symforge 11.7.1

SymForge — in-memory code intelligence MCP server for Claude Code, Codex, and Gemini CLI
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
use crate::domain::SymbolRecord;

/// The exact byte plan used by both MCP replacement and embedded replacement.
pub(crate) struct PreparedReplace {
    pub new_content: Vec<u8>,
    pub old_bytes: usize,
    pub inserted_bytes: usize,
}

pub(crate) fn prepare_replace(
    file_content: &[u8],
    symbol: &SymbolRecord,
    new_body: &str,
) -> Result<PreparedReplace, &'static str> {
    let (start, end) = symbol.byte_range;
    let (splice_start, _) = replacement_splice_range(file_content, symbol, new_body)?;
    let indent = detect_indentation(file_content, start);
    let line_ending = detect_line_ending(file_content);
    let normalized = normalize_line_endings(new_body.as_bytes(), line_ending);
    let normalized_str = std::str::from_utf8(&normalized).unwrap_or(new_body);
    let indented = apply_indentation(normalized_str, &indent, line_ending);
    let new_content = apply_splice(file_content, (splice_start, end), &indented);
    Ok(PreparedReplace {
        new_content,
        old_bytes: (end - start) as usize,
        inserted_bytes: indented.len(),
    })
}

/// The physical source bytes consumed by `prepare_replace`, including any
/// leading modifiers, indentation, or orphaned docs it actually removes.
pub(crate) fn replacement_splice_range(
    file_content: &[u8],
    symbol: &SymbolRecord,
    new_body: &str,
) -> Result<(u32, u32), &'static str> {
    let (start, end) = symbol.byte_range;
    if start > end || end as usize > file_content.len() || symbol.effective_start() > start {
        return Err("symbol span is outside the captured file");
    }
    let new_body_supplies_docs = body_starts_with_doc_comment(new_body);
    let effective = if new_body_supplies_docs {
        symbol.effective_start() as usize
    } else {
        start as usize
    };
    let raw_line_start = file_content[..effective]
        .iter()
        .rposition(|&byte| byte == b'\n')
        .map(|position| position + 1)
        .unwrap_or(0);
    let splice_start = if new_body_supplies_docs {
        extend_past_orphaned_docs(file_content, raw_line_start, symbol)
    } else {
        docless_replacement_splice_start(file_content, raw_line_start, start as usize)
    };
    Ok((splice_start as u32, end))
}

// ---------------------------------------------------------------------------
// Core splice
// ---------------------------------------------------------------------------

/// Splice `replacement` bytes into `content` at the given byte range [start, end).
pub(crate) fn apply_splice(content: &[u8], range: (u32, u32), replacement: &[u8]) -> Vec<u8> {
    let (start, end) = (range.0 as usize, range.1 as usize);
    let mut result = Vec::with_capacity(content.len() - (end - start) + replacement.len());
    result.extend_from_slice(&content[..start]);
    result.extend_from_slice(replacement);
    result.extend_from_slice(&content[end..]);
    result
}

// ---------------------------------------------------------------------------
// Line ending detection and normalization
// ---------------------------------------------------------------------------

/// Detected line ending style of a file.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub(crate) enum LineEnding {
    Lf,
    CrLf,
}

impl LineEnding {
    /// Returns the byte sequence for this line ending.
    pub(crate) fn as_bytes(&self) -> &[u8] {
        match self {
            LineEnding::Lf => b"\n",
            LineEnding::CrLf => b"\r\n",
        }
    }
}

/// Detect the dominant line ending style in file content.
/// Counts \r\n pairs vs lone \n. If \r\n > lone \n → CrLf, else Lf.
/// Empty or no-newline content defaults to Lf.
pub(crate) fn detect_line_ending(content: &[u8]) -> LineEnding {
    let mut crlf_count: usize = 0;
    let mut lf_count: usize = 0;
    let mut i = 0;
    while i < content.len() {
        if i + 1 < content.len() && content[i] == b'\r' && content[i + 1] == b'\n' {
            crlf_count += 1;
            i += 2;
        } else if content[i] == b'\n' {
            lf_count += 1;
            i += 1;
        } else {
            i += 1;
        }
    }
    if crlf_count > lf_count {
        LineEnding::CrLf
    } else {
        LineEnding::Lf
    }
}

