Files
lukaszraczylo 9af2801b1b refactor(edit): remove auto-indentation and add line-ending normalization
- [x] Remove auto-indentation from text mode edits (caller controls whitespace)
- [x] Add line-ending detection and normalization for both AST and text modes
- [x] Share edit logic via new `spliceContent` function for both modes
- [x] Fix diff to emit "No newline at end of file" markers
- [x] Fix diff to strip raw CR from CRLF file output
- [x] Remove double-unescape of backslash sequences in new_content
- [x] Fix countDiffLines to be hunk-aware (correctly count lines starting with +/-)
- [x] Fix block-comment stripping to remove standalone lines cleanly
- [x] Fix Python license header stripping to preserve separator blank lines
2026-05-29 00:17:36 +01:00

880 lines
27 KiB
Go

// Package edit provides AST-aware file editing capabilities.
package edit
import (
"bytes"
"context"
"fmt"
"os"
"slices"
"strings"
"sync"
"github.com/lukaszraczylo/mcp-filepuff/internal/parser"
"github.com/lukaszraczylo/mcp-filepuff/internal/util"
"github.com/lukaszraczylo/mcp-filepuff/pkg/errors"
"github.com/lukaszraczylo/mcp-filepuff/pkg/protocol"
"github.com/sergi/go-diff/diffmatchpatch"
sitter "github.com/smacker/go-tree-sitter"
)
// EditOperation defines the type of edit operation.
type EditOperation string
const (
EditReplace EditOperation = "replace"
EditInsertBefore EditOperation = "insert_before"
EditInsertAfter EditOperation = "insert_after"
EditDelete EditOperation = "delete"
)
// ASTEdit represents an AST-aware edit request.
type ASTEdit struct {
File string `json:"file"`
Operation EditOperation `json:"operation"`
NewContent string `json:"new_content,omitempty"`
Selector ASTSelector `json:"selector"`
}
// ASTSelector specifies how to find the target node.
type ASTSelector struct {
Kind string `json:"kind,omitempty"`
Name string `json:"name,omitempty"`
Pattern string `json:"pattern,omitempty"`
Text string `json:"text,omitempty"`
TextPattern string `json:"text_pattern,omitempty"`
AtLine int `json:"at_line,omitempty"`
Index int `json:"index,omitempty"`
LineEnd int `json:"line_end,omitempty"`
}
// EditResult contains the result of an edit operation.
type EditResult struct {
Diff string `json:"diff,omitempty"`
OriginalContent string `json:"original_content,omitempty"`
NewContent string `json:"new_content,omitempty"`
Error string `json:"error,omitempty"`
Success bool `json:"success"`
Applied bool `json:"applied"`
}
// Engine performs AST-aware edits.
type Engine struct {
registry *parser.Registry
dmp *diffmatchpatch.DiffMatchPatch
fileLocks sync.Map // map[string]*sync.Mutex for per-file locking
}
// NewEngine creates a new edit engine.
func NewEngine(registry *parser.Registry) *Engine {
return &Engine{
registry: registry,
dmp: diffmatchpatch.New(),
fileLocks: sync.Map{},
}
}
// lockFile acquires a lock for the specified file and returns an unlock function.
// This prevents concurrent edits to the same file which could cause corruption.
func (e *Engine) lockFile(filePath string) func() {
// Get or create mutex for this file
actual, _ := e.fileLocks.LoadOrStore(filePath, &sync.Mutex{})
mu := actual.(*sync.Mutex)
mu.Lock()
return mu.Unlock
}
// Preview generates a preview of an edit without applying it.
func (e *Engine) Preview(ctx context.Context, edit *ASTEdit) (*EditResult, error) {
return e.performEdit(ctx, edit, false)
}
// Apply performs an edit and writes the result to disk.
