Support EditorConfig (#13056)

book/src/editor.md

@@ -60,6 +60,7 @@
 | `jump-label-alphabet` | The characters that are used to generate two character jump labels. Characters at the start of the alphabet are used first. | `"abcdefghijklmnopqrstuvwxyz"`
 | `end-of-line-diagnostics` | Minimum severity of diagnostics to render at the end of the line. Set to `disable` to disable entirely. Refer to the setting about `inline-diagnostics` for more details | "disable"
 | `clipboard-provider` | Which API to use for clipboard interaction. One of `pasteboard` (MacOS), `wayland`, `x-clip`, `x-sel`, `win-32-yank`, `termux`, `tmux`, `windows`, `termcode`, `none`, or a custom command set. | Platform and environment specific. |
+| `editor-config` | Whether to read settings from [EditorConfig](https://editorconfig.org) files | `true` |
 
 ### `[editor.clipboard-provider]` Section
 

helix-core/src/editor_config.rs

@@ -0,0 +1,333 @@
+//! Support for [EditorConfig](https://EditorConfig.org) configuration loading.
+//!
+//! EditorConfig is an editor-agnostic format for specifying configuration in an INI-like, human
+//! friendly syntax in `.editorconfig` files (which are intended to be checked into VCS). This
+//! module provides functions to search for all `.editorconfig` files that apply to a given path
+//! and returns an `EditorConfig` type containing any specified configuration options.
+//!
+//! At time of writing, this module follows the [spec](https://spec.editorconfig.org/) at
+//! version 0.17.2.
+
+use std::{
+    collections::HashMap,
+    fs,
+    num::{NonZeroU16, NonZeroU8},
+    path::Path,
+    str::FromStr,
+};
+
+use encoding_rs::Encoding;
+use globset::{GlobBuilder, GlobMatcher};
+
+use crate::{
+    indent::{IndentStyle, MAX_INDENT},
+    LineEnding,
+};
+
+/// Configuration declared for a path in `.editorconfig` files.
+#[derive(Debug, Default, PartialEq, Eq)]
+pub struct EditorConfig {
+    pub indent_style: Option<IndentStyle>,
+    pub tab_width: Option<NonZeroU8>,
+    pub line_ending: Option<LineEnding>,
+    pub encoding: Option<&'static Encoding>,
+    // pub spelling_language: Option<SpellingLanguage>,
+    pub trim_trailing_whitespace: Option<bool>,
+    pub insert_final_newline: Option<bool>,
+    pub max_line_length: Option<NonZeroU16>,
+}
+
+impl EditorConfig {
+    /// Finds any configuration in `.editorconfig` files which applies to the given path.
+    ///
+    /// If no configuration applies then `EditorConfig::default()` is returned.
+    pub fn find(path: &Path) -> Self {
+        let mut configs = Vec::new();
+        // <https://spec.editorconfig.org/#file-processing>
+        for ancestor in path.ancestors() {
+            let editor_config_file = ancestor.join(".editorconfig");
+            let Ok(contents) = fs::read_to_string(&editor_config_file) else {
+                continue;
+            };
+            let ini = match contents.parse::<Ini>() {
+                Ok(ini) => ini,
+                Err(err) => {
+                    log::warn!("Ignoring EditorConfig file at '{editor_config_file:?}' because a glob failed to compile: {err}");
+                    continue;
+                }
+            };
+            let is_root = ini.pairs.get("root").map(AsRef::as_ref) == Some("true");
+            configs.push((ini, ancestor));
+            // > The search shall stop if an EditorConfig file is found with the `root` key set to
+            // > `true` in the preamble or when reaching the root filesystem directory.
+            if is_root {
+                break;
+            }
+        }
+
+        let mut pairs = Pairs::new();
+        // Reverse the configuration stack so that the `.editorconfig` files closest to `path`
+        // are applied last and overwrite settings in files closer to the search ceiling.
+        //
+        // > If multiple EditorConfig files have matching sections, the pairs from the closer
+        // > EditorConfig file are read last, so pairs in closer files take precedence.
+        for (config, dir) in configs.into_iter().rev() {
+            let relative_path = path.strip_prefix(dir).expect("dir is an ancestor of path");
+
+            for section in config.sections {
+                if section.glob.is_match(relative_path) {
+                    log::info!(
+                        "applying EditorConfig from section '{}' in file {:?}",
+                        section.glob.glob(),
+                        dir.join(".editorconfig")
+                    );
+                    pairs.extend(section.pairs);
+                }
+            }
