eure-ls 0.2.0

//! LSP-specific queries that convert to LSP types.

use eure::query::{
    CompletionItem, CompletionKind, DiagnosticMessage, DiagnosticSeverity, GetFileDiagnostics,
    GetSemanticTokens, SemanticToken, TextFile, get_completions, get_hover,
};
use lsp_types::{
    CompletionItem as LspCompletionItem, CompletionItemKind, CompletionTextEdit, Diagnostic,
    DiagnosticSeverity as LspSeverity, Documentation, Hover, HoverContents, MarkupContent,
    MarkupKind, NumberOrString, Position, Range, SemanticToken as LspSemanticToken, SemanticTokens,
    TextEdit,
};
use query_flow::{Db, QueryError, query};

/// LSP-formatted completion.
///
/// Wraps `get_completions` and converts to LSP `CompletionItem`s. `offset` is
/// the cursor position as a byte offset (see [`position_to_offset`]).
///
/// Like `get_completions`, this is a plain function rather than a query so
/// that per-cursor results are not memoized.
pub fn lsp_completion(
    db: &impl Db,
    file: &TextFile,
    offset: u32,
) -> Result<Vec<LspCompletionItem>, QueryError> {
    let items = get_completions(db, file, offset)?;
    let source: std::sync::Arc<eure::query::TextFileContent> = db.asset(file.clone())?;
    let line_offsets = compute_line_offsets(source.get());
    Ok(items
        .iter()
        .map(|item| convert_completion_item(item, source.get(), &line_offsets))
        .collect())
}

fn convert_completion_item(
    item: &CompletionItem,
    source: &str,
    line_offsets: &[usize],
) -> LspCompletionItem {
    let range = Range {
        start: offset_to_lsp_position(item.replace.start as usize, source, line_offsets),
        end: offset_to_lsp_position(item.replace.end as usize, source, line_offsets),
    };
    LspCompletionItem {
        label: item.label.clone(),
        kind: Some(convert_completion_kind(item.kind)),
        detail: item.detail.clone(),
        documentation: item.documentation.as_ref().map(|value| {
            Documentation::MarkupContent(MarkupContent {
                kind: MarkupKind::Markdown,
                value: value.clone(),
            })
        }),
        deprecated: item.deprecated.then_some(true),
        filter_text: Some(item.label.clone()),
        text_edit: Some(CompletionTextEdit::Edit(TextEdit {
            range,
            new_text: item.label.clone(),
        })),
        ..Default::default()
    }
}

fn convert_completion_kind(kind: CompletionKind) -> CompletionItemKind {
    match kind {
        CompletionKind::Field => CompletionItemKind::FIELD,
        CompletionKind::Extension => CompletionItemKind::PROPERTY,
        CompletionKind::Variant => CompletionItemKind::ENUM_MEMBER,
        CompletionKind::Value => CompletionItemKind::VALUE,
    }
}

/// LSP-formatted hover.
///
/// Wraps `get_hover` and converts to an LSP `Hover` with markdown contents
/// and the range of the hovered key or value. `offset` is the cursor
/// position as a byte offset (see [`position_to_offset`]).
///
/// Like `get_hover`, this is a plain function rather than a query so that
/// per-cursor results are not memoized.
pub fn lsp_hover(db: &impl Db, file: &TextFile, offset: u32) -> Result<Option<Hover>, QueryError> {
    let Some(hover) = get_hover(db, file, offset)? else {
        return Ok(None);
    };
    let source: std::sync::Arc<eure::query::TextFileContent> = db.asset(file.clone())?;
    let line_offsets = compute_line_offsets(source.get());
    Ok(Some(Hover {
        contents: HoverContents::Markup(MarkupContent {
            kind: MarkupKind::Markdown,
            value: hover.contents,
        }),
        range: Some(Range {
            start: offset_to_lsp_position(hover.span.start as usize, source.get(), &line_offsets),
            end: offset_to_lsp_position(hover.span.end as usize, source.get(), &line_offsets),
        }),
    }))
}