// Uses file locking to prevent concurrent edits to the same file.
func (e *Engine) Apply(ctx context.Context, edit *ASTEdit) (*EditResult, error) {
unlock := e.lockFile(edit.File)
defer unlock()
return e.performEdit(ctx, edit, true)
}
// performEdit executes an edit operation.
func (e *Engine) performEdit(ctx context.Context, edit *ASTEdit, apply bool) (*EditResult, error) {
// Determine if we should use text mode
useTextMode := e.shouldUseTextMode(edit)
if useTextMode {
return e.performTextEdit(ctx, edit, apply)
}
return e.performASTEdit(ctx, edit, apply)
}
// shouldUseTextMode determines if text-based editing should be used.
func (e *Engine) shouldUseTextMode(edit *ASTEdit) bool {
// Use text mode if text-specific selectors are provided
if edit.Selector.Text != "" || edit.Selector.TextPattern != "" {
return true
}
// Use text mode if line range is specified without AST selectors
if edit.Selector.AtLine > 0 && edit.Selector.LineEnd > 0 &&
edit.Selector.Kind == "" && edit.Selector.Name == "" && edit.Selector.Pattern == "" {
return true
}
// Use text mode if language is not supported for AST
lang := protocol.DetectLanguage(edit.File)
return lang == protocol.LangUnknown
}
// performASTEdit executes an AST-aware edit operation.
func (e *Engine) performASTEdit(ctx context.Context, edit *ASTEdit, apply bool) (*EditResult, error) {
// Validate operation
if err := e.validateASTEdit(edit); err != nil {
return &EditResult{Success: false, Error: err.Error()}, nil
}
// Read file
content, err := os.ReadFile(edit.File)
if err != nil {
structuredErr := errors.NewFileNotReadableError(edit.File, err)
return &EditResult{Success: false, Error: structuredErr.Error()}, nil
}
// Parse file
parseResult, err := e.registry.Parse(ctx, edit.File, content)
if err != nil {
return &EditResult{Success: false, Error: err.Error()}, nil
}
// Find target node
node, err := e.resolveSelector(edit.Selector, parseResult.Tree, content)
if err != nil {
return &EditResult{Success: false, Error: err.Error()}, nil
}
// Apply edit
newContent, err := e.applyEdit(edit, node, content)
if err != nil {
return &EditResult{Success: false, Error: err.Error()}, nil
}
// Validate new content (re-parse)
_, err = e.registry.Parse(ctx, edit.File, newContent)
if err != nil {
structuredErr := errors.NewEditValidationError(edit.File, err)
return &EditResult{
Success: false,
Error: structuredErr.Error(),
}, nil
}
// Generate diff
diff := e.generateDiff(string(content), string(newContent), edit.File)
result := &EditResult{
Success: true,
Diff: diff,
Applied: false,
}
// Apply changes if requested
if apply {
// Preserve original file permissions
fileInfo, err := os.Stat(edit.File)
perm := os.FileMode(0o600) // default fallback
if err == nil {
perm = fileInfo.Mode().Perm()
}
if err := os.WriteFile(edit.File, newContent, perm); err != nil {
structuredErr := errors.NewFileNotWritableError(edit.File, err)
return &EditResult{
Success: false,
Error: structuredErr.Error(),
}, nil
}
result.Applied = true
}
return result, nil
}
// performTextEdit executes a text-based edit operation for non-AST files.
func (e *Engine) performTextEdit(_ context.Context, edit *ASTEdit, apply bool) (*EditResult, error) {
// Validate operation
if err := e.validateTextEdit(edit); err != nil {
return &EditResult{Success: false, Error: err.Error()}, nil
}
// Read file
content, err := os.ReadFile(edit.File)
if err != nil {
structuredErr := errors.NewFileNotReadableError(edit.File, err)
return &EditResult{Success: false, Error: structuredErr.Error()}, nil
}
// Find the text selection (byte range)
start, end, err := e.resolveTextSelector(edit.Selector, content)
if err != nil {
return &EditResult{Success: false, Error: err.Error()}, nil
}
// Apply edit
newContent, err := e.applyTextEditOperation(edit.Operation, content, start, end, edit.NewContent)
if err != nil {
return &EditResult{Success: false, Error: err.Error()}, nil
}
// Generate diff
diff := e.generateDiff(string(content), string(newContent), edit.File)
result := &EditResult{
Success: true,
Diff: diff,
Applied: false,
}
// Apply changes if requested
if apply {
// Preserve original file permissions
fileInfo, err := os.Stat(edit.File)
perm := os.FileMode(0o600) // default fallback
if err == nil {
perm = fileInfo.Mode().Perm()
}
if err := os.WriteFile(edit.File, newContent, perm); err != nil {
structuredErr := errors.NewFileNotWritableError(edit.File, err)
return &EditResult{
Success: false,
Error: structuredErr.Error(),
}, nil
}
result.Applied = true
}
return result, nil
}
// validateBaseEdit checks common edit request fields.