+        }
+
+        Self::from_pairs(pairs)
+    }
+
+    fn from_pairs(pairs: Pairs) -> Self {
+        enum IndentSize {
+            Tab,
+            Spaces(NonZeroU8),
+        }
+
+        // <https://spec.editorconfig.org/#supported-pairs>
+        let indent_size = pairs.get("indent_size").and_then(|value| {
+            if value.as_ref() == "tab" {
+                Some(IndentSize::Tab)
+            } else if let Ok(spaces) = value.parse::<NonZeroU8>() {
+                Some(IndentSize::Spaces(spaces))
+            } else {
+                None
+            }
+        });
+        let tab_width = pairs
+            .get("tab_width")
+            .and_then(|value| value.parse::<NonZeroU8>().ok())
+            .or(match indent_size {
+                Some(IndentSize::Spaces(spaces)) => Some(spaces),
+                _ => None,
+            });
+        let indent_style = pairs
+            .get("indent_style")
+            .and_then(|value| match value.as_ref() {
+                "tab" => Some(IndentStyle::Tabs),
+                "space" => {
+                    let spaces = match indent_size {
+                        Some(IndentSize::Spaces(spaces)) => spaces.get(),
+                        Some(IndentSize::Tab) => tab_width.map(|n| n.get()).unwrap_or(4),
+                        None => 4,
+                    };
+                    Some(IndentStyle::Spaces(spaces.clamp(1, MAX_INDENT)))
+                }
+                _ => None,
+            });
+        let line_ending = pairs
+            .get("end_of_line")
+            .and_then(|value| match value.as_ref() {
+                "lf" => Some(LineEnding::LF),
+                "crlf" => Some(LineEnding::Crlf),
+                #[cfg(feature = "unicode-lines")]
+                "cr" => Some(LineEnding::CR),
+                _ => None,
+            });
+        let encoding = pairs.get("charset").and_then(|value| match value.as_ref() {
+            "latin1" => Some(encoding_rs::WINDOWS_1252),
+            "utf-8" => Some(encoding_rs::UTF_8),
+            // `utf-8-bom` is intentionally ignored.
+            // > `utf-8-bom` is discouraged.
+            "utf-16le" => Some(encoding_rs::UTF_16LE),
+            "utf-16be" => Some(encoding_rs::UTF_16BE),
+            _ => None,
+        });
+        let trim_trailing_whitespace =
+            pairs
+                .get("trim_trailing_whitespace")
+                .and_then(|value| match value.as_ref() {
+                    "true" => Some(true),
+                    "false" => Some(false),
+                    _ => None,
+                });
+        let insert_final_newline = pairs
+            .get("insert_final_newline")
+            .and_then(|value| match value.as_ref() {
+                "true" => Some(true),
+                "false" => Some(false),
+                _ => None,
+            });
+        // This option is not in the spec but is supported by some editors.
+        // <https://github.com/editorconfig/editorconfig/wiki/EditorConfig-Properties#max_line_length>
+        let max_line_length = pairs
+            .get("max_line_length")
+            .and_then(|value| value.parse::<NonZeroU16>().ok());
+
+        Self {
+            indent_style,
+            tab_width,
+            line_ending,
+            encoding,
+            trim_trailing_whitespace,
+            insert_final_newline,
+            max_line_length,
+        }
+    }
+}
+
+type Pairs = HashMap<Box<str>, Box<str>>;
+
+#[derive(Debug)]
+struct Section {
+    glob: GlobMatcher,
+    pairs: Pairs,
+}
+
+#[derive(Debug, Default)]
+struct Ini {
+    pairs: Pairs,
+    sections: Vec<Section>,
+}
+
+impl FromStr for Ini {
+    type Err = globset::Error;
+
+    fn from_str(source: &str) -> Result<Self, Self::Err> {
+        // <https://spec.editorconfig.org/#file-format>
+        let mut ini = Ini::default();
+        // > EditorConfig files are in an INI-like file format. To read an EditorConfig file, take
+        // > one line at a time, from beginning to end. For each line:
+        for full_line in source.lines() {
+            // > 1. Remove all leading and trailing whitespace.
+            let line = full_line.trim();
+            // > 2. Process the remaining text as specified for its type below.
+            // > The types of lines are:
+            // > * Blank: contains nothing. Blank lines are ignored.
+            if line.is_empty() {
+                continue;
+            }
+            // > * Comment: starts with a ';' or '#'. Comment lines are ignored.
+            if line.starts_with([';', '#']) {
