Add Grep text binary override

Author: SqlRush

docs/claude-code-go-rewrite-plan.md

@@ -207,6 +207,7 @@ test/parity/                 # golden tests against TS/official behavior
 - 本轮补充：Grep 文件列表输出现在显式支持 `files_with_match(es)`/`filesWithMatch(es)`/`files-with-match(es)`/`--files-with-match(es)`/`-l`，并接受 `output_mode` 的 `files_with_match` alias，统一归一为 `files_with_matches`。
 - 本轮补充：Grep 路径过滤现在支持 ripgrep 风格 `--glob`/`-g` 和 `--type`/`-t` aliases，执行和 structured content 都统一使用归一化后的 glob/type 过滤值。
 - 本轮补充：Grep `glob`/`--glob`/`-g` 路径过滤现在支持 ripgrep 风格 `!pattern` 排除规则，可与正向 glob、逗号/空白多 pattern 和 brace alternation 组合；只有排除规则时默认包含未被排除的路径。
+- 本轮补充：Grep 搜索现在支持 ripgrep 风格 `text`/`--text`/`-a`，可显式把二进制扩展名文件按文本读取参与匹配，structured content 会回传 `text` 状态；`-a` 同样兼容 quoted semantic boolean。
 - 本轮补充：Grep 常用布尔参数继续补齐 ripgrep 长参数 aliases，覆盖 `--line-number`、`--ignore-case`、`--fixed-strings`、`--word-regexp`、`--invert-match` 和 `--only-matching`，并兼容 quoted semantic boolean。
 - 本轮补充：Grep multiline 搜索现在支持 ripgrep 风格 `-U`、`--multiline`、`multiline-dotall` 和 `--multiline-dotall` aliases，统一映射到既有跨行 dotall 匹配逻辑并兼容 quoted semantic boolean。
 - 本轮补充：Grep 搜索现在支持 `no_ignore`/`noIgnore`/`no-ignore`/`--no-ignore`，可跳过 `.gitignore`/`.ignore` 规则，同时继续排除 VCS metadata 目录并保留 Read deny 额外 ignore 保护；`--no-ignore` 兼容 quoted boolean。

docs/first-second-parity-audit.md

@@ -134,6 +134,7 @@ Covered behavior:
 - `Grep` files-with-matches output now mirrors the official modified-time sort: newest files first, path tie-breaker, and pagination applied after sorting.
 - `Grep` now accepts ripgrep-style result ordering through `sort`/`--sort` and `sortr`/`--sortr`, covering path, modified-time, and none ordering for files/content/count modes with structured sort metadata.
 - `Grep` glob filtering now accepts ripgrep-style negated `!pattern` rules, including combinations with positive patterns, comma/whitespace splitting, and brace-expanded globs.
+- `Grep` now accepts `text`/`--text`/`-a` to search binary-extension files as text while preserving the default binary-extension skip behavior.
 - `Glob`/`Grep` traversal now applies `Read(...)` deny rules from the permission context as extra search ignore rules, hiding denied basename, path, and directory patterns from search results.
 - `Read`/`Edit` now coerce quoted semantic strings for `offset`/`limit` and `replace_all`, including whole-decimal numeric strings such as `"2.0"` for integer fields while keeping fractional values rejected.
 - `Bash`/`BashOutput` now coerce quoted semantic strings for `timeout`, `run_in_background`/`runInBackground`, and `tail_lines`/`tailLines`, matching official SDK-style number/boolean inputs without relaxing unknown-field validation.

internal/tools/file/helpers.go

@@ -81,6 +81,15 @@ func readText(path string) (string, error) {
 	return string(data), nil
 }
 