func (e *Engine) validateBaseEdit(edit *ASTEdit) error {
if edit.File == "" {
return errors.NewInvalidEditError("file is required")
}
if edit.Operation == "" {
return errors.NewInvalidEditError("operation is required")
}
// Validate operation type
switch edit.Operation {
case EditReplace, EditInsertBefore, EditInsertAfter:
if edit.NewContent == "" {
return errors.NewInvalidEditError(fmt.Sprintf("new_content is required for %s operation", edit.Operation))
}
case EditDelete:
// new_content not required
default:
return errors.NewInvalidEditError(fmt.Sprintf("unknown operation: %s", edit.Operation))
}
return nil
}
// validateASTEdit checks if an AST edit request is valid.
func (e *Engine) validateASTEdit(edit *ASTEdit) error {
if err := e.validateBaseEdit(edit); err != nil {
return err
}
// Validate AST selector
if edit.Selector.Kind == "" && edit.Selector.Name == "" && edit.Selector.Pattern == "" && edit.Selector.AtLine == 0 {
return errors.NewInvalidEditError("AST selector must specify at least one of: kind, name, pattern, or at_line")
}
return nil
}
// validateTextEdit checks if a text edit request is valid.
func (e *Engine) validateTextEdit(edit *ASTEdit) error {
if err := e.validateBaseEdit(edit); err != nil {
return err
}
// Validate text selector - need at least one text selection method
hasTextSelector := edit.Selector.Text != "" ||
edit.Selector.TextPattern != "" ||
edit.Selector.AtLine > 0
if !hasTextSelector {
return errors.NewInvalidEditError("text selector must specify at least one of: text, text_pattern, or at_line")
}
// Validate regex pattern if provided (uses cached compilation)
if edit.Selector.TextPattern != "" {
if _, err := util.CompileRegex(edit.Selector.TextPattern); err != nil {
return errors.Wrap(errors.ErrInvalidEdit, "invalid text_pattern regex", err)
}
}
return nil
}
// resolveSelector finds the target node based on the selector.
func (e *Engine) resolveSelector(sel ASTSelector, tree *sitter.Tree, content []byte) (*sitter.Node, error) {
if tree == nil {
return nil, errors.NewNodeNotFoundError("no AST tree available")
}
root := tree.RootNode()
if root == nil {
return nil, errors.NewNodeNotFoundError("empty AST tree")
}
var matches []*sitter.Node
parser.WalkTree(root, func(n *sitter.Node) bool {
if e.matchesSelector(sel, n, content) {
matches = append(matches, n)
}
return true
})
if len(matches) == 0 {
selectorDesc := fmt.Sprintf("kind=%s name=%s pattern=%s line=%d", sel.Kind, sel.Name, sel.Pattern, sel.AtLine)
return nil, errors.NewNodeNotFoundError(selectorDesc)
}
// When using AtLine without a specific Kind, prefer the smallest (most specific) node.
// This prevents matching large parent nodes like source_file when we want a specific declaration.
if sel.AtLine > 0 && sel.Kind == "" {
matches = sortBySpecificity(matches)
}
// Use index to select specific match
index := sel.Index
if index < 0 || index >= len(matches) {
return nil, errors.NewInvalidSelectionError(fmt.Sprintf("selector matched %d nodes, but index %d is out of range", len(matches), index))
}
return matches[index], nil
}
// sortBySpecificity sorts nodes so that the most useful nodes come first.