+                continue;
+            }
+            if let Some(section) = line.strip_prefix('[').and_then(|s| s.strip_suffix(']')) {
+                // > * Section Header: starts with a `[` and ends with a `]`. These lines define
+                // >   globs...
+
+                // <https://spec.editorconfig.org/#glob-expressions>
+                // We need to modify the glob string slightly since EditorConfig's glob flavor
+                // doesn't match `globset`'s exactly. `globset` only allows '**' at the beginning
+                // or end of a glob or between two '/'s. (This replacement is not very fancy but
+                // should cover most practical cases.)
+                let mut glob_str = section.replace("**.", "**/*.");
+                if !is_glob_relative(section) {
+                    glob_str.insert_str(0, "**/");
+                }
+                let glob = GlobBuilder::new(&glob_str)
+                    .literal_separator(true)
+                    .backslash_escape(true)
+                    .build()?;
+                ini.sections.push(Section {
+                    glob: glob.compile_matcher(),
+                    pairs: Pairs::new(),
+                });
+            } else if let Some((key, value)) = line.split_once('=') {
+                // > * Key-Value Pair (or Pair): contains a key and a value, separated by an `=`.
+                // >     * Key: The part before the first `=` on the line.
+                // >     * Value: The part, if any, after the first `=` on the line.
+                // >     * Keys and values are trimmed of leading and trailing whitespace, but
+                // >       include any whitespace that is between non-whitespace characters.
+                // >     * If a value is not provided, then the value is an empty string.
+                let key = key.trim().to_lowercase().into_boxed_str();
+                let value = value.trim().to_lowercase().into_boxed_str();
+                if let Some(section) = ini.sections.last_mut() {
+                    section.pairs.insert(key, value);
+                } else {
+                    ini.pairs.insert(key, value);
+                }
+            }
+        }
+        Ok(ini)
+    }
+}
+
+/// Determines whether a glob is relative to the directory of the config file.
+fn is_glob_relative(source: &str) -> bool {
+    // > If the glob contains a path separator (a `/` not inside square brackets), then the
+    // > glob is relative to the directory level of the particular `.editorconfig` file itself.
+    let mut idx = 0;
+    while let Some(open) = source[idx..].find('[').map(|open| idx + open) {
+        if source[..open].contains('/') {
+            return true;
+        }
+        idx = source[open..]
+            .find(']')
+            .map_or(source.len(), |close| idx + close);
+    }
+    source[idx..].contains('/')
+}
+
+#[cfg(test)]
+mod test {
+    use super::*;
+
+    #[test]
+    fn is_glob_relative_test() {
+        assert!(is_glob_relative("subdir/*.c"));
+        assert!(!is_glob_relative("*.txt"));
+        assert!(!is_glob_relative("[a/b].c"));
+    }
+
+    fn editor_config(path: impl AsRef<Path>, source: &str) -> EditorConfig {
+        let path = path.as_ref();
+        let ini = source.parse::<Ini>().unwrap();
+        let pairs = ini
+            .sections
+            .into_iter()
+            .filter(|section| section.glob.is_match(path))
+            .fold(Pairs::new(), |mut acc, section| {
+                acc.extend(section.pairs);
+                acc
+            });
+        EditorConfig::from_pairs(pairs)
+    }
+
+    #[test]
+    fn parse_test() {
+        let source = r#"
+        [*]
+        indent_style = space
+
+        [Makefile]
+        indent_style = tab
+
+        [docs/**.txt]
+        insert_final_newline = true
+        "#;
+
+        assert_eq!(
+            editor_config("a.txt", source),
+            EditorConfig {
+                indent_style: Some(IndentStyle::Spaces(4)),
+                ..Default::default()
+            }
+        );
+        assert_eq!(
+            editor_config("pkg/Makefile", source),
+            EditorConfig {
+                indent_style: Some(IndentStyle::Tabs),
+                ..Default::default()
+            }
+        );
+        assert_eq!(
+            editor_config("docs/config/editor.txt", source),
+            EditorConfig {
+                indent_style: Some(IndentStyle::Spaces(4)),
+                insert_final_newline: Some(true),
+                ..Default::default()
+            }
+        );
+    }
+}