/// Normalize line endings in generated/replacement text to match the target style.
/// 1. Convert \r\n → \n  2. Convert lone \r → \n  3. If target is CrLf, convert \n → \r\n
pub(crate) fn normalize_line_endings(text: &[u8], target: LineEnding) -> Vec<u8> {
    // Step 1+2: canonicalize to \n
    let mut canonical = Vec::with_capacity(text.len());
    let mut i = 0;
    while i < text.len() {
        if i + 1 < text.len() && text[i] == b'\r' && text[i + 1] == b'\n' {
            canonical.push(b'\n');
            i += 2;
        } else if text[i] == b'\r' {
            canonical.push(b'\n');
            i += 1;
        } else {
            canonical.push(text[i]);
            i += 1;
        }
    }
    match target {
        LineEnding::Lf => canonical,
        LineEnding::CrLf => {
            let mut result = Vec::with_capacity(canonical.len() * 2);
            for &byte in &canonical {
                if byte == b'\n' {
                    result.extend_from_slice(b"\r\n");
                } else {
                    result.push(byte);
                }
            }
            result
        }
    }
}

// ---------------------------------------------------------------------------
// Indentation utilities
// ---------------------------------------------------------------------------

/// Detect the leading whitespace on the line containing `byte_offset`.
pub(crate) fn detect_indentation(content: &[u8], byte_offset: u32) -> Vec<u8> {
    let offset = byte_offset as usize;
    let line_start = content[..offset]
        .iter()
        .rposition(|&b| b == b'\n')
        .map(|p| p + 1)
        .unwrap_or(0);
    let indent_end = content[line_start..]
        .iter()
        .position(|b| !b.is_ascii_whitespace() || *b == b'\n')
        .unwrap_or(0);
    content[line_start..line_start + indent_end].to_vec()
}

/// The longest leading-whitespace prefix common to every line of `lines` that
/// has non-whitespace content (blank / whitespace-only lines are ignored). This
/// is the body's uniform base indent — empty when any content line is already
/// flush-left (the normal case). Mirrors the prefix `textwrap.dedent` strips.
fn common_leading_whitespace<'a>(lines: &[&'a str]) -> &'a str {
    let mut common: Option<&'a str> = None;
    for raw in lines {
        let line = raw.strip_suffix('\r').unwrap_or(raw);
        if line.trim().is_empty() {
            continue;
        }
        let ws = &line[..line.len() - line.trim_start().len()];
        common = Some(match common {
            None => ws,
            Some(prev) => {
                let max = prev.len().min(ws.len());
                let (pb, wb) = (prev.as_bytes(), ws.as_bytes());
                let mut end = 0;
                while end < max && pb[end] == wb[end] {
                    end += 1;
                }
                &prev[..end]
            }
        });
        if common == Some("") {
            break;
        }
    }
    common.unwrap_or("")
}

/// Re-column `text` to `indent`: strip the body's uniform base indent, then
/// prefix each non-empty line with `indent`, using the given line ending.
///
/// Stripping the common base indent first means a body the caller pasted at
/// some other column (e.g. an 8-space chat-context indent) is re-columned to
/// exactly the symbol's `indent` rather than COMPOUNDING to base+indent. When
/// the body is already flush-left (its first content line has no leading
/// whitespace — the normal case) the base is empty and this is a pure prefix,
/// so existing callers are unaffected.
pub(crate) fn apply_indentation(text: &str, indent: &[u8], line_ending: LineEnding) -> Vec<u8> {
    let mut result = Vec::new();
    // Use split('\n') instead of lines() so that trailing newlines produce a trailing
    // empty element, preserving them. str::lines() silently strips all trailing newlines.
    let parts: Vec<&str> = text.split('\n').collect();
    let base = common_leading_whitespace(&parts);
    for (i, line) in parts.iter().enumerate() {
        // Strip '\r' left behind by split('\n') on CRLF input; re-emit via line_ending.
        let line = line.strip_suffix('\r').unwrap_or(line);
        if i > 0 {
            result.extend_from_slice(line_ending.as_bytes());
        }
        if !line.is_empty() {
            // Dedent the uniform base, then apply the symbol's column.
            let dedented = line.strip_prefix(base).unwrap_or(line);
            result.extend_from_slice(indent);
            result.extend_from_slice(dedented.as_bytes());
        }
    }
    result
}

// ---------------------------------------------------------------------------
// Insert helpers
// ---------------------------------------------------------------------------

