rigg_diff/

text_diff.rs

//! Line-level text diffing for long text properties.
//!
//! Produces structured diffs with context lines for human-readable
//! comparison of instructions, descriptions, and other long text fields.
//! Supports word-level highlighting within modified paragraphs.

use similar::{ChangeTag, TextDiff};

/// Minimum string length to trigger line-level diffing.
pub const LONG_TEXT_THRESHOLD: usize = 120;

/// Per-line threshold: lines longer than this are split at sentence boundaries.
const SENTENCE_SPLIT_THRESHOLD: usize = 120;

/// Target width when word-wrapping lines with no sentence boundaries.
const WORD_WRAP_WIDTH: usize = 80;

/// Maximum display length for context lines before truncation.
pub const CONTEXT_TRUNCATE_LEN: usize = 80;

/// A segment of a word-level diff within a single line.
#[derive(Debug, Clone, PartialEq)]
pub struct WordSegment {
    pub text: String,
    pub kind: WordSegmentKind,
}

/// Whether a word segment is unchanged or changed.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum WordSegmentKind {
    Equal,
    Changed,
}

/// A single line in a text diff.
#[derive(Debug, Clone, PartialEq)]
pub enum DiffLine {
    /// Unchanged context line.
    Equal(String),
    /// Deleted line (only in old text).
    Delete(String),
    /// Inserted line (only in new text).
    Insert(String),
    /// A modified line with word-level diff segments.
    /// Contains old line segments and new line segments.
    Modified {
        old_segments: Vec<WordSegment>,
        new_segments: Vec<WordSegment>,
    },
}

/// A group of contiguous diff lines with surrounding context.
#[derive(Debug, Clone)]
pub struct TextDiffHunk {
    pub lines: Vec<DiffLine>,
}

/// Result of a line-level text diff.
#[derive(Debug, Clone)]
pub struct TextDiffResult {
    pub hunks: Vec<TextDiffHunk>,
    pub deletions: usize,
    pub insertions: usize,
}

/// Returns `true` if the string is long enough or multi-line to warrant
/// line-level diffing instead of inline comparison.
pub fn is_long_text(s: &str) -> bool {
    s.len() >= LONG_TEXT_THRESHOLD || s.contains('\n')
}

/// Compute a line-level diff between two strings.
///
/// Returns hunks with 1 line of surrounding context per change group.
/// Consecutive Delete+Insert pairs are converted to Modified entries
/// with word-level diff segments.
pub fn diff_text(old: &str, new: &str) -> TextDiffResult {
    let diff = TextDiff::from_lines(old, new);
    let mut deletions = 0;
    let mut insertions = 0;
    let mut hunks = Vec::new();

    for group in diff.grouped_ops(1) {
        let mut lines = Vec::new();
        for op in &group {
            for change in diff.iter_changes(op) {
                let text = change.value().to_string();
                match change.tag() {
                    ChangeTag::Equal => lines.push(DiffLine::Equal(text)),
                    ChangeTag::Delete => {
                        deletions += 1;
                        lines.push(DiffLine::Delete(text));
                    }
                    ChangeTag::Insert => {
                        insertions += 1;
                        lines.push(DiffLine::Insert(text));
                    }
                }
            }
        }
        if !lines.is_empty() {
            lines = pair_modifications(lines);
            hunks.push(TextDiffHunk { lines });
        }
    }

    TextDiffResult {
        hunks,
        deletions,
        insertions,
    }
}

/// Convert consecutive Delete+Insert runs into Modified entries with word-level diffs.
///
/// Pairs them 1:1 — if there are 3 deletes and 2 inserts, the first 2 become
/// Modified pairs and the 3rd remains a standalone Delete.
fn pair_modifications(lines: Vec<DiffLine>) -> Vec<DiffLine> {
    let mut result = Vec::with_capacity(lines.len());
    let mut i = 0;

    while i < lines.len() {
        // Look for a run of Delete lines followed by Insert lines
        if matches!(&lines[i], DiffLine::Delete(_)) {
            let delete_start = i;
            while i < lines.len() && matches!(&lines[i], DiffLine::Delete(_)) {
                i += 1;
            }
            let delete_end = i;

            let insert_start = i;
            while i < lines.len() && matches!(&lines[i], DiffLine::Insert(_)) {
                i += 1;
            }
            let insert_end = i;

            let delete_count = delete_end - delete_start;
            let insert_count = insert_end - insert_start;
            let pair_count = delete_count.min(insert_count);

