.Edits to LSP TextEdits.
+// See https://microsoft.github.io/language-server-protocol/specifications/lsp/3.17/specification/#textEditArray
+func ToProtocolEdits(m *protocol.Mapper, edits []diff.Edit) ([]protocol.TextEdit, error) {
+ // LSP doesn't require TextEditArray to be sorted:
+ // this is the receiver's concern. But govim, and perhaps
+ // other clients have historically relied on the order.
+ edits = append([]diff.Edit(nil), edits...)
+ diff.SortEdits(edits)
+
+ result := make([]protocol.TextEdit, len(edits))
+ for i, edit := range edits {
+ rng, err := m.OffsetRange(edit.Start, edit.End)
+ if err != nil {
+ return nil, err
+ }
+ result[i] = protocol.TextEdit{
+ Range: rng,
+ NewText: edit.New,
+ }
+ }
+ return result, nil
+}
+
+// FromProtocolEdits converts LSP TextEdits to diff.Edits.
+// See https://microsoft.github.io/language-server-protocol/specifications/lsp/3.17/specification/#textEditArray
+func FromProtocolEdits(m *protocol.Mapper, edits []protocol.TextEdit) ([]diff.Edit, error) {
+ if edits == nil {
+ return nil, nil
+ }
+ result := make([]diff.Edit, len(edits))
+ for i, edit := range edits {
+ start, end, err := m.RangeOffsets(edit.Range)
+ if err != nil {
+ return nil, err
+ }
+ result[i] = diff.Edit{
+ Start: start,
+ End: end,
+ New: edit.NewText,
+ }
+ }
+ return result, nil
+}
+
+// ApplyProtocolEdits applies the patch (edits) to m.Content and returns the result.
+// It also returns the edits converted to diff-package form.
+func ApplyProtocolEdits(m *protocol.Mapper, edits []protocol.TextEdit) ([]byte, []diff.Edit, error) {
+ diffEdits, err := FromProtocolEdits(m, edits)
+ if err != nil {
+ return nil, nil, err
+ }
+ out, err := diff.ApplyBytes(m.Content, diffEdits)
+ return out, diffEdits, err
+}