rustledger-lsp 0.21.0

//! Linked editing range handler for simultaneous editing.
//!
//! Provides ranges that can be edited together:
//! - Account names: edit all occurrences simultaneously
//! - Currency names: edit all occurrences simultaneously

use lsp_types::{LinkedEditingRangeParams, LinkedEditingRanges, Position, Range};
use rustledger_parser::ParseResult;

use super::utils::{
    LineIndex, PositionEncoding, get_word_at_position, is_account_like, is_currency_like,
};

/// Handle a linked editing range request.
pub fn handle_linked_editing_range(
    params: &LinkedEditingRangeParams,
    source: &str,
    parse_result: &ParseResult,
    encoding: PositionEncoding,
) -> Option<LinkedEditingRanges> {
    let position = params.text_document_position_params.position;
    let line_idx = position.line as usize;
    let lines: Vec<&str> = source.lines().collect();
    let line = lines.get(line_idx)?;

    // Get the word at the cursor position
    let (word, _, _) = get_word_at_position(line, position.character as usize, encoding)?;

    let mut ranges = Vec::new();
    // Build the line index once and share it across collectors —
    // otherwise each posting/directive lookup is an O(n) scan.
    let line_index = LineIndex::new(source, encoding);

    // Check if it's an account
    if is_account_like(&word) {
        collect_account_ranges(parse_result, &line_index, &word, &mut ranges);
    }
    // Check if it's a currency
    else if is_currency_like(&word, parse_result) {
        collect_currency_ranges(parse_result, &line_index, &word, &mut ranges);
    }

    if ranges.is_empty() {
        None
    } else {
        // Account pattern: uppercase start, can contain colons, letters, numbers, hyphens
        let word_pattern = if is_account_like(&word) {
            Some(r"[A-Z][A-Za-z0-9:-]*".to_string())
        } else {
            Some(r"[A-Z][A-Z0-9]*".to_string())
        };

        Some(LinkedEditingRanges {
            ranges,
            word_pattern,
        })
    }
}

/// Collect all linked-edit ranges for an account.
///
/// Walks the parser's `account_occurrences` index (every `ACCOUNT`
/// token with exact source spans) and emits one `Range` per
/// occurrence matching `account`. Same shape as
/// `collect_currency_ranges`. The previous shape walked the typed
/// directives and ran a substring search inside each directive's
/// source bytes, which produced false-positive linked-edit ranges
/// for any account-name fragment textually present in a payee
/// string, STRING-typed metadata value, or comment - so as the
/// user retyped one occurrence, unrelated text in the same
/// directive would mutate in lock-step.
fn collect_account_ranges(
    parse_result: &ParseResult,
    line_index: &LineIndex,
    account: &str,
    ranges: &mut Vec<Range>,
) {
    for occurrence in &parse_result.account_occurrences {
        if occurrence.value != account {
            continue;
        }
        let (start_line, start_col) = line_index.offset_to_position(occurrence.span.start);
        let (end_line, end_col) = line_index.offset_to_position(occurrence.span.end);
        ranges.push(Range {
            start: Position::new(start_line, start_col),
            end: Position::new(end_line, end_col),
        });
    }

    ranges.sort_by(|a, b| {
        a.start
            .line
            .cmp(&b.start.line)
            .then(a.start.character.cmp(&b.start.character))
    });
    ranges.dedup_by(|a, b| a == b);
}

/// Collect all ranges for a currency.
/// Collect linked-edit ranges for a currency.
///
/// Walks the parser's `currency_occurrences` index for exact spans
/// — same pattern as `rename::collect_currency_rename_edits` and
/// siblings. The previous string-search implementation matched
/// currency-code substrings in payee strings, account-name segments,
/// and comments, then included them in the linked-edit range set —
/// so as the user typed, the unrelated text would mutate alongside
/// the real currency tokens.
fn collect_currency_ranges(
    parse_result: &ParseResult,
    line_index: &LineIndex,
    currency: &str,
    ranges: &mut Vec<Range>,
) {
    for occurrence in &parse_result.currency_occurrences {
        if occurrence.value != currency {
            continue;
        }
        let (start_line, start_col) = line_index.offset_to_position(occurrence.span.start);
        let (end_line, end_col) = line_index.offset_to_position(occurrence.span.end);
        ranges.push(Range {
            start: Position::new(start_line, start_col),
            end: Position::new(end_line, end_col),
        });
    }

    // Defensive dedup — see `rename::collect_currency_rename_edits`
    // for the rationale; the parser is forward-advancing so today
    // every occurrence is unique, but the dedup costs nothing and
    // protects against future parser refactors.
    ranges.sort_by(|a, b| {
        a.start
            .line
            .cmp(&b.start.line)
            .then(a.start.character.cmp(&b.start.character))
    });
    ranges.dedup();
}

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

    #[test]
    fn test_linked_editing_account() {
        let source = r#"2024-01-01 open Assets:Bank USD
2024-01-15 * "Coffee"
  Assets:Bank  -5.00 USD
  Expenses:Food
2024-01-31 balance Assets:Bank 100 USD
"#;
        let result = parse(source);
        let uri: lsp_types::Uri = "file:///test.beancount".parse().unwrap();

        let params = LinkedEditingRangeParams {
            text_document_position_params: lsp_types::TextDocumentPositionParams {
                text_document: lsp_types::TextDocumentIdentifier { uri },
                position: Position::new(0, 16), // On "Assets:Bank"
            },
            work_done_progress_params: Default::default(),
        };

        let result = handle_linked_editing_range(&params, source, &result, PositionEncoding::Utf16);
        assert!(result.is_some());

        let ranges = result.unwrap();
        // Should find: open, posting, balance = 3 ranges
        assert_eq!(ranges.ranges.len(), 3);
        // Should have account word pattern
        assert!(ranges.word_pattern.is_some());
    }