            // Pair deletes with inserts as Modified entries
            for j in 0..pair_count {
                let old_text = match &lines[delete_start + j] {
                    DiffLine::Delete(s) => s.clone(),
                    _ => unreachable!(),
                };
                let new_text = match &lines[insert_start + j] {
                    DiffLine::Insert(s) => s.clone(),
                    _ => unreachable!(),
                };

                let (old_segments, new_segments) = diff_words(&old_text, &new_text);
                result.push(DiffLine::Modified {
                    old_segments,
                    new_segments,
                });
            }

            // Emit remaining unpaired deletes
            for j in pair_count..delete_count {
                result.push(lines[delete_start + j].clone());
            }

            // Emit remaining unpaired inserts
            for j in pair_count..insert_count {
                result.push(lines[insert_start + j].clone());
            }
        } else {
            result.push(lines[i].clone());
            i += 1;
        }
    }

    result
}

/// Compute word-level diff segments between two strings.
///
/// Returns (old_segments, new_segments) where each segment indicates
/// whether it's equal or changed relative to the other side.
pub fn diff_words(old: &str, new: &str) -> (Vec<WordSegment>, Vec<WordSegment>) {
    let diff = TextDiff::from_words(old, new);
    let mut old_segments = Vec::new();
    let mut new_segments = Vec::new();

    for change in diff.iter_all_changes() {
        let text = change.value().to_string();
        match change.tag() {
            ChangeTag::Equal => {
                old_segments.push(WordSegment {
                    text: text.clone(),
                    kind: WordSegmentKind::Equal,
                });
                new_segments.push(WordSegment {
                    text,
                    kind: WordSegmentKind::Equal,
                });
            }
            ChangeTag::Delete => {
                old_segments.push(WordSegment {
                    text,
                    kind: WordSegmentKind::Changed,
                });
            }
            ChangeTag::Insert => {
                new_segments.push(WordSegment {
                    text,
                    kind: WordSegmentKind::Changed,
                });
            }
        }
    }

    (old_segments, new_segments)
}

/// Truncate a string to approximately `max_len` characters with `" ... "` in the middle.
///
/// If the string is shorter than or equal to `max_len`, returns it unchanged.
/// Uses `char_indices()` for Unicode safety.
pub fn truncate_context(s: &str, max_len: usize) -> String {
    if s.len() <= max_len {
        return s.to_string();
    }

    let ellipsis = " ... ";
    let ellipsis_len = ellipsis.len();

    if max_len <= ellipsis_len + 4 {
        // Too short for meaningful truncation
        return s.chars().take(max_len).collect();
    }

    let half = (max_len - ellipsis_len) / 2;

    // Find the byte index for the prefix (first `half` chars)
    let prefix_end = s
        .char_indices()
        .nth(half)
        .map(|(i, _)| i)
        .unwrap_or(s.len());

    // Find the byte index for the suffix (last `half` chars)
    let total_chars = s.chars().count();
    let suffix_start_char = total_chars.saturating_sub(half);
    let suffix_start = s
        .char_indices()
        .nth(suffix_start_char)
        .map(|(i, _)| i)
        .unwrap_or(s.len());

    format!("{}{}{}", &s[..prefix_end], ellipsis, &s[suffix_start..])
}

/// Normalize text for more readable line-level diffs.
///
/// Long lines (single paragraphs without newlines) are split at sentence
/// boundaries so that the diff can show which sentence changed, rather than
/// showing the entire paragraph as one changed block.
///
/// Lines shorter than the threshold are left unchanged. When no sentence
/// boundaries are found in a long line, it falls back to word-wrapping.
pub fn normalize_for_diff(text: &str) -> String {
    let mut result = String::new();
    let parts: Vec<&str> = text.split('\n').collect();
    let line_count = parts.len();
    for (idx, line) in parts.into_iter().enumerate() {
        // Preserve trailing newline semantics: if the original text ended with
        // \n, the split produces a final empty element that we should skip.
        if line.is_empty() && idx == line_count - 1 {
            break;
        }
        if line.len() <= SENTENCE_SPLIT_THRESHOLD {
            result.push_str(line);
            result.push('\n');
        } else {
            let sentences = split_at_sentence_boundaries(line);
            if sentences.len() <= 1 {
                // No sentence boundaries found — word-wrap instead
                let wrapped = word_wrap(line.trim(), WORD_WRAP_WIDTH);
                for wrap_line in &wrapped {
                    result.push_str(wrap_line);
                    result.push('\n');
                }
            } else {
                for sentence in &sentences {
                    let trimmed = sentence.trim();
                    if trimmed.is_empty() {
                        continue;
                    }
                    if trimmed.len() > SENTENCE_SPLIT_THRESHOLD {
                        let wrapped = word_wrap(trimmed, WORD_WRAP_WIDTH);
                        for wrap_line in &wrapped {
                            result.push_str(wrap_line);
                            result.push('\n');
                        }
                    } else {
                        result.push_str(trimmed);
                        result.push('\n');
                    }
                }
            }
        }
    }
    result
}