// Prefers: 1) Named nodes (declarations/statements) over anonymous tokens
// 2) Smaller nodes over larger ones (more specific)
func sortBySpecificity(nodes []*sitter.Node) []*sitter.Node {
if len(nodes) <= 1 {
return nodes
}
result := make([]*sitter.Node, len(nodes))
copy(result, nodes)
slices.SortFunc(result, func(a, b *sitter.Node) int {
if shouldPrefer(a, b) {
return -1
}
if shouldPrefer(b, a) {
return 1
}
return 0
})
return result
}
// shouldPrefer returns true if node a should come before node b.
func shouldPrefer(a, b *sitter.Node) bool {
// Prefer named nodes over anonymous tokens
aIsNamed := a.IsNamed()
bIsNamed := b.IsNamed()
if aIsNamed && !bIsNamed {
return true
}
if !aIsNamed && bIsNamed {
return false
}
// Both named or both anonymous: prefer smaller meaningful nodes
// But filter out very small nodes (likely just identifiers/literals)
aSize := a.EndByte() - a.StartByte()
bSize := b.EndByte() - b.StartByte()
// If both are named, prefer "declaration" or "statement" types
aIsDecl := isDeclarationLike(a.Type())
bIsDecl := isDeclarationLike(b.Type())
if aIsDecl && !bIsDecl {
return true
}
if !aIsDecl && bIsDecl {
return false
}
// Same category: prefer smaller
return aSize < bSize
}
// isDeclarationLike returns true for node types that represent declarations or statements.
func isDeclarationLike(nodeType string) bool {
// Common declaration/statement patterns across languages
declarationPatterns := []string{
"declaration", "definition", "statement", "spec", "clause",
"function", "method", "class", "struct", "interface", "type",
"import", "package", "module", "const", "var", "let",
}
for _, pattern := range declarationPatterns {
if strings.Contains(nodeType, pattern) {
return true
}
}
return false
}
// matchesSelector checks if a node matches the selector criteria.
func (e *Engine) matchesSelector(sel ASTSelector, n *sitter.Node, content []byte) bool {
// Check kind
if sel.Kind != "" && n.Type() != sel.Kind {
return false
}
// Check name (look for identifier in the node)
if sel.Name != "" {
nameNode := n.ChildByFieldName("name")
if nameNode == nil {
// Also try to find an identifier child
found := false
for i := 0; i < int(n.NamedChildCount()); i++ {
child := n.NamedChild(i)
if child != nil && child.Type() == "identifier" {
if parser.GetNodeText(child, content) == sel.Name {
found = true
break
}
}
}
if !found {
return false
}
} else if parser.GetNodeText(nameNode, content) != sel.Name {
return false
}
}
// Check line
if sel.AtLine > 0 {
startLine := int(n.StartPoint().Row) + 1
endLine := int(n.EndPoint().Row) + 1
if sel.AtLine < startLine || sel.AtLine > endLine {
return false
}
}
// Pattern matching is handled separately (simplified here)
if sel.Pattern != "" {
nodeText := parser.GetNodeText(n, content)
if !strings.Contains(nodeText, sel.Pattern) {
return false
}
}
return true
}
// applyEdit applies an AST-mode edit. new_content is inserted verbatim — the caller
// controls indentation — and only line endings are normalized to the file's convention.
func (e *Engine) applyEdit(edit *ASTEdit, node *sitter.Node, content []byte) ([]byte, error) {
eol := detectLineEnding(content)
newContent := normalizeLineEndings(edit.NewContent, eol)
return spliceContent(edit.Operation, content, int(node.StartByte()), int(node.EndByte()), newContent, eol)
}
// detectLineEnding reports the file's dominant line-ending convention: "\r\n" when any
// CRLF terminator is present, otherwise "\n".
func detectLineEnding(content []byte) string {
if bytes.Contains(content, []byte("\r\n")) {
return "\r\n"
}
return "\n"
}
// normalizeLineEndings rewrites every line ending in s to eol. It first collapses CRLF to
// LF, then expands to the target, so mixed input becomes uniform and new_content can never
// introduce a line ending foreign to the file being edited.
func normalizeLineEndings(s, eol string) string {
if s == "" {
return s
}
s = strings.ReplaceAll(s, "\r\n", "\n")
if eol != "\n" {
s = strings.ReplaceAll(s, "\n", eol)
}
return s
}
func endsWithNewline(s string) bool { return strings.HasSuffix(s, "\n") }
func startsWithNewline(s string) bool {
return s != "" && (s[0] == '\n' || s[0] == '\r')
}
// spliceContent applies an edit operation by splicing newContent into content over the
// byte range [start, end). It is shared by AST and text modes — once auto-indentation is
// removed the two are identical. Restored terminators and separators use eol so the
// file's line-ending convention is preserved.
func spliceContent(op EditOperation, content []byte, start, end int, newContent, eol string) ([]byte, error) {
// A line-based selection on a CRLF file can land `end` between the \r (treated as
// line content) and the \n of a terminator. Pull it back so the full \r\n stays
// intact in the tail and is never split into a bare LF.
if end > start && end < len(content) && content[end-1] == '\r' && content[end] == '\n' {
end--
}
var result []byte
switch op {
case EditReplace:
result = append(result, content[:start]...)
result = append(result, newContent...)