/// Definition ranges use the target file's own UTF-16 coordinates.
pub fn lsp_definition(
    db: &impl Db,
    file: &TextFile,
    offset: u32,
) -> Result<Vec<lsp_types::LocationLink>, QueryError> {
    let definitions = eure::query::get_definition(db, file, offset)?;
    let source = db.asset(file.clone())?;
    let offsets = compute_line_offsets(source.get());
    definitions
        .into_iter()
        .map(|definition| {
            let target = db.asset(definition.file.clone())?;
            let target_offsets = compute_line_offsets(target.get());
            Ok(lsp_types::LocationLink {
                origin_selection_range: Some(Range {
                    start: offset_to_lsp_position(
                        definition.origin.start as usize,
                        source.get(),
                        &offsets,
                    ),
                    end: offset_to_lsp_position(
                        definition.origin.end as usize,
                        source.get(),
                        &offsets,
                    ),
                }),
                target_uri: crate::uri_utils::text_file_to_uri(&definition.file).parse()?,
                target_range: Range {
                    start: offset_to_lsp_position(
                        definition.range.start as usize,
                        target.get(),
                        &target_offsets,
                    ),
                    end: offset_to_lsp_position(
                        definition.range.end as usize,
                        target.get(),
                        &target_offsets,
                    ),
                },
                target_selection_range: Range {
                    start: offset_to_lsp_position(
                        definition.selection.start as usize,
                        target.get(),
                        &target_offsets,
                    ),
                    end: offset_to_lsp_position(
                        definition.selection.end as usize,
                        target.get(),
                        &target_offsets,
                    ),
                },
            })
        })
        .collect()
}

/// Convert an LSP position (UTF-16 based) to a byte offset in `source`.
///
/// Positions past the end of a line clamp to the line end; positions past
/// the last line clamp to the end of the source.
pub fn position_to_offset(source: &str, position: Position) -> usize {
    let line_offsets = compute_line_offsets(source);
    let Some(&line_start) = line_offsets.get(position.line as usize) else {
        return source.len();
    };
    let line_end = line_offsets
        .get(position.line as usize + 1)
        .map(|&next| next - 1)
        .unwrap_or(source.len());
    let line = &source[line_start..line_end];
    let mut utf16_units = 0u32;
    for (byte_index, c) in line.char_indices() {
        if utf16_units >= position.character {
            return line_start + byte_index;
        }
        utf16_units += c.len_utf16() as u32;
    }
    line_end
}

/// LSP-formatted semantic tokens query.
///
/// Wraps `GetSemanticTokens` and converts to LSP `SemanticTokens` format.
#[query]
pub fn lsp_semantic_tokens(
    db: &impl Db,
    file: TextFile,
    source: String,
) -> Result<SemanticTokens, QueryError> {
    let tokens = db.query(GetSemanticTokens::new(file.clone()))?;
    Ok(convert_tokens(&tokens, &source))
}

/// LSP-formatted diagnostics query, grouped by file.
///
/// Wraps `GetFileDiagnostics` and converts to LSP `Diagnostic` format.
/// Returns diagnostics grouped by file, so that each file can receive
/// its own publishDiagnostics notification.
#[query]
pub fn lsp_diagnostics(
    db: &impl Db,
    file: TextFile,
) -> Result<Vec<(TextFile, Vec<Diagnostic>)>, QueryError> {
    let diagnostics = db.query(GetFileDiagnostics::new(file.clone()))?;

    // Group diagnostics by file
    let mut by_file: std::collections::HashMap<TextFile, Vec<DiagnosticMessage>> =
        std::collections::HashMap::new();

    // Always include the target file so diagnostics are cleared when errors are fixed
    by_file.insert(file, vec![]);

    for d in diagnostics.iter() {
        by_file.entry(d.file.clone()).or_default().push(d.clone());
    }

    // Convert each group to LSP diagnostics using the correct source
    let mut result = Vec::new();
    for (diag_file, file_diagnostics) in by_file {
        let source: std::sync::Arc<eure::query::TextFileContent> = db.asset(diag_file.clone())?;
        let line_offsets = compute_line_offsets(source.get());
        let lsp_diagnostics: Vec<Diagnostic> = file_diagnostics
            .iter()
            .map(|d| convert_diagnostic(d, source.get(), &line_offsets))
            .collect();
        result.push((diag_file, lsp_diagnostics));
    }