/// Build the bytes to insert before a symbol: indented content + separator + existing content.
/// Splices at the start of the line (before existing indentation) so indentation isn't doubled.
/// Uses `\n\n` when the target symbol has no doc comments and no blank line already precedes
/// the splice point (visual separation between definitions), and `\n` otherwise (avoids triple
/// newlines when a blank line already exists, and keeps doc comments tight against their symbol).
pub(crate) fn build_insert_before(
    file_content: &[u8],
    sym: &SymbolRecord,
    new_code: &str,
    line_ending: LineEnding,
) -> Vec<u8> {
    let line_start = insert_before_position(file_content, sym);
    let indent = detect_indentation(file_content, sym.byte_range.0);
    let normalized = normalize_line_endings(new_code.as_bytes(), line_ending);
    let normalized_str = std::str::from_utf8(&normalized).unwrap_or(new_code);
    let indented = apply_indentation(normalized_str, &indent, line_ending);
    let mut insertion = indented;
    let le = line_ending.as_bytes();
    let separator: Vec<u8> = if sym.doc_byte_range.is_some() {
        le.to_vec()
    } else {
        // Use single newline only when a blank line already precedes the symbol
        // (avoids creating triple-newline sequences). At start-of-file (empty prefix),
        // there's no existing blank line, so use double newline for visual separation.
        let prefix = &file_content[..line_start as usize];
        let already_has_blank = match line_ending {
            LineEnding::CrLf => {
                prefix.len() >= 4
                    && prefix[prefix.len() - 2] == b'\r'
                    && prefix[prefix.len() - 1] == b'\n'
                    && prefix[prefix.len() - 4] == b'\r'
                    && prefix[prefix.len() - 3] == b'\n'
            }
            LineEnding::Lf => {
                prefix.len() >= 2
                    && prefix[prefix.len() - 1] == b'\n'
                    && prefix[prefix.len() - 2] == b'\n'
            }
        };
        if already_has_blank {
            le.to_vec()
        } else {
            let mut sep = Vec::with_capacity(le.len() * 2);
            sep.extend_from_slice(le);
            sep.extend_from_slice(le);
            sep
        }
    };
    insertion.extend_from_slice(&separator);
    apply_splice(file_content, (line_start, line_start), &insertion)
}

pub(crate) fn insert_before_position(file_content: &[u8], sym: &SymbolRecord) -> u32 {
    let sym_start = sym.effective_start() as usize;
    file_content[..sym_start]
        .iter()
        .rposition(|&byte| byte == b'\n')
        .map(|position| position + 1)
        .unwrap_or(0) as u32
}

/// Build the bytes to insert after a symbol: existing content + blank line + indented content.
///
/// Handles the C/C++ quirk where struct/enum/class definitions end their tree-sitter
/// node at `}` while the actual declaration includes a trailing `;`.  When the byte
/// immediately following the symbol end (skipping spaces/tabs) is `;`, the insertion
/// point moves past it so the result stays syntactically valid.
pub(crate) fn build_insert_after(
    file_content: &[u8],
    sym: &SymbolRecord,
    new_code: &str,
    line_ending: LineEnding,
) -> Vec<u8> {
    let indent = detect_indentation(file_content, sym.byte_range.0);
    let normalized = normalize_line_endings(new_code.as_bytes(), line_ending);
    let normalized_str = std::str::from_utf8(&normalized).unwrap_or(new_code);
    let indented = apply_indentation(normalized_str, &indent, line_ending);
    let le = line_ending.as_bytes();
    let mut insertion = Vec::new();
    insertion.extend_from_slice(le);
    insertion.extend_from_slice(le);
    insertion.extend_from_slice(&indented);
    // Skip past a trailing `;` that belongs to the parent declaration (C/C++
    // struct/enum/class: tree-sitter node ends at `}`, declaration at `};`).
    let insert_pos = insert_after_position(file_content, sym);
    apply_splice(file_content, (insert_pos, insert_pos), &insertion)
}

pub(crate) fn insert_after_position(file_content: &[u8], sym: &SymbolRecord) -> u32 {
    skip_trailing_semicolon(file_content, sym.byte_range.1 as usize) as u32
}

/// If the byte(s) immediately after `pos` (skipping spaces and tabs, but not
/// newlines) form a `;`, return the position just past it.  Otherwise return `pos`.
fn skip_trailing_semicolon(content: &[u8], pos: usize) -> usize {
    let mut i = pos;
    while i < content.len() && (content[i] == b' ' || content[i] == b'\t') {
        i += 1;
    }
    if i < content.len() && content[i] == b';' {
        i + 1
    } else {
        pos
    }
}

// ---------------------------------------------------------------------------
// Delete helper
// ---------------------------------------------------------------------------