// Restore a line terminator if the replaced range ended with one but the
// replacement does not, to prevent merging with the following line.
if end > start && content[end-1] == '\n' && !endsWithNewline(newContent) {
result = append(result, eol...)
}
result = append(result, content[end:]...)
case EditInsertBefore:
insertion := newContent
if !endsWithNewline(insertion) {
insertion += eol
}
result = append(result, content[:start]...)
result = append(result, insertion...)
result = append(result, content[start:]...)
case EditInsertAfter:
insertion := newContent
// Separate from preceding content.
if end > 0 && content[end-1] != '\n' && !startsWithNewline(insertion) {
insertion = eol + insertion
}
// Separate from following content.
if !endsWithNewline(insertion) && end < len(content) && content[end] != '\n' {
insertion += eol
}
result = append(result, content[:end]...)
result = append(result, insertion...)
result = append(result, content[end:]...)
case EditDelete:
result = append(result, content[:start]...)
result = append(result, content[end:]...)
default:
return nil, errors.NewInvalidEditError(fmt.Sprintf("unknown operation: %s", op))
}
return result, nil
}
// noNewlineMarker is the git-style annotation emitted after a diff line whose source
// version has no trailing newline.
const noNewlineMarker = "\\ No newline at end of file\n"
// diffLine represents a single line in the diff with its type and content.
type diffLine struct {
op diffmatchpatch.Operation
text string // line content without trailing newline
oldN int // 1-based line number in original (0 if insert)
newN int // 1-based line number in modified (0 if delete)
}
// indexRange is an inclusive [start, end] range of diffLine indices forming one hunk.
type indexRange struct{ start, end int }
// generateDiff creates a unified diff between original and modified content.
// Uses a line-level Myers diff and outputs a unified diff with 3 lines of context
// before/after each change, merging close hunks.
func (e *Engine) generateDiff(original, modified, filename string) string {
dmp := e.dmp
// Line-level diffing: encode each line as a single rune, diff the encoded strings,
// then decode back to real lines.
chars1, chars2, lineArray := dmp.DiffLinesToChars(original, modified)
diffs := dmp.DiffCharsToLines(dmp.DiffMain(chars1, chars2, false), lineArray)
diffs = dmp.DiffCleanupSemantic(diffs)
// Track whether each version lacks a final newline, so the diff is annotated
// git-style ("\ No newline at end of file") instead of implying a phantom one.
origNoEOL := len(original) > 0 && !strings.HasSuffix(original, "\n")
modNoEOL := len(modified) > 0 && !strings.HasSuffix(modified, "\n")
lines, maxOldN, maxNewN := flattenDiffLines(diffs)
ranges := diffHunkRanges(lines)
if len(ranges) == 0 {
return "" // no changes
}
var buf bytes.Buffer
fmt.Fprintf(&buf, "--- %s\n", filename)
fmt.Fprintf(&buf, "+++ %s\n", filename)
for _, r := range ranges {
oldStart, oldCount, newStart, newCount := hunkBounds(lines, r.start, r.end)
fmt.Fprintf(&buf, "@@ -%d,%d +%d,%d @@\n", oldStart, oldCount, newStart, newCount)
writeDiffBody(&buf, lines, r.start, r.end, origNoEOL, modNoEOL, maxOldN, maxNewN)
}
return buf.String()
}
// flattenDiffLines expands diff segments into per-line records with 1-based line numbers,
// returning the lines plus the final line number of each version (for no-newline marking).