    Ok(result)
}

/// LSP-formatted diagnostics for a single file.
///
/// Wraps `GetFileDiagnostics` and converts to LSP `Diagnostic` format.
/// Returns diagnostics only for the specified file.
#[query]
pub fn lsp_file_diagnostics(db: &impl Db, file: TextFile) -> Result<Vec<Diagnostic>, QueryError> {
    let diagnostics = db.query(GetFileDiagnostics::new(file.clone()))?;

    // Get source for position conversion
    let source: std::sync::Arc<eure::query::TextFileContent> = db.asset(file.clone())?;
    let line_offsets = compute_line_offsets(source.get());

    // Convert to LSP diagnostics
    let lsp_diagnostics: Vec<Diagnostic> = diagnostics
        .iter()
        .filter(|d| d.file == file) // Only include diagnostics for this file
        .map(|d| convert_diagnostic(d, source.get(), &line_offsets))
        .collect();

    Ok(lsp_diagnostics)
}

/// Convert internal semantic tokens to LSP format.
///
/// LSP semantic tokens use a delta encoding:
/// - Each token is encoded as (deltaLine, deltaStartChar, length, tokenType, tokenModifiers)
/// - deltaLine is relative to the previous token's line
/// - deltaStartChar is relative to the previous token's start (or line start if on new line)
/// - All character positions and lengths are in UTF-16 code units
fn convert_tokens(tokens: &[SemanticToken], source: &str) -> SemanticTokens {
    let line_offsets = compute_line_offsets(source);

    let mut data = Vec::new();
    let mut prev_line = 0u32;
    let mut prev_start = 0u32;

    for token in tokens {
        let start = token.start as usize;
        let end = start + token.length as usize;
        let (line, char) = offset_to_position(start, source, &line_offsets);
        let length = byte_len_to_utf16_len(source, start, end);

        let delta_line = line - prev_line;
        let delta_start = if delta_line == 0 {
            char - prev_start
        } else {
            char
        };

        data.push(LspSemanticToken {
            delta_line,
            delta_start,
            length,
            token_type: token.token_type as u32,
            token_modifiers_bitset: token.modifiers,
        });

        prev_line = line;
        prev_start = char;
    }

    SemanticTokens {
        result_id: None,
        data,
    }
}

/// Convert internal diagnostic to LSP format.
fn convert_diagnostic(msg: &DiagnosticMessage, source: &str, line_offsets: &[usize]) -> Diagnostic {
    let start = offset_to_lsp_position(msg.start, source, line_offsets);
    let end = offset_to_lsp_position(msg.end, source, line_offsets);

    Diagnostic {
        range: Range { start, end },
        severity: Some(convert_severity(msg.severity)),
        code: msg.code.clone().map(NumberOrString::String),
        code_description: None,
        source: Some("eure".to_string()),
        message: msg.message.clone(),
        related_information: None,
        tags: None,
        data: None,
    }
}

/// Convert internal severity to LSP severity.
fn convert_severity(severity: DiagnosticSeverity) -> LspSeverity {
    match severity {
        DiagnosticSeverity::Error => LspSeverity::ERROR,
        DiagnosticSeverity::Warning => LspSeverity::WARNING,
        DiagnosticSeverity::Info => LspSeverity::INFORMATION,
        DiagnosticSeverity::Hint => LspSeverity::HINT,
    }
}

/// Compute line offsets for a source string.
///
/// Returns a vector where `line_offsets[i]` is the byte offset of line `i`.
fn compute_line_offsets(source: &str) -> Vec<usize> {
    let mut offsets = vec![0];
    for (i, c) in source.char_indices() {
        if c == '\n' {
            offsets.push(i + 1);
        }
    }
    offsets
}

/// Convert a byte offset to (line, character) position.
///
/// Line is 0-indexed. Character is in UTF-16 code units (as required by LSP).
fn offset_to_position(offset: usize, source: &str, line_offsets: &[usize]) -> (u32, u32) {
    let line = line_offsets.iter().rposition(|&o| o <= offset).unwrap_or(0);
    let line_start = line_offsets[line];
    // Count UTF-16 code units from line start to offset
    let end = offset.min(source.len());
    let line_content = &source[line_start..end];
    let utf16_offset: usize = line_content.chars().map(|c| c.len_utf16()).sum();
    (line as u32, utf16_offset as u32)
}