/// Build file content with the symbol removed, including leading whitespace and trailing newlines.
/// Collapses runs of 3+ consecutive blank lines down to 1 after deletion.
/// Scan upward from `line_start` to include orphaned doc comments when
/// `doc_byte_range` is `None`. Returns the (possibly earlier) byte offset
/// that includes the orphaned comments. Used by `build_delete` and
/// `replace_symbol_body` to handle blank-line-separated doc comments.
pub(crate) fn extend_past_orphaned_docs(
    file_content: &[u8],
    line_start: usize,
    sym: &SymbolRecord,
) -> usize {
    if sym.doc_byte_range.is_some() {
        return line_start;
    }
    let above = &file_content[..line_start];
    let lines: Vec<&[u8]> = above.split(|&b| b == b'\n').collect();
    let mut i = lines.len();
    // Skip trailing empty element from split
    if i > 0 && lines[i - 1].is_empty() {
        i -= 1;
    }
    // Skip exactly one blank line
    if i > 0 && lines[i - 1].iter().all(|b| b.is_ascii_whitespace()) {
        i -= 1;
        // Collect consecutive comment lines above the blank line
        let mut found_comments = false;
        while i > 0 {
            let line_text = std::str::from_utf8(lines[i - 1]).unwrap_or("");
            let trimmed = line_text.trim_start();
            if trimmed.starts_with("///")
                || trimmed.starts_with("//!")
                || trimmed.starts_with("/**")
                || trimmed.starts_with("* ")
                || trimmed == "*/"
                || trimmed.starts_with("# ")
                || trimmed == "#"
            {
                found_comments = true;
                i -= 1;
            } else {
                break;
            }
        }
        if found_comments {
            // split('\n') leaves \r in slices for CRLF; +1 accounts for the \n separator
            return lines[..i].iter().map(|l| l.len() + 1).sum();
        }
    }
    line_start
}

/// Walk upward from `line_start` (the first byte of the symbol's opening
/// line) and include contiguous Rust outer-attribute lines (`#[...]`) that sit
/// directly above the item with no blank line between. Attributes belong to
/// the item; leaving them behind on a delete orphans them onto the following
/// item — for example a stray `#[test]` that then fails to compile. Inner
/// attributes (`#![...]`) are not consumed (they belong to the enclosing
/// scope), and only the leading line of a multi-line attribute is recognized,
/// which is never worse than the previous behaviour of consuming none.
fn extend_past_leading_attributes(file_content: &[u8], line_start: usize) -> usize {
    let mut start = line_start;
    while start > 0 {
        // `start` sits just past a '\n'; find the bounds of the line above it.
        let prev_line_end = start - 1;
        let prev_line_start = file_content[..prev_line_end]
            .iter()
            .rposition(|&b| b == b'\n')
            .map(|p| p + 1)
            .unwrap_or(0);
        let line = &file_content[prev_line_start..prev_line_end];
        let trimmed = std::str::from_utf8(line).unwrap_or("").trim();
        if trimmed.starts_with("#[") {
            start = prev_line_start;
        } else {
            break;
        }
    }
    start
}

/// Whether `body` begins (first non-blank line) with a doc-comment marker.
///
/// Used by `replace_symbol_body` to decide whether the caller intends to
/// supply fresh docs for the symbol. When true, the splice range extends
/// past the existing docs so the old ones are replaced. When false, the
/// splice starts at the signature line so attached docs are preserved.
///
/// Conservative on purpose: only matches markers that are unambiguously
/// doc comments across the grammars SymForge indexes. Line comments like
/// `//` and `#` are NOT counted because they may be ordinary code
/// comments or, for `#`, Rust attributes (e.g., `#[inline]`).
pub(crate) fn body_starts_with_doc_comment(body: &str) -> bool {
    let Some(first) = body.lines().find(|l| !l.trim().is_empty()) else {
        return false;
    };
    let trimmed = first.trim_start();
    trimmed.starts_with("///")
        || trimmed.starts_with("//!")
        || trimmed.starts_with("/**")
        || trimmed.starts_with("/*!")
        || trimmed.starts_with("#[doc")
}

/// Return the splice start for a docless replacement.
///
/// Normally this is the start of the symbol's source line. When a doc marker
/// shares the line with the symbol, preserve the marker and its separator, then
/// replace the old modifiers/signature with the caller's `new_body`.
pub(crate) fn docless_replacement_splice_start(
    file_content: &[u8],
    raw_line_start: usize,
    symbol_start: usize,
) -> usize {
    if raw_line_start >= symbol_start || symbol_start > file_content.len() {
        return raw_line_start;
    }

    let prefix = &file_content[raw_line_start..symbol_start];
    same_line_doc_prefix_end(prefix)
        .map(|end| raw_line_start + end)
        .unwrap_or(raw_line_start)
}