func flattenDiffLines(diffs []diffmatchpatch.Diff) (lines []diffLine, maxOldN, maxNewN int) {
oldLine, newLine := 1, 1
for _, d := range diffs {
for _, raw := range strings.SplitAfter(d.Text, "\n") {
if raw == "" {
continue
}
// Strip the terminator for display; also drop a trailing CR so CRLF files
// do not leak raw carriage returns into the rendered diff.
text := strings.TrimSuffix(strings.TrimSuffix(raw, "\n"), "\r")
switch d.Type {
case diffmatchpatch.DiffEqual:
lines = append(lines, diffLine{op: d.Type, text: text, oldN: oldLine, newN: newLine})
oldLine++
newLine++
case diffmatchpatch.DiffDelete:
lines = append(lines, diffLine{op: d.Type, text: text, oldN: oldLine})
oldLine++
case diffmatchpatch.DiffInsert:
lines = append(lines, diffLine{op: d.Type, text: text, newN: newLine})
newLine++
}
}
}
return lines, oldLine - 1, newLine - 1
}
// diffHunkRanges returns the inclusive index ranges to emit: each changed line padded by
// 3 lines of context, with overlapping/adjacent ranges merged.
func diffHunkRanges(lines []diffLine) []indexRange {
const contextSize = 3
var ranges []indexRange
for i, l := range lines {
if l.op == diffmatchpatch.DiffEqual {
continue
}
rStart := max(i-contextSize, 0)
rEnd := min(i+contextSize, len(lines)-1)
if len(ranges) > 0 && rStart <= ranges[len(ranges)-1].end+1 {
ranges[len(ranges)-1].end = rEnd // merge with previous
} else {
ranges = append(ranges, indexRange{rStart, rEnd})
}
}
return ranges
}
// hunkBounds computes the unified-diff hunk header line numbers and counts for
// lines[start:end+1]. newStart/oldStart for one-sided lines are approximate.
func hunkBounds(lines []diffLine, start, end int) (oldStart, oldCount, newStart, newCount int) {
for i := start; i <= end; i++ {
l := lines[i]
switch l.op {
case diffmatchpatch.DiffEqual:
if oldCount == 0 {
oldStart = l.oldN
}
if newCount == 0 {
newStart = l.newN
}
oldCount++
newCount++
case diffmatchpatch.DiffDelete:
if oldCount == 0 {
oldStart = l.oldN
}
if newCount == 0 {
newStart = l.oldN // approximate
}
oldCount++
case diffmatchpatch.DiffInsert:
if newCount == 0 {
newStart = l.newN
}
if oldCount == 0 {
oldStart = l.newN // approximate
}
newCount++
}
}
return
}
// writeDiffBody writes the space/-/+ body lines for one hunk, appending the git-style
// no-newline marker after the final line of any version that lacks a trailing newline.
func writeDiffBody(buf *bytes.Buffer, lines []diffLine, start, end int, origNoEOL, modNoEOL bool, maxOldN, maxNewN int) {
for i := start; i <= end; i++ {
l := lines[i]
switch l.op {
case diffmatchpatch.DiffEqual:
fmt.Fprintf(buf, " %s\n", l.text)
if (origNoEOL && l.oldN == maxOldN) || (modNoEOL && l.newN == maxNewN) {
buf.WriteString(noNewlineMarker)
}
case diffmatchpatch.DiffDelete:
fmt.Fprintf(buf, "-%s\n", l.text)
if origNoEOL && l.oldN == maxOldN {
buf.WriteString(noNewlineMarker)
}
case diffmatchpatch.DiffInsert:
fmt.Fprintf(buf, "+%s\n", l.text)
if modNoEOL && l.newN == maxNewN {
buf.WriteString(noNewlineMarker)
}
}
}
}
// resolveTextSelector finds the byte range for a text-based selection.
func (e *Engine) resolveTextSelector(sel ASTSelector, content []byte) (start, end int, err error) {
switch {
case sel.Text != "":
return e.findExactText(content, sel.Text, sel.Index)
case sel.TextPattern != "":
return e.findRegexPattern(content, sel.TextPattern, sel.Index)
case sel.AtLine > 0:
return e.findLineRange(content, sel.AtLine, sel.LineEnd)
default:
return 0, 0, errors.NewInvalidEditError("text selector requires text, text_pattern, or at_line")
}
}
// findExactText finds an exact text match in content.