/// Convert a byte offset to LSP Position with UTF-16 character position.
fn offset_to_lsp_position(offset: usize, source: &str, line_offsets: &[usize]) -> Position {
    let (line, character) = offset_to_position(offset, source, line_offsets);
    Position { line, character }
}

/// Convert a byte length to UTF-16 code unit length.
fn byte_len_to_utf16_len(source: &str, start: usize, end: usize) -> u32 {
    let end = end.min(source.len());
    let start = start.min(end);
    source[start..end]
        .chars()
        .map(|c| c.len_utf16())
        .sum::<usize>() as u32
}

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

    #[test]
    fn test_compute_line_offsets() {
        let source = "hello\nworld\n";
        let offsets = compute_line_offsets(source);
        assert_eq!(offsets, vec![0, 6, 12]);
    }

    #[test]
    fn test_offset_to_position_ascii() {
        let source = "hello\nworld\n";
        let offsets = compute_line_offsets(source);
        assert_eq!(offset_to_position(0, source, &offsets), (0, 0));
        assert_eq!(offset_to_position(5, source, &offsets), (0, 5));
        assert_eq!(offset_to_position(6, source, &offsets), (1, 0));
        assert_eq!(offset_to_position(11, source, &offsets), (1, 5));
    }

    #[test]
    fn test_offset_to_position_utf8() {
        // "日本語" is 9 bytes (3 chars × 3 bytes each), but 3 UTF-16 code units
        let source = "日本語\ntest";
        let offsets = compute_line_offsets(source);
        // Byte offset 0 -> (line 0, char 0)
        assert_eq!(offset_to_position(0, source, &offsets), (0, 0));
        // Byte offset 3 (after 日) -> (line 0, char 1)
        assert_eq!(offset_to_position(3, source, &offsets), (0, 1));
        // Byte offset 6 (after 日本) -> (line 0, char 2)
        assert_eq!(offset_to_position(6, source, &offsets), (0, 2));
        // Byte offset 9 (after 日本語) -> (line 0, char 3)
        assert_eq!(offset_to_position(9, source, &offsets), (0, 3));
        // Byte offset 10 (after \n) -> (line 1, char 0)
        assert_eq!(offset_to_position(10, source, &offsets), (1, 0));
    }

    #[test]
    fn test_offset_to_position_emoji() {
        // "😀" is 4 bytes in UTF-8, but 2 UTF-16 code units (surrogate pair)
        let source = "😀a";
        let offsets = compute_line_offsets(source);
        // Byte offset 0 -> (line 0, char 0)
        assert_eq!(offset_to_position(0, source, &offsets), (0, 0));
        // Byte offset 4 (after 😀) -> (line 0, char 2) because emoji is 2 UTF-16 units
        assert_eq!(offset_to_position(4, source, &offsets), (0, 2));
        // Byte offset 5 (after 😀a) -> (line 0, char 3)
        assert_eq!(offset_to_position(5, source, &offsets), (0, 3));
    }

    #[test]
    fn test_position_to_offset() {
        let source = "日本語\ntest";
        assert_eq!(position_to_offset(source, Position::new(0, 0)), 0);
        assert_eq!(position_to_offset(source, Position::new(0, 2)), 6);
        assert_eq!(position_to_offset(source, Position::new(0, 3)), 9);
        // Past the end of the line clamps to the line end (before the newline)
        assert_eq!(position_to_offset(source, Position::new(0, 10)), 9);
        assert_eq!(position_to_offset(source, Position::new(1, 4)), 14);
        // Past the last line clamps to the end of the source
        assert_eq!(position_to_offset(source, Position::new(5, 0)), 14);
        // Surrogate pair counts as two UTF-16 units
        assert_eq!(position_to_offset("😀a", Position::new(0, 2)), 4);
    }

    #[test]
    fn test_byte_len_to_utf16_len() {
        // ASCII: 1 byte = 1 UTF-16 unit
        assert_eq!(byte_len_to_utf16_len("hello", 0, 5), 5);
        // Japanese: 3 bytes per char, 1 UTF-16 unit per char
        assert_eq!(byte_len_to_utf16_len("日本語", 0, 9), 3);
        // Emoji: 4 bytes, 2 UTF-16 units
        assert_eq!(byte_len_to_utf16_len("😀", 0, 4), 2);
    }
}