/// Split text at sentence boundaries (`. `, `? `, `! ` followed by an
/// uppercase letter). The punctuation stays with the preceding sentence.
fn split_at_sentence_boundaries(text: &str) -> Vec<&str> {
    let bytes = text.as_bytes();
    let mut result = Vec::new();
    let mut start = 0;

    let mut i = 0;
    while i + 2 < bytes.len() {
        if (bytes[i] == b'.' || bytes[i] == b'?' || bytes[i] == b'!')
            && bytes[i + 1] == b' '
            && bytes[i + 2].is_ascii_uppercase()
        {
            // Include the punctuation on this segment, skip the space
            result.push(&text[start..=i]);
            start = i + 2;
            i = start;
        } else {
            i += 1;
        }
    }
    if start < text.len() {
        result.push(&text[start..]);
    }
    result
}

/// Word-wrap text at the given width, breaking at space boundaries.
fn word_wrap(text: &str, width: usize) -> Vec<String> {
    let mut lines = Vec::new();
    let mut current = String::new();

    for word in text.split(' ') {
        if word.is_empty() {
            continue;
        }
        if current.is_empty() {
            current.push_str(word);
        } else if current.len() + 1 + word.len() > width {
            lines.push(current);
            current = word.to_string();
        } else {
            current.push(' ');
            current.push_str(word);
        }
    }
    if !current.is_empty() {
        lines.push(current);
    }
    lines
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn test_is_long_text_short() {
        assert!(!is_long_text("hello"));
    }

    #[test]
    fn test_is_long_text_by_length() {
        let s = "a".repeat(LONG_TEXT_THRESHOLD);
        assert!(is_long_text(&s));
    }

    #[test]
    fn test_is_long_text_multiline() {
        assert!(is_long_text("line one\nline two"));
    }

    #[test]
    fn test_diff_identical() {
        let result = diff_text("hello\n", "hello\n");
        assert_eq!(result.deletions, 0);
        assert_eq!(result.insertions, 0);
        assert!(result.hunks.is_empty());
    }

    #[test]
    fn test_diff_single_line_change() {
        let old = "line one\nline two\nline three\n";
        let new = "line one\nline TWO\nline three\n";
        let result = diff_text(old, new);
        assert_eq!(result.deletions, 1);
        assert_eq!(result.insertions, 1);
        assert_eq!(result.hunks.len(), 1);

        // Should produce a Modified entry (paired delete+insert)
        let lines = &result.hunks[0].lines;
        assert!(lines.iter().any(|l| matches!(l, DiffLine::Modified { .. })));
    }

    #[test]
    fn test_diff_addition() {
        let old = "line one\nline three\n";
        let new = "line one\nline two\nline three\n";
        let result = diff_text(old, new);
        assert_eq!(result.deletions, 0);
        assert_eq!(result.insertions, 1);
    }

    #[test]
    fn test_diff_deletion() {
        let old = "line one\nline two\nline three\n";
        let new = "line one\nline three\n";
        let result = diff_text(old, new);
        assert_eq!(result.deletions, 1);
        assert_eq!(result.insertions, 0);
    }

    #[test]
    fn test_diff_multiple_hunks() {
        let old = "a\nb\nc\nd\ne\nf\ng\nh\ni\nj\n";
        let new = "a\nB\nc\nd\ne\nf\ng\nH\ni\nj\n";
        let result = diff_text(old, new);
        assert_eq!(result.deletions, 2);
        assert_eq!(result.insertions, 2);
        // With 1 context line, changes at positions 2 and 8 (0-indexed)
        // should form separate hunks since they're >2 lines apart
        assert_eq!(result.hunks.len(), 2);
    }

    #[test]
    fn test_diff_context_lines() {
        let old = "a\nb\nc\nd\ne\n";
        let new = "a\nb\nC\nd\ne\n";
        let result = diff_text(old, new);
        assert_eq!(result.hunks.len(), 1);

        let lines = &result.hunks[0].lines;
        // With 1 context line, we should have: b (context), Modified(c->C), d (context)
        assert!(
            lines
                .iter()
                .any(|l| matches!(l, DiffLine::Equal(s) if s.starts_with('b')))
        );
        assert!(lines.iter().any(|l| matches!(l, DiffLine::Modified { .. })));
        assert!(
            lines
                .iter()
                .any(|l| matches!(l, DiffLine::Equal(s) if s.starts_with('d')))
        );
    }