fn same_line_doc_prefix_end(prefix: &[u8]) -> Option<usize> {
    let Ok(text) = std::str::from_utf8(prefix) else {
        return None;
    };
    let leading = text.len() - text.trim_start().len();
    let trimmed = &text[leading..];

    if trimmed.starts_with("/**") || trimmed.starts_with("/*!") {
        let marker_end = trimmed.find("*/")? + 2;
        let after_padding = trimmed[marker_end..]
            .find(|c: char| !c.is_whitespace())
            .map(|pos| marker_end + pos)
            .unwrap_or(trimmed.len());
        return Some(leading + after_padding);
    }

    if trimmed.starts_with("#[doc") {
        let marker_end = trimmed.find(']')? + 1;
        let after_padding = trimmed[marker_end..]
            .find(|c: char| !c.is_whitespace())
            .map(|pos| marker_end + pos)
            .unwrap_or(trimmed.len());
        return Some(leading + after_padding);
    }

    None
}

pub(crate) fn build_delete(
    file_content: &[u8],
    sym: &SymbolRecord,
    line_ending: LineEnding,
) -> Vec<u8> {
    let (start, end) = delete_source_range(file_content, sym, line_ending);
    let spliced = apply_splice(file_content, (start, end), b"");
    collapse_blank_lines(&spliced, line_ending)
}

pub(crate) fn delete_source_range(
    file_content: &[u8],
    sym: &SymbolRecord,
    line_ending: LineEnding,
) -> (u32, u32) {
    // Extend to start of line (include leading whitespace, attached attributes,
    // and orphaned doc comments).
    let start = {
        let s = sym.effective_start() as usize;
        let line_start = file_content[..s]
            .iter()
            .rposition(|&b| b == b'\n')
            .map(|p| p + 1)
            .unwrap_or(0);
        // Consume contiguous outer-attribute lines (`#[...]`) directly above the
        // item so they are removed with it instead of being orphaned onto the
        // next item (e.g. a stray `#[test]`, which then fails to compile).
        let after_attrs = extend_past_leading_attributes(file_content, line_start);
        extend_past_orphaned_docs(file_content, after_attrs, sym) as u32
    };
    // Extend past trailing newlines (consume up to one blank line).
    // CRLF-aware: on CRLF files, a line ending is \r\n not just \n.
    let end = {
        let e = sym.byte_range.1 as usize;
        let mut pos = e;
        // Skip to end of current line (past any trailing non-newline chars).
        while pos < file_content.len() && file_content[pos] != b'\n' {
            pos += 1;
        }
        // Consume the \n (or \r\n).
        if pos < file_content.len() && file_content[pos] == b'\n' {
            pos += 1;
        }
        // Consume one more blank line if present.
        match line_ending {
            LineEnding::CrLf => {
                if pos + 1 < file_content.len()
                    && file_content[pos] == b'\r'
                    && file_content[pos + 1] == b'\n'
                {
                    pos += 2;
                }
            }
            LineEnding::Lf => {
                if pos < file_content.len() && file_content[pos] == b'\n' {
                    pos += 1;
                }
            }
        }
        pos as u32
    };
    (start, end)
}

#[derive(Clone, Copy, Debug)]
pub(crate) enum SpliceFootprint {
    Bytes(u32, u32),
    Insert(u32),
}

pub(crate) fn splice_footprints_overlap(left: SpliceFootprint, right: SpliceFootprint) -> bool {
    match (left, right) {
        (SpliceFootprint::Bytes(a, b), SpliceFootprint::Bytes(c, d)) => a < d && c < b,
        (SpliceFootprint::Insert(at), SpliceFootprint::Bytes(start, end))
        | (SpliceFootprint::Bytes(start, end), SpliceFootprint::Insert(at)) => {
            start <= at && at < end
        }
        (SpliceFootprint::Insert(left), SpliceFootprint::Insert(right)) => left == right,
    }
}

/// Extra source ranges removed by `build_delete`'s whole-file blank-line
/// cleanup, expressed in the same original byte coordinates as the main
/// deletion. A batch must account for these before applying cached spans.
pub(crate) fn delete_cleanup_ranges(
    file_content: &[u8],
    sym: &SymbolRecord,
    line_ending: LineEnding,
) -> Vec<(u32, u32)> {
    let (start, end) = delete_source_range(file_content, sym, line_ending);
    let spliced = apply_splice(file_content, (start, end), b"");
    let mut removed = Vec::new();
    match line_ending {
        LineEnding::Lf => {
            let mut consecutive = 0;
            for (position, byte) in spliced.iter().enumerate() {
                if *byte == b'\n' {
                    consecutive += 1;
                    if consecutive > 2 {
                        push_cleanup_range(&mut removed, position, position + 1, start, end);
                    }
                } else {
                    consecutive = 0;
                }
            }
        }
        LineEnding::CrLf => {
            let mut consecutive = 0;
            let mut position = 0;
            while position < spliced.len() {
                if spliced.get(position..position + 2) == Some(b"\r\n") {
                    consecutive += 1;
                    if consecutive > 2 {
                        push_cleanup_range(&mut removed, position, position + 2, start, end);
                    }
                    position += 2;
                } else {
                    consecutive = 0;
                    position += 1;
                }
            }
        }
    }
    removed
}

