openidea/crates/ide-core/src/text.rs

//! Text coordinates used across the IPC boundary.
//!
//! Convention: **all offsets and columns crossing IPC are UTF-16 code units**,
//! matching JavaScript strings, CodeMirror and the LSP default encoding.
//! Lines are 0-based. Conversion to/from byte offsets lives in `ide-text`.

use serde::{Deserialize, Serialize};
use specta::Type;
use std::path::PathBuf;

/// A 0-based line / UTF-16 column position.
#[derive(Clone, Copy, Debug, Default, PartialEq, Eq, PartialOrd, Ord, Hash, Serialize, Deserialize, Type)]
pub struct Pos {
    pub line: u32,
    pub col: u32,
}

impl Pos {
    pub const fn new(line: u32, col: u32) -> Self {
        Self { line, col }
    }
}

/// Half-open range `[start, end)`.
#[derive(Clone, Copy, Debug, Default, PartialEq, Eq, PartialOrd, Ord, Hash, Serialize, Deserialize, Type)]
pub struct Range {
    pub start: Pos,
    pub end: Pos,
}

impl Range {
    pub const fn new(start: Pos, end: Pos) -> Self {
        Self { start, end }
    }
    pub const fn point(p: Pos) -> Self {
        Self { start: p, end: p }
    }
    pub fn is_empty(&self) -> bool {
        self.start == self.end
    }
}

/// A range inside a specific file, or inside a virtual document (library
/// class contents, decompiled code…) that only a language server can provide.
#[derive(Clone, Debug, PartialEq, Eq, Hash, Serialize, Deserialize, Type)]
pub struct Location {
    /// File path; for virtual targets the URI itself (equal to `uri`), which
    /// the UI opens as a read-only virtual document.
    pub path: crate::IpcPath,
    pub range: Range,
    /// Set for non-`file:` targets, e.g. `jdt://contents/…`.
    #[specta(optional)]
    pub uri: Option<String>,
}

/// Whether `s` is a non-`file:` URI (`scheme:` with a scheme of 2+ chars, so
/// Windows drive letters don't count).
pub fn is_virtual_uri(s: &str) -> bool {
    let Some(colon) = s.find(':') else { return false };
    let scheme = &s[..colon];
    scheme.len() >= 2
        && !scheme.eq_ignore_ascii_case("file")
        && scheme.starts_with(|c: char| c.is_ascii_alphabetic())
        && scheme.chars().all(|c| c.is_ascii_alphanumeric() || matches!(c, '+' | '-' | '.'))
}

impl Location {
    pub fn file(path: PathBuf, range: Range) -> Self {
        Self { path: path.into(), range, uri: None }
    }
}

/// Replace the text in `[from, to)` (UTF-16 offsets into the document *before*
/// the change set is applied) with `insert`.
#[derive(Clone, Debug, PartialEq, Eq, Serialize, Deserialize, Type)]
pub struct TextChange {
    pub from: u32,
    pub to: u32,
    pub insert: String,
}

/// Ordered, non-overlapping changes, all expressed in the coordinates of the
/// original document (CodeMirror `ChangeSet` semantics).
#[derive(Clone, Debug, Default, PartialEq, Eq, Serialize, Deserialize, Type)]
pub struct ChangeSet {
    pub changes: Vec<TextChange>,
}

/// An edit expressed in line/column coordinates (LSP `TextEdit`).
#[derive(Clone, Debug, PartialEq, Eq, Serialize, Deserialize, Type)]
pub struct TextEdit {
    pub range: Range,
    pub new_text: String,
}

/// Edits to many files, applied atomically from the user's point of view.
#[derive(Clone, Debug, Default, PartialEq, Eq, Serialize, Deserialize, Type)]
pub struct WorkspaceEdit {
    pub files: Vec<FileEdit>,
}

impl WorkspaceEdit {
    /// Coalesces aliases of the same file before applying edits against its original text.
    /// Resource operations separate batches because later edits may use the new contents.
    pub fn deduplicate(&mut self) {
        let mut files = Vec::new();
        let mut batch = std::collections::HashMap::<std::path::PathBuf, usize>::new();
        let mut resolve_aliases = true;
        for file in std::mem::take(&mut self.files) {
            match file {
                FileEdit::Edit { path, edits } => {
                    let path = if resolve_aliases {
                        dunce::canonicalize(&path).unwrap_or_else(|_| path.to_path_buf())
                    } else {
                        path.to_path_buf()
                    };
                    let index = *batch.entry(path.clone()).or_insert_with(|| {
                        files.push(FileEdit::Edit { path: path.into(), edits: Vec::new() });
                        files.len() - 1
                    });
                    let FileEdit::Edit { edits: own, .. } = &mut files[index] else { unreachable!() };
                    for edit in edits {
                        if !own.contains(&edit) {
                            own.push(edit);
                        }
                    }
                }
                op => {
                    batch.clear();
                    // Later paths refer to a filesystem state the resource operation has yet to create.
                    resolve_aliases = false;
                    files.push(op);
                }
            }
        }
        self.files = files;
    }
}