    #[test]
    fn test_linked_editing_currency() {
        let source = r#"2024-01-01 open Assets:Bank USD
2024-01-15 * "Coffee"
  Assets:Bank  -5.00 USD
  Expenses:Food  5.00 USD
"#;
        let result = parse(source);
        let uri: lsp_types::Uri = "file:///test.beancount".parse().unwrap();

        let params = LinkedEditingRangeParams {
            text_document_position_params: lsp_types::TextDocumentPositionParams {
                text_document: lsp_types::TextDocumentIdentifier { uri },
                position: Position::new(0, 28), // On "USD"
            },
            work_done_progress_params: Default::default(),
        };

        let result = handle_linked_editing_range(&params, source, &result, PositionEncoding::Utf16);
        assert!(result.is_some());

        let ranges = result.unwrap();
        // Should find USD in: open, posting 1, posting 2 = 3 ranges
        assert_eq!(ranges.ranges.len(), 3);
    }

    /// Regression test for currency linked-editing false positives.
    /// See `rename::test_rename_currency_no_false_positives` for the
    /// fuller rationale.
    ///
    /// Linked editing is especially sensitive to false positives:
    /// each range gets mutated *together* as the user types, so a
    /// false-positive range inside an account name or a payee
    /// string would silently corrupt that text. The AST-driven
    /// `currency_occurrences` walk makes that impossible.
    #[test]
    fn test_linked_editing_currency_no_false_positives() {
        let source = r#"2024-01-01 open Assets:USD-Reserve
2024-01-01 commodity USD
2024-01-15 * "USD-to-EUR transfer"
  Assets:USD-Reserve  -100 USD
  Assets:Bank          100 USD
"#;
        let result = parse(source);
        assert!(
            result.errors.is_empty(),
            "parse errors: {:?}",
            result.errors
        );
        let uri: lsp_types::Uri = "file:///test.beancount".parse().unwrap();

        let params = LinkedEditingRangeParams {
            text_document_position_params: lsp_types::TextDocumentPositionParams {
                text_document: lsp_types::TextDocumentIdentifier { uri },
                position: Position::new(1, 21), // on `USD` of `commodity USD`
            },
            work_done_progress_params: Default::default(),
        };

        let ranges = handle_linked_editing_range(&params, source, &result, PositionEncoding::Utf16)
            .expect("linked editing returns Some");

        // Expected: 3 ranges — `commodity USD`, `-100 USD`,
        // `100 USD`. Bespoke string-search would have produced 5
        // (the payee substring and the account-name substring
        // would have been mutated alongside, corrupting unrelated
        // text as the user typed).
        assert_eq!(
            ranges.ranges.len(),
            3,
            "expected 3 linked-edit ranges, got {}: {:#?}",
            ranges.ranges.len(),
            ranges.ranges
        );
    }

    /// Regression test for account linked-editing false positives -
    /// phase 5.5 of the CST migration (#1262). Linked editing
    /// mutates each range *together* as the user types, so a
    /// false-positive range inside a payee string or comment used
    /// to silently corrupt that text. The CST-backed walk over
    /// `account_occurrences` makes the corruption impossible.
    #[test]
    fn test_linked_editing_account_no_false_positives() {
        let source = r#"2024-01-01 open Assets:Bank USD
2024-01-15 * "Assets:Bank transfer note"
  Assets:Bank  -5.00 USD
    memo: "moved Assets:Bank balance"
  Expenses:Food
; rebalanced Assets:Bank yesterday
"#;
        let result = parse(source);
        assert!(
            result.errors.is_empty(),
            "parse errors: {:?}",
            result.errors
        );
        let uri: lsp_types::Uri = "file:///test.beancount".parse().unwrap();
        let params = LinkedEditingRangeParams {
            text_document_position_params: lsp_types::TextDocumentPositionParams {
                text_document: lsp_types::TextDocumentIdentifier { uri },
                position: Position::new(0, 16), // on `Assets:Bank` of the open
            },
            work_done_progress_params: Default::default(),
        };
        let ranges = handle_linked_editing_range(&params, source, &result, PositionEncoding::Utf16)
            .expect("linked editing returns Some");
        // Expected: 2 ranges - the open's ACCOUNT token and the
        // posting's ACCOUNT token. The substring-search shape
        // would have produced 5 (including 3 phantoms in payee,
        // STRING-typed metadata value, and comment - any of which
        // would corrupt user text as they typed).
        assert_eq!(
            ranges.ranges.len(),
            2,
            "expected 2 linked-edit ranges, got {}: {:#?}",
            ranges.ranges.len(),
            ranges.ranges
        );
        // Pin the lines + range widths so a future regression that
        // emits two ranges on the same line, or two zero-width
        // ranges, would still fail. Linked editing is the most
        // safety-critical of the three handlers: a wrong-position
        // range corrupts user text the moment they start typing.
        let summary: Vec<(u32, u32, u32)> = ranges
            .ranges
            .iter()
            .map(|r| {
                (
                    r.start.line,
                    r.start.character,
                    r.end.character - r.start.character,
                )
            })
            .collect();
        assert_eq!(
            summary,
            vec![(0, 16, 11), (2, 2, 11)],
            "expected (line 0, col 16, width 11) + (line 2, col 2, width 11); got {summary:?}"
        );
    }
}