+func readTextAllowBinary(path string) (string, error) {
+	data, err := os.ReadFile(path)
+	if err != nil {
+		return "", err
+	}
+	data = bytes.TrimPrefix(data, []byte{0xef, 0xbb, 0xbf})
+	return string(data), nil
+}
+
 func readTextForEdit(path string) (content string, existed bool, crlf bool, mode os.FileMode, err error) {
 	info, statErr := os.Stat(path)
 	if statErr != nil {

internal/tools/file/search_tools.go

@@ -39,6 +39,7 @@ var allowedGrepInputKeys = map[string]struct{}{
 	"case_sensitive": {}, "caseSensitive": {}, "case-sensitive": {}, "--case-sensitive": {}, "-s": {},
 	"smart_case": {}, "smartCase": {}, "smart-case": {}, "--smart-case": {}, "-S": {},
 	"fixed_strings": {}, "fixedStrings": {}, "fixed-strings": {}, "--fixed-strings": {}, "-F": {}, "multiline": {}, "--multiline": {}, "multiline-dotall": {}, "--multiline-dotall": {}, "-U": {},
+	"text": {}, "--text": {}, "-a": {},
 	"word_regexp": {}, "wordRegexp": {}, "word-regexp": {}, "--word-regexp": {}, "-w": {},
 	"invert_match": {}, "invertMatch": {}, "invert-match": {}, "--invert-match": {}, "-v": {},
 	"only_matching": {}, "onlyMatching": {}, "only-matching": {}, "--only-matching": {}, "-o": {},
@@ -63,6 +64,7 @@ var grepSemanticBooleanKeys = map[string]struct{}{
 	"case_sensitive": {}, "caseSensitive": {}, "case-sensitive": {}, "--case-sensitive": {}, "-s": {},
 	"smart_case": {}, "smartCase": {}, "smart-case": {}, "--smart-case": {}, "-S": {},
 	"fixed_strings": {}, "fixedStrings": {}, "fixed-strings": {}, "--fixed-strings": {}, "-F": {}, "multiline": {}, "--multiline": {}, "multiline-dotall": {}, "--multiline-dotall": {}, "-U": {},
+	"text": {}, "--text": {}, "-a": {},
 	"word_regexp": {}, "wordRegexp": {}, "word-regexp": {}, "--word-regexp": {}, "-w": {},
 	"invert_match": {}, "invertMatch": {}, "invert-match": {}, "--invert-match": {}, "-v": {},
 	"only_matching": {}, "onlyMatching": {}, "only-matching": {}, "--only-matching": {}, "-o": {},
@@ -155,6 +157,9 @@ type grepInput struct {
 	FixedStringsDash        bool   `json:"fixed-strings,omitempty"`
 	LongFixedStrings        bool   `json:"--fixed-strings,omitempty"`
 	ShortFixedStrings       bool   `json:"-F,omitempty"`
+	Text                    bool   `json:"text,omitempty"`
+	LongText                bool   `json:"--text,omitempty"`
+	ShortText               bool   `json:"-a,omitempty"`
 	WordRegexp              bool   `json:"word_regexp,omitempty"`
 	WordRegexpAlt           bool   `json:"wordRegexp,omitempty"`
 	WordRegexpDash          bool   `json:"word-regexp,omitempty"`
@@ -236,6 +241,7 @@ type grepOptions struct {
 	OnlyMatching  bool
 	CountMatches  bool
 	ColumnNumbers bool
+	Text          bool
 	SortMode      string
 	SortReverse   bool
 	SortExplicit  bool
@@ -372,6 +378,9 @@ func NewGrepTool() tool.Tool {
 					"fixed-strings":    map[string]any{"type": "boolean"},
 					"--fixed-strings":  map[string]any{"type": "boolean"},
 					"-F":               map[string]any{"type": "boolean"},
+					"text":             map[string]any{"type": "boolean"},
+					"--text":           map[string]any{"type": "boolean"},
+					"-a":               map[string]any{"type": "boolean"},
 					"word_regexp":      map[string]any{"type": "boolean"},
 					"wordRegexp":       map[string]any{"type": "boolean"},
 					"word-regexp":      map[string]any{"type": "boolean"},
@@ -453,7 +462,7 @@ func NewGrepTool() tool.Tool {
 			},
 		},
 		PromptFunc: func(tool.PromptContext) (string, error) {