#[derive(Clone, Debug, PartialEq, Eq, Serialize, Deserialize, Type)]
#[serde(tag = "kind", rename_all = "camelCase")]
pub enum FileEdit {
    Edit { path: crate::IpcPath, edits: Vec<TextEdit> },
    Create { path: crate::IpcPath, overwrite: bool },
    Rename { from: crate::IpcPath, to: crate::IpcPath },
    Delete { path: crate::IpcPath, recursive: bool },
}

#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash, Serialize, Deserialize, Type)]
#[serde(rename_all = "camelCase")]
pub enum Severity {
    Error,
    Warning,
    Info,
    Hint,
}

#[derive(Clone, Debug, PartialEq, Eq, Serialize, Deserialize, Type)]
pub struct Diagnostic {
    pub range: Range,
    pub severity: Severity,
    pub message: String,
    /// Producer, e.g. `rust-analyzer`, `eslint`.
    pub source: Option<String>,
    pub code: Option<String>,
    #[serde(default)]
    pub unnecessary: bool,
    #[serde(default)]
    pub deprecated: bool,
    /// Language server (or tool) that reported it, e.g. `basedpyright`; `source`
    /// is what the producer calls itself. Inspection profiles key on this.
    #[serde(default)]
    #[specta(optional)]
    pub server: Option<String>,
}

/// Symbol kinds, a stable subset of LSP `SymbolKind`.
#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash, Serialize, Deserialize, Type)]
#[serde(rename_all = "camelCase")]
pub enum SymbolKind {
    File,
    Module,
    Namespace,
    Package,
    Class,
    Interface,
    Struct,
    Enum,
    EnumMember,
    Trait,
    Function,
    Method,
    Constructor,
    Field,
    Property,
    Variable,
    Constant,
    TypeParameter,
    TypeAlias,
    Macro,
    Event,
    Operator,
    Key,
    Unknown,
}

impl SymbolKind {
    /// "Class-like" symbols are what IDEA's *Go to Class* (⌘O) searches.
    pub fn is_type_like(self) -> bool {
        matches!(self, Self::Class | Self::Interface | Self::Struct | Self::Enum | Self::Trait | Self::TypeAlias)
    }
}

/// Declared (or conventional) visibility of an outline symbol.
#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash, Serialize, Deserialize, Type)]
#[serde(rename_all = "camelCase")]
pub enum Visibility {
    Public,
    Protected,
    /// Package / module / assembly scope (Java default, `internal`, `pub(crate)`, Go unexported).
    Package,
    Private,
}

/// A node of a document outline (Structure tool window, breadcrumbs).
#[derive(Clone, Debug, PartialEq, Eq, Serialize, Deserialize, Type)]
pub struct OutlineItem {
    pub name: String,
    pub detail: Option<String>,
    pub kind: SymbolKind,
    /// Whole extent of the symbol.
    pub range: Range,
    /// The identifier itself; where navigation lands.
    pub selection_range: Range,
    pub children: Vec<OutlineItem>,
    /// Unknown (`None`) for language-server symbols and languages without the notion.
    #[serde(default)]
    #[specta(optional)]
    pub visibility: Option<Visibility>,
}

/// Fuzzy-match highlight: indices (in chars of the displayed string) that matched.
pub type MatchIndices = Vec<u32>;

#[cfg(all(test, unix))]
mod workspace_edit_tests {
    use super::*;

    #[test]
    fn aliases_are_one_edit_but_resource_operations_preserve_order() {
        let dir = tempfile::tempdir().unwrap();
        let real = dir.path().join("main.cpp");
        let alias = dir.path().join("alias.cpp");
        std::fs::write(&real, "add()").unwrap();
        std::os::unix::fs::symlink(&real, &alias).unwrap();
        let change = TextEdit { range: Range::new(Pos::new(0, 0), Pos::new(0, 3)), new_text: "sum_values".into() };
        let edit = |path: PathBuf| FileEdit::Edit { path: path.into(), edits: vec![change.clone()] };
        let mut workspace = WorkspaceEdit {
            files: vec![
                edit(real.clone()),
                edit(alias),
                FileEdit::Create { path: dir.path().join("other").into(), overwrite: false },
                edit(real),
            ],
        };
        workspace.deduplicate();
        assert_eq!(workspace.files.len(), 3);
        let FileEdit::Edit { edits, .. } = &workspace.files[0] else { panic!() };
        assert_eq!(edits, &[change]);
        let once = workspace.clone();
        workspace.deduplicate();
        assert_eq!(workspace, once);
    }
}