fn push_cleanup_range(
    removed: &mut Vec<(u32, u32)>,
    spliced_start: usize,
    spliced_end: usize,
    main_start: u32,
    main_end: u32,
) {
    let boundary = main_start as usize;
    for (start, end) in [
        (spliced_start, spliced_end.min(boundary)),
        (spliced_start.max(boundary), spliced_end),
    ] {
        if start >= end {
            continue;
        }
        let shift = if start >= boundary {
            (main_end - main_start) as usize
        } else {
            0
        };
        let source_start = (start + shift) as u32;
        let source_end = (end + shift) as u32;
        if let Some(last) = removed.last_mut()
            && last.1 == source_start
        {
            last.1 = source_end;
        } else {
            removed.push((source_start, source_end));
        }
    }
}

/// Any removed byte before an unprocessed splice shifts its captured offset;
/// removal within a splice changes its target bytes. Both require refusal.
pub(crate) fn cleanup_invalidates_splice(cleanup: &[(u32, u32)], other: SpliceFootprint) -> bool {
    let last_target_byte = match other {
        SpliceFootprint::Bytes(_, end) => end,
        SpliceFootprint::Insert(at) => at.saturating_add(1),
    };
    cleanup.iter().any(|(start, _)| *start < last_target_byte)
}

/// Collapse runs of 3+ consecutive newlines down to 2 (one blank line).
/// On CRLF files, counts `\r\n` pairs; on LF files, counts `\n` bytes.
pub(crate) fn collapse_blank_lines(content: &[u8], line_ending: LineEnding) -> Vec<u8> {
    let mut result = Vec::with_capacity(content.len());
    match line_ending {
        LineEnding::Lf => {
            let mut consecutive_newlines = 0u32;
            for &b in content {
                if b == b'\n' {
                    consecutive_newlines += 1;
                    if consecutive_newlines <= 2 {
                        result.push(b);
                    }
                } else {
                    consecutive_newlines = 0;
                    result.push(b);
                }
            }
        }
        LineEnding::CrLf => {
            // Count \r\n pairs as line endings; threshold at 2 pairs (one blank line).
            let mut consecutive_line_endings = 0u32;
            let mut i = 0;
            while i < content.len() {
                if i + 1 < content.len() && content[i] == b'\r' && content[i + 1] == b'\n' {
                    consecutive_line_endings += 1;
                    if consecutive_line_endings <= 2 {
                        result.push(b'\r');
                        result.push(b'\n');
                    }
                    i += 2;
                } else {
                    consecutive_line_endings = 0;
                    result.push(content[i]);
                    i += 1;
                }
            }
        }
    }
    result
}

// ---------------------------------------------------------------------------
// Edit-within helper
// ---------------------------------------------------------------------------

/// Find-and-replace text within a symbol's byte range. Returns (new_content, replacement_count).
pub(crate) fn build_edit_within(
    file_content: &[u8],
    sym: &SymbolRecord,
    old_text: &str,
    new_text: &str,
    replace_all: bool,
) -> Result<(Vec<u8>, usize), String> {
    let sym_start = sym.effective_start() as usize;
    let sym_end = sym.byte_range.1 as usize;
    let body = &file_content[sym_start..sym_end];
    let body_str =
        std::str::from_utf8(body).map_err(|_| "Symbol body is not valid UTF-8.".to_string())?;