-			return "Searches text files under path using a regular expression or fixed string. pattern is the canonical search expression; regex/regexp/--regexp/-e are accepted aliases. output_mode may be files_with_matches, files_without_matches, content, or count; glob/-g/--glob and type/-t/--type optionally filter file paths. glob accepts whitespace/comma-separated patterns and brace alternation. content mode supports context, before_context, after_context, -C, -B, -A, -n/--line-number and -N/--no-line-number line-number control, --column column-number output, offset, head_limit pagination, max_count/-m per-file match limiting, max_columns/--max-columns long-line omission, and only_matching/-o/--only-matching matched-text output. Use files_with_matches or -l to list files with matches, files_without_match or -L to list files without matches, and count/--count/-c for count mode. Count mode supports count_matches/--count-matches for occurrence counts. Use sort/--sort or sortr/--sortr with path or modified to control result ordering. Use fixed_strings/-F/--fixed-strings for literal matching, word_regexp/-w/--word-regexp for whole-word matches, ignore_case/-i/--ignore-case for case-insensitive search, case_sensitive/-s/--case-sensitive to force case-sensitive matching, smart_case/-S/--smart-case for lowercase-only patterns, and invert_match/-v/--invert-match to select non-matching lines. Set no_ignore/--no-ignore to skip .gitignore/.ignore files while still excluding VCS metadata and read-denied paths. Set multiline to allow patterns to span lines with dot matching newlines.", nil
+			return "Searches text files under path using a regular expression or fixed string. pattern is the canonical search expression; regex/regexp/--regexp/-e are accepted aliases. output_mode may be files_with_matches, files_without_matches, content, or count; glob/-g/--glob and type/-t/--type optionally filter file paths. glob accepts whitespace/comma-separated patterns and brace alternation. content mode supports context, before_context, after_context, -C, -B, -A, -n/--line-number and -N/--no-line-number line-number control, --column column-number output, offset, head_limit pagination, max_count/-m per-file match limiting, max_columns/--max-columns long-line omission, and only_matching/-o/--only-matching matched-text output. Use files_with_matches or -l to list files with matches, files_without_match or -L to list files without matches, and count/--count/-c for count mode. Count mode supports count_matches/--count-matches for occurrence counts. Use sort/--sort or sortr/--sortr with path or modified to control result ordering. Use fixed_strings/-F/--fixed-strings for literal matching, text/-a/--text to search binary-extension files as text, word_regexp/-w/--word-regexp for whole-word matches, ignore_case/-i/--ignore-case for case-insensitive search, case_sensitive/-s/--case-sensitive to force case-sensitive matching, smart_case/-S/--smart-case for lowercase-only patterns, and invert_match/-v/--invert-match to select non-matching lines. Set no_ignore/--no-ignore to skip .gitignore/.ignore files while still excluding VCS metadata and read-denied paths. Set multiline to allow patterns to span lines with dot matching newlines.", nil
 		},
 		NormalizeFunc:   normalizeGrepRawInput,
 		ValidateFunc:    validateGrep,
@@ -629,6 +638,7 @@ func callGrep(ctx tool.Context, raw json.RawMessage, _ tool.ProgressSink) (contr
 		OnlyMatching:  onlyMatching,
 		CountMatches:  countMatches,
 		ColumnNumbers: grepColumnNumbers(input),
+		Text:          grepText(input),
 		SortMode:      sortMode,
 		SortReverse:   sortReverse,
 		SortExplicit:  sortExplicit,
@@ -667,6 +677,7 @@ func callGrep(ctx tool.Context, raw json.RawMessage, _ tool.ProgressSink) (contr
 			"case_sensitive":      grepCaseSensitive(input),
 			"smart_case":          grepSmartCase(input),
 			"fixed_strings":       grepFixedStrings(input),
+			"text":                options.Text,
 			"word_regexp":         grepWordRegexp(input),
 			"invert_match":        grepInvertMatch(input),
 			"only_matching":       onlyMatching,
@@ -843,10 +854,10 @@ func collectGrepMatches(root string, displayRoot string, glob string, typeFilter
 		if len(typeExtensions) > 0 && !grepTypeMatches(path, typeExtensions) {
 			return nil
 		}
-		if hasBinaryExtension(path) {
+		if hasBinaryExtension(path) && !options.Text {
 			return nil
 		}
-		content, err := readText(path)
+		content, err := readGrepText(path, options.Text)
 		if err != nil {
 			return nil
 		}
@@ -936,6 +947,13 @@ func globBaseDirectory(pattern string) (string, string) {
 	return filepath.Clean(base), pattern[lastSep+1:]
 }
 