func (e *Engine) findExactText(content []byte, text string, index int) (start, end int, err error) {
if text == "" {
return 0, 0, errors.NewInvalidEditError("text selector cannot be empty")
}
textBytes := []byte(text)
type match struct{ start, end int }
var matches []match
offset := 0
for {
idx := bytes.Index(content[offset:], textBytes)
if idx == -1 {
break
}
matches = append(matches, match{
start: offset + idx,
end: offset + idx + len(textBytes),
})
offset += idx + 1
}
if len(matches) == 0 {
return 0, 0, errors.NewInvalidSelectionError(fmt.Sprintf("text not found: %q", truncateString(text, 50)))
}
if index >= len(matches) {
return 0, 0, errors.NewInvalidSelectionError(fmt.Sprintf("selector_index %d out of range (found %d matches)", index, len(matches)))
}
return matches[index].start, matches[index].end, nil
}
// findRegexPattern finds a regex pattern match in content.
func (e *Engine) findRegexPattern(content []byte, pattern string, index int) (start, end int, err error) {
re, err := util.CompileRegex(pattern)
if err != nil {
return 0, 0, errors.Wrap(errors.ErrInvalidEdit, "invalid regex pattern", err)
}
matches := re.FindAllIndex(content, -1)
if len(matches) == 0 {
return 0, 0, errors.NewInvalidSelectionError(fmt.Sprintf("pattern not found: %q", truncateString(pattern, 50)))
}
if index >= len(matches) {
return 0, 0, errors.NewInvalidSelectionError(fmt.Sprintf("selector_index %d out of range (found %d matches)", index, len(matches)))
}
return matches[index][0], matches[index][1], nil
}
// findLineRange finds the byte range for a line range selection.
func (e *Engine) findLineRange(content []byte, lineStart, lineEnd int) (start, end int, err error) {
if lineEnd == 0 {
lineEnd = lineStart
}
if lineStart < 1 {
return 0, 0, errors.NewInvalidEditError(fmt.Sprintf("line number must be >= 1, got %d", lineStart))
}
if lineEnd < lineStart {
return 0, 0, errors.NewInvalidEditError(fmt.Sprintf("line_end (%d) must be >= line (%d)", lineEnd, lineStart))
}
lines := bytes.Split(content, []byte("\n"))
// Trim phantom empty element from trailing newline
if len(lines) > 0 && len(lines[len(lines)-1]) == 0 {
lines = lines[:len(lines)-1]
}
totalLines := len(lines)
// Convert to 0-indexed
startIdx := lineStart - 1
endIdx := lineEnd - 1
if startIdx >= totalLines {
return 0, 0, errors.NewInvalidSelectionError(fmt.Sprintf("line %d out of range (file has %d lines)", lineStart, totalLines))
}
if endIdx >= totalLines {
return 0, 0, errors.NewInvalidSelectionError(fmt.Sprintf("line_end %d out of range (file has %d lines)", lineEnd, totalLines))
}
// Calculate byte positions
start = 0
for i := range startIdx {
start += len(lines[i]) + 1 // +1 for newline
}
end = start
for i := startIdx; i <= endIdx; i++ {
end += len(lines[i])
if i < totalLines-1 {
end += 1 // newline
}
}
return start, end, nil
}
// applyTextEditOperation applies a text-mode edit. Like AST mode, new_content is inserted
// verbatim (no auto-indentation) with its line endings normalized to the file's convention.
func (e *Engine) applyTextEditOperation(op EditOperation, content []byte, start, end int, newContent string) ([]byte, error) {
eol := detectLineEnding(content)
return spliceContent(op, content, start, end, normalizeLineEndings(newContent, eol), eol)
}
// truncateString truncates a string to maxLen with ellipsis.
func truncateString(s string, maxLen int) string {
if len(s) <= maxLen {
return s
}
return s[:maxLen-3] + "..."
}
// ValidateLanguage checks if AST editing is supported for a file.
// Returns nil for supported languages, error for unsupported.
// Note: Text-based editing is always available regardless of this check.
func ValidateLanguage(filename string) error {
lang := protocol.DetectLanguage(filename)
if lang == protocol.LangUnknown {
return fmt.Errorf("unsupported file type for AST editing: %s (text-based editing is available)", filename)
}
return nil
}