    // Callers (LLMs) almost always supply `\n`-separated text regardless of the
    // file's on-disk convention. Normalize both the search needle and the
    // replacement to the file's dominant line ending so matches succeed in
    // CRLF files and the splice never introduces mixed line endings.
    let line_ending = detect_line_ending(file_content);
    let needle = String::from_utf8(normalize_line_endings(old_text.as_bytes(), line_ending))
        .map_err(|_| "Normalized search text is not valid UTF-8.".to_string())?;
    let replacement = String::from_utf8(normalize_line_endings(new_text.as_bytes(), line_ending))
        .map_err(|_| "Normalized replacement text is not valid UTF-8.".to_string())?;

    let (new_body, count) = if replace_all {
        let count = body_str.matches(needle.as_str()).count();
        if count == 0 {
            return Err(format!(
                "`{old_text}` not found within symbol `{}`",
                sym.name
            ));
        }
        (
            body_str.replace(needle.as_str(), replacement.as_str()),
            count,
        )
    } else {
        match body_str.find(needle.as_str()) {
            Some(_) => (
                body_str.replacen(needle.as_str(), replacement.as_str(), 1),
                1,
            ),
            None => {
                return Err(format!(
                    "`{old_text}` not found within symbol `{}`",
                    sym.name
                ));
            }
        }
    };

    let effective_range = (sym.effective_start(), sym.byte_range.1);
    let new_content = apply_splice(file_content, effective_range, new_body.as_bytes());
    Ok((new_content, count))
}

/// Return the leading whitespace of the first non-blank line.
fn indent_of_first_nonempty<'a>(lines: &[&'a str]) -> &'a str {
    for line in lines {
        let trimmed = line.trim_start();
        if !trimmed.is_empty() {
            return &line[..line.len() - trimmed.len()];
        }
    }
    ""
}

/// Re-indent `line` from `old_base` indentation to `file_base`.
fn reindent_line(line: &str, old_base: &str, file_base: &str) -> String {
    if line.trim().is_empty() {
        return String::new();
    }
    match line.strip_prefix(old_base) {
        Some(rest) => format!("{file_base}{rest}"),
        None => {
            // Line has different indent depth than the base.
            let line_indent = line.len() - line.trim_start().len();
            let old_indent = old_base.len();
            if line_indent < old_indent {
                // Less indented (e.g. closing brace) — preserve relative de-indent.
                let deficit = old_indent - line_indent;
                if file_base.len() > deficit {
                    format!(
                        "{}{}",
                        &file_base[..file_base.len() - deficit],
                        line.trim_start()
                    )
                } else {
                    line.trim_start().to_string()
                }
            } else {
                // More indented but prefix mismatch (tabs vs spaces mix).
                let extra = &line[old_indent..line_indent];
                format!("{file_base}{extra}{}", line.trim_start())
            }
        }
    }
}

/// Attempt a whitespace-flexible find-and-replace within `body`.
///
/// When an exact match of `old_text` fails, this tries matching lines
/// with leading whitespace stripped.  If found, `new_text` is re-indented
/// to match the file's actual indentation before replacement.
///
/// Returns `Some((new_body, count))` on success, `None` if no flexible
/// match is found either.
pub(crate) fn try_whitespace_flexible_replace(
    body: &str,
    old_text: &str,
    new_text: &str,
    replace_all: bool,
) -> Option<(String, usize)> {
    let body_lines: Vec<&str> = body.lines().collect();
    let old_lines: Vec<&str> = old_text.lines().collect();

    if old_lines.is_empty() || old_lines.iter().all(|l| l.trim().is_empty()) {
        return None;
    }

    let old_trimmed: Vec<&str> = old_lines.iter().map(|l| l.trim_start()).collect();
    let window = old_trimmed.len();

    // Find matching positions (line-aligned, trimmed comparison).
    let mut matches: Vec<usize> = Vec::new();
    for start in 0..=body_lines.len().saturating_sub(window) {
        let hit = old_trimmed
            .iter()
            .enumerate()
            .all(|(i, ot)| body_lines[start + i].trim_start() == *ot);
        if hit {
            matches.push(start);
            if !replace_all {
                break;
            }
        }
    }

    if matches.is_empty() {
        return None;
    }

    // Pre-compute byte offset of each line start.
    let mut line_starts: Vec<usize> = vec![0];
    for (i, b) in body.bytes().enumerate() {
        if b == b'\n' {
            line_starts.push(i + 1);
        }
    }

    let count = matches.len();
    let mut result = body.to_string();

    // Process in reverse so earlier byte offsets remain valid.
    for &m in matches.iter().rev() {
        let byte_start = line_starts[m];
        let byte_end = if m + window < line_starts.len() {
            line_starts[m + window]
        } else {
            body.len()
        };

        let matched_lines = &body_lines[m..m + window];
        let old_base = indent_of_first_nonempty(&old_lines);
        let file_base = indent_of_first_nonempty(matched_lines);

        let reindented: Vec<String> = new_text
            .lines()
            .map(|l| reindent_line(l, old_base, file_base))
            .collect();
        let mut replacement = reindented.join("\n");