+func readGrepText(path string, allowBinary bool) (string, error) {
+	if allowBinary {
+		return readTextAllowBinary(path)
+	}
+	return readText(path)
+}
+
 func grepFileMatches(path string, content string, expr *regexp.Regexp, options grepOptions) []grepMatch {
 	content = normalizeGrepContent(content)
 	lines := strings.Split(content, "\n")
@@ -1510,6 +1528,10 @@ func grepFixedStrings(input grepInput) bool {
 		input.ShortFixedStrings
 }
 
+func grepText(input grepInput) bool {
+	return input.Text || input.LongText || input.ShortText
+}
+
 func grepWordRegexp(input grepInput) bool {
 	return input.WordRegexp ||
 		input.WordRegexpAlt ||

internal/tools/file/tools_test.go

@@ -1919,6 +1919,52 @@ func TestGrepToolColumnNumbers(t *testing.T) {
 	}
 }
 
+func TestGrepToolTextSearchesBinaryExtensionFiles(t *testing.T) {
+	dir := t.TempDir()
+	if err := os.WriteFile(filepath.Join(dir, "payload.bin"), []byte("Needle\x00inside\n"), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	executor := fileExecutor(t)
+	ctx := fileToolContext(dir)
+
+	defaultResult, err := executor.Execute(ctx, contracts.ToolUse{
+		ID:    "toolu_grep_binary_default",
+		Name:  "Grep",
+		Input: json.RawMessage(`{"pattern":"Needle"}`),
+	}, nil)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if defaultResult.Content != "No files found" || defaultResult.StructuredContent["text"] != false {
+		t.Fatalf("default binary result = %#v", defaultResult)
+	}
+
+	textResult, err := executor.Execute(ctx, contracts.ToolUse{
+		ID:    "toolu_grep_binary_text",
+		Name:  "Grep",
+		Input: json.RawMessage(`{"pattern":"Needle","--text":true}`),
+	}, nil)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if textResult.Content != "Found 1 file\npayload.bin" || textResult.StructuredContent["text"] != true {
+		t.Fatalf("text binary result = %#v", textResult)
+	}
+
+	shortTextResult, err := executor.Execute(ctx, contracts.ToolUse{
+		ID:    "toolu_grep_binary_short_text",
+		Name:  "Grep",
+		Input: json.RawMessage(`{"pattern":"Needle","output_mode":"count","-a":"true"}`),
+	}, nil)
+	if err != nil {
+		t.Fatal(err)
+	}
+	wantShortText := "payload.bin:1\n\nFound 1 total occurrence across 1 file."
+	if shortTextResult.Content != wantShortText || shortTextResult.StructuredContent["text"] != true {
+		t.Fatalf("short text binary result = %#v", shortTextResult)
+	}
+}
+
 func TestGrepToolMaxColumnsOmission(t *testing.T) {
 	dir := t.TempDir()
 	longMatch := strings.Repeat("x", defaultGrepMaxColumns-len("Needle")) + "Needle"