helix-core/src/lib.rs

@@ -10,6 +10,7 @@ pub mod config;
 pub mod diagnostic;
 pub mod diff;
 pub mod doc_formatter;
+pub mod editor_config;
 pub mod fuzzy;
 pub mod graphemes;
 pub mod history;

helix-term/src/commands/typed.rs

@@ -326,13 +326,13 @@ fn write_impl(cx: &mut compositor::Context, path: Option<&str>, force: bool) ->
     let jobs = &mut cx.jobs;
     let (view, doc) = current!(cx.editor);
 
-    if config.trim_trailing_whitespace {
+    if doc.trim_trailing_whitespace() {
         trim_trailing_whitespace(doc, view.id);
     }
     if config.trim_final_newlines {
         trim_final_newlines(doc, view.id);
     }
-    if config.insert_final_newline {
+    if doc.insert_final_newline() {
         insert_final_newline(doc, view.id);
     }
 
@@ -738,13 +738,13 @@ pub fn write_all_impl(
         let doc = doc_mut!(cx.editor, &doc_id);
         let view = view_mut!(cx.editor, target_view);
 
-        if config.trim_trailing_whitespace {
+        if doc.trim_trailing_whitespace() {
             trim_trailing_whitespace(doc, target_view);
         }
         if config.trim_final_newlines {
             trim_final_newlines(doc, target_view);
         }
-        if config.insert_final_newline {
+        if doc.insert_final_newline() {
             insert_final_newline(doc, target_view);
         }
 
@@ -2146,7 +2146,6 @@ fn reflow(cx: &mut compositor::Context, args: Args, event: PromptEvent) -> anyho
     }
 
     let scrolloff = cx.editor.config().scrolloff;
-    let cfg_text_width: usize = cx.editor.config().text_width;
     let (view, doc) = current!(cx.editor);
 
     // Find the text_width by checking the following sources in order:
@@ -2157,8 +2156,7 @@ fn reflow(cx: &mut compositor::Context, args: Args, event: PromptEvent) -> anyho
         .first()
         .map(|num| num.parse::<usize>())
         .transpose()?
-        .or_else(|| doc.language_config().and_then(|config| config.text_width))
-        .unwrap_or(cfg_text_width);
+        .unwrap_or_else(|| doc.text_width());
 
     let rope = doc.text();
 

helix-view/src/document.rs

@@ -31,6 +31,7 @@ use std::sync::{Arc, Weak};
 use std::time::SystemTime;
 
 use helix_core::{
+    editor_config::EditorConfig,
     encoding,
     history::{History, State, UndoKind},
     indent::{auto_detect_indent_style, IndentStyle},
@@ -50,6 +51,7 @@ use crate::{
 const BUF_SIZE: usize = 8192;
 
 const DEFAULT_INDENT: IndentStyle = IndentStyle::Tabs;
+const DEFAULT_TAB_WIDTH: usize = 4;
 
 pub const DEFAULT_LANGUAGE_NAME: &str = "text";
 