        // Preserve trailing newline when the matched region included one.
        if byte_end > byte_start
            && result.as_bytes().get(byte_end - 1) == Some(&b'\n')
            && !replacement.ends_with('\n')
        {
            replacement.push('\n');
        }

        result.replace_range(byte_start..byte_end, &replacement);
    }

    Some((result, count))
}

/// The same edit-within match selection used by the MCP writer and Embed.
/// Errors carry no source text or search needle so callers can apply their
/// own bounded, secret-safe presentation policy.
pub(crate) struct WithinSelection {
    pub new_body: String,
    pub count: usize,
    pub untargeted_extra: usize,
}

#[derive(Clone, Copy, Debug, PartialEq, Eq)]
pub(crate) enum WithinSelectionError {
    InvalidSpan,
    InvalidUtf8,
    ConflictingTargeting,
    OccurrenceOutOfRange { requested: u32, total: usize },
    NotFound,
}

pub(crate) fn select_edit_within_body(
    file_content: &[u8],
    sym: &SymbolRecord,
    old_text: &str,
    new_text: &str,
    replace_all: bool,
    occurrence: Option<u32>,
    near_line: Option<u32>,
) -> Result<WithinSelection, WithinSelectionError> {
    let sym_start = sym.effective_start() as usize;
    let sym_end = sym.byte_range.1 as usize;
    let body = file_content
        .get(sym_start..sym_end)
        .ok_or(WithinSelectionError::InvalidSpan)?;
    let body_str = std::str::from_utf8(body).map_err(|_| WithinSelectionError::InvalidUtf8)?;
    let line_ending = detect_line_ending(file_content);
    let normalized_old = normalize_line_endings(old_text.as_bytes(), line_ending);
    let normalized_old_str =
        String::from_utf8(normalized_old).unwrap_or_else(|_| old_text.to_string());
    let normalized_new = normalize_line_endings(new_text.as_bytes(), line_ending);
    let normalized_new_str =
        String::from_utf8(normalized_new).unwrap_or_else(|_| new_text.to_string());
    let targeting_modes = usize::from(replace_all)
        + usize::from(occurrence.is_some())
        + usize::from(near_line.is_some());
    if targeting_modes > 1 {
        return Err(WithinSelectionError::ConflictingTargeting);
    }
    let (new_body, count, untargeted_extra) = if replace_all {
        let count = body_str.matches(&normalized_old_str).count();
        if count > 0 {
            (
                body_str.replace(&normalized_old_str, &normalized_new_str),
                count,
                0,
            )
        } else {
            match try_whitespace_flexible_replace(
                body_str,
                &normalized_old_str,
                &normalized_new_str,
                true,
            ) {
                Some((body, count)) => (body, count, 0),
                None => return Err(WithinSelectionError::NotFound),
            }
        }
    } else if occurrence.is_some() || near_line.is_some() {
        let positions: Vec<usize> = body_str
            .match_indices(normalized_old_str.as_str())
            .map(|(pos, _)| pos)
            .collect();
        if positions.is_empty() {
            return Err(WithinSelectionError::NotFound);
        }
        let index = if let Some(n) = occurrence {
            let n_usize = n as usize;
            if n_usize == 0 || n_usize > positions.len() {
                return Err(WithinSelectionError::OccurrenceOutOfRange {
                    requested: n,
                    total: positions.len(),
                });
            }
            n_usize - 1
        } else {
            let target = i64::from(near_line.unwrap_or(1));
            let line_of = |pos: usize| {
                1 + file_content[..sym_start + pos]
                    .iter()
                    .filter(|&&byte| byte == b'\n')
                    .count() as i64
            };
            (0..positions.len())
                .min_by_key(|&i| (line_of(positions[i]) - target).abs())
                .unwrap_or(0)
        };
        let pos = positions[index];
        let mut body = String::with_capacity(body_str.len() + normalized_new_str.len());
        body.push_str(&body_str[..pos]);
        body.push_str(&normalized_new_str);
        body.push_str(&body_str[pos + normalized_old_str.len()..]);
        (body, 1, 0)
    } else if body_str.contains(normalized_old_str.as_str()) {
        (
            body_str.replacen(&normalized_old_str, &normalized_new_str, 1),
            1,
            body_str
                .matches(&normalized_old_str)
                .count()
                .saturating_sub(1),
        )
    } else {
        match try_whitespace_flexible_replace(
            body_str,
            &normalized_old_str,
            &normalized_new_str,
            false,
        ) {
            Some((body, count)) => (body, count, 0),
            None => return Err(WithinSelectionError::NotFound),
        }
    };
    Ok(WithinSelection {
        new_body,
        count,
        untargeted_extra,
    })
}

// ---------------------------------------------------------------------------