    // === Word-level diff tests ===

    #[test]
    fn test_diff_words_single_word_change() {
        let (old_segs, new_segs) = diff_words("hello world", "hello earth");
        // "hello " is equal, "world"/"earth" is changed
        assert!(
            old_segs
                .iter()
                .any(|s| s.kind == WordSegmentKind::Changed && s.text == "world")
        );
        assert!(
            new_segs
                .iter()
                .any(|s| s.kind == WordSegmentKind::Changed && s.text == "earth")
        );
        assert!(
            old_segs
                .iter()
                .any(|s| s.kind == WordSegmentKind::Equal && s.text.contains("hello"))
        );
    }

    #[test]
    fn test_diff_words_no_change() {
        let (old_segs, new_segs) = diff_words("same text", "same text");
        assert!(old_segs.iter().all(|s| s.kind == WordSegmentKind::Equal));
        assert!(new_segs.iter().all(|s| s.kind == WordSegmentKind::Equal));
    }

    #[test]
    fn test_diff_words_completely_different() {
        let (old_segs, new_segs) = diff_words("foo bar", "baz qux");
        assert!(old_segs.iter().any(|s| s.kind == WordSegmentKind::Changed));
        assert!(new_segs.iter().any(|s| s.kind == WordSegmentKind::Changed));
    }

    // === pair_modifications tests ===

    #[test]
    fn test_pair_modifications_basic() {
        let lines = vec![
            DiffLine::Equal("context\n".to_string()),
            DiffLine::Delete("old line\n".to_string()),
            DiffLine::Insert("new line\n".to_string()),
            DiffLine::Equal("context\n".to_string()),
        ];
        let result = pair_modifications(lines);
        assert_eq!(result.len(), 3); // Equal, Modified, Equal
        assert!(matches!(&result[1], DiffLine::Modified { .. }));
    }

    #[test]
    fn test_pair_modifications_uneven_more_deletes() {
        let lines = vec![
            DiffLine::Delete("old 1\n".to_string()),
            DiffLine::Delete("old 2\n".to_string()),
            DiffLine::Delete("old 3\n".to_string()),
            DiffLine::Insert("new 1\n".to_string()),
            DiffLine::Insert("new 2\n".to_string()),
        ];
        let result = pair_modifications(lines);
        // 2 Modified + 1 leftover Delete
        assert_eq!(result.len(), 3);
        assert!(matches!(&result[0], DiffLine::Modified { .. }));
        assert!(matches!(&result[1], DiffLine::Modified { .. }));
        assert!(matches!(&result[2], DiffLine::Delete(_)));
    }

    #[test]
    fn test_pair_modifications_uneven_more_inserts() {
        let lines = vec![
            DiffLine::Delete("old 1\n".to_string()),
            DiffLine::Insert("new 1\n".to_string()),
            DiffLine::Insert("new 2\n".to_string()),
        ];
        let result = pair_modifications(lines);
        // 1 Modified + 1 leftover Insert
        assert_eq!(result.len(), 2);
        assert!(matches!(&result[0], DiffLine::Modified { .. }));
        assert!(matches!(&result[1], DiffLine::Insert(_)));
    }

    #[test]
    fn test_pair_modifications_no_pairs() {
        let lines = vec![
            DiffLine::Equal("context\n".to_string()),
            DiffLine::Insert("new\n".to_string()),
            DiffLine::Equal("context\n".to_string()),
        ];
        let result = pair_modifications(lines);
        assert_eq!(result.len(), 3);
        assert!(matches!(&result[1], DiffLine::Insert(_)));
    }

    // === truncate_context tests ===

    #[test]
    fn test_truncate_context_short_string() {
        let s = "short string";
        assert_eq!(truncate_context(s, 80), s);
    }

    #[test]
    fn test_truncate_context_exact_length() {
        let s = "a".repeat(80);
        assert_eq!(truncate_context(&s, 80), s);
    }

    #[test]
    fn test_truncate_context_long_string() {
        let s = "a".repeat(200);
        let truncated = truncate_context(&s, 80);
        assert!(truncated.len() <= 85); // allow slight overshoot from char boundaries
        assert!(truncated.contains(" ... "));
    }

    #[test]
    fn test_truncate_context_preserves_ends() {
        let s = "START middle padding that is quite long and should be truncated away END";
        let truncated = truncate_context(&s, 40);
        assert!(truncated.starts_with("START"));
        assert!(truncated.ends_with("END"));
        assert!(truncated.contains(" ... "));
    }

    #[test]
    fn test_truncate_context_unicode() {
        let s = "aaa\u{00e9}\u{00e9}\u{00e9}".repeat(20);
        let truncated = truncate_context(&s, 40);
        // Should not panic on unicode boundaries
        assert!(truncated.len() < s.len());
    }

    // === Modified variant in diff output ===

    #[test]
    fn test_diff_produces_modified_for_paired_changes() {
        let old = "line one\nold paragraph content here\nline three\n";
        let new = "line one\nnew paragraph content here\nline three\n";
        let result = diff_text(old, new);

        let has_modified = result.hunks.iter().any(|h| {
            h.lines
                .iter()
                .any(|l| matches!(l, DiffLine::Modified { .. }))
        });
        assert!(has_modified, "Paired delete+insert should become Modified");
    }

    #[test]
    fn test_modified_has_word_segments() {
        let old = "The IRMA framework requires validation.\n";
        let new = "The revised IRMA framework v2 requires validation.\n";
        let result = diff_text(old, new);

        for hunk in &result.hunks {
            for line in &hunk.lines {
                if let DiffLine::Modified {
                    old_segments,
                    new_segments,
                } = line
                {
                    // Old side should have equal segments and no "revised"/"v2"
                    assert!(
                        old_segments
                            .iter()
                            .any(|s| s.kind == WordSegmentKind::Equal)
                    );
                    // New side should have "revised" and "v2" as changed
                    let new_changed: String = new_segments
                        .iter()
                        .filter(|s| s.kind == WordSegmentKind::Changed)
                        .map(|s| s.text.as_str())
                        .collect();
                    assert!(
                        new_changed.contains("revised") || new_changed.contains("v2"),
                        "Expected changed words in new_segments, got: {}",
                        new_changed
                    );
                    return; // Found and verified
                }
            }
        }
        panic!("Expected a Modified DiffLine");
    }

    // === normalize_for_diff tests ===

    #[test]
    fn test_normalize_short_lines_unchanged() {
        let text = "Line one.\nLine two.\nLine three.\n";
        assert_eq!(normalize_for_diff(text), text);
    }

    #[test]
    fn test_normalize_splits_long_paragraph_at_sentences() {
        let text = "You are a helpful assistant. You should always respond politely. Use formal language when speaking to customers. Never share personal information. Always verify the caller's identity before proceeding.";
        let normalized = normalize_for_diff(text);
        let lines: Vec<&str> = normalized.lines().collect();
        assert!(
            lines.len() >= 4,
            "Expected at least 4 sentence lines, got {}: {:?}",
            lines.len(),
            lines
        );
        assert!(lines[0].ends_with('.'));
        assert!(lines[1].ends_with('.'));
    }

    #[test]
    fn test_normalize_preserves_existing_multiline() {
        let text = "Short line one.\nShort line two.\nShort line three.\n";
        assert_eq!(normalize_for_diff(text), text);
    }

    #[test]
    fn test_normalize_word_wraps_when_no_sentences() {
        // A long line with no sentence boundaries (no ". A" pattern)
        let text = "this is a very long line without any sentence boundaries or uppercase letters after periods and it just keeps going and going and going and going for a really long time";
        let normalized = normalize_for_diff(text);
        let lines: Vec<&str> = normalized.lines().collect();
        assert!(
            lines.len() > 1,
            "Expected word-wrap to split into multiple lines, got: {:?}",
            lines
        );
        for line in &lines {
            assert!(
                line.len() <= WORD_WRAP_WIDTH + 20,
                "Line too long after wrapping: {} chars",
                line.len()
            );
        }
    }

    #[test]
    fn test_normalize_improves_diff_granularity() {
        let old = "You are a helpful assistant. Always respond politely. Use formal language. Never share secrets.";
        let new = "You are a helpful assistant. Always respond very politely. Use formal language. Never share secrets.";
        // Without normalization: 1 delete + 1 insert (whole paragraph)
        let raw_result = diff_text(old, new);
        // With normalization: only the changed sentence differs
        let norm_old = normalize_for_diff(old);
        let norm_new = normalize_for_diff(new);
        let norm_result = diff_text(&norm_old, &norm_new);
        // The normalized version should have fewer changed lines
        assert!(
            norm_result.deletions <= raw_result.deletions,
            "Normalized diff should have same or fewer deletions"
        );
        assert_eq!(norm_result.deletions, 1);
        assert_eq!(norm_result.insertions, 1);
    }
}