@@ -157,6 +159,7 @@ pub struct Document {
 
     /// Current indent style.
     pub indent_style: IndentStyle,
+    editor_config: EditorConfig,
 
     /// The document's default line ending.
     pub line_ending: LineEnding,
@@ -678,6 +681,7 @@ impl Document {
             inlay_hints_oudated: false,
             view_data: Default::default(),
             indent_style: DEFAULT_INDENT,
+            editor_config: EditorConfig::default(),
             line_ending,
             restore_cursor: false,
             syntax: None,
@@ -712,7 +716,7 @@ impl Document {
     /// overwritten with the `encoding` parameter.
     pub fn open(
         path: &Path,
-        encoding: Option<&'static Encoding>,
+        mut encoding: Option<&'static Encoding>,
         config_loader: Option<Arc<ArcSwap<syntax::Loader>>>,
         config: Arc<dyn DynAccess<Config>>,
     ) -> Result<Self, DocumentOpenError> {
@@ -721,12 +725,21 @@ impl Document {
             return Err(DocumentOpenError::IrregularFile);
         }
 
+        let editor_config = if config.load().editor_config {
+            EditorConfig::find(path)
+        } else {
+            EditorConfig::default()
+        };
+        encoding = encoding.or(editor_config.encoding);
+
         // Open the file if it exists, otherwise assume it is a new file (and thus empty).
         let (rope, encoding, has_bom) = if path.exists() {
             let mut file = std::fs::File::open(path)?;
             from_reader(&mut file, encoding)?
         } else {
-            let line_ending: LineEnding = config.load().default_line_ending.into();
+            let line_ending = editor_config
+                .line_ending
+                .unwrap_or_else(|| config.load().default_line_ending.into());
             let encoding = encoding.unwrap_or(encoding::UTF_8);
             (Rope::from(line_ending.as_str()), encoding, false)
         };
@@ -739,6 +752,7 @@ impl Document {
             doc.detect_language(loader);
         }
 
+        doc.editor_config = editor_config;
         doc.detect_indent_and_line_ending();
 
         Ok(doc)
@@ -1090,16 +1104,32 @@ impl Document {
     /// configured in `languages.toml`, with a fallback to tabs if it isn't specified. Line ending
     /// is likewise auto-detected, and will remain unchanged if no line endings were detected.
     pub fn detect_indent_and_line_ending(&mut self) {
-        self.indent_style = auto_detect_indent_style(&self.text).unwrap_or_else(|| {
-            self.language_config()
-                .and_then(|config| config.indent.as_ref())
-                .map_or(DEFAULT_INDENT, |config| IndentStyle::from_str(&config.unit))
-        });
-        if let Some(line_ending) = auto_detect_line_ending(&self.text) {
+        self.indent_style = if let Some(indent_style) = self.editor_config.indent_style {
+            indent_style
+        } else {
+            auto_detect_indent_style(&self.text).unwrap_or_else(|| {
+                self.language_config()
+                    .and_then(|config| config.indent.as_ref())
+                    .map_or(DEFAULT_INDENT, |config| IndentStyle::from_str(&config.unit))
+            })
+        };
+        if let Some(line_ending) = self
+            .editor_config
+            .line_ending
+            .or_else(|| auto_detect_line_ending(&self.text))
+        {
             self.line_ending = line_ending;
         }
     }
 
+    pub(crate) fn detect_editor_config(&mut self) {
+        if self.config.load().editor_config {
+            if let Some(path) = self.path.as_ref() {
+                self.editor_config = EditorConfig::find(path);
+            }
+        }
+    }
+
     pub fn pickup_last_saved_time(&mut self) {
         self.last_saved_time = match self.path() {
             Some(path) => match path.metadata() {
@@ -1819,9 +1849,14 @@ impl Document {
 
     /// The width that the tab character is rendered at
     pub fn tab_width(&self) -> usize {
-        self.language_config()
-            .and_then(|config| config.indent.as_ref())
-            .map_or(4, |config| config.tab_width) // fallback to 4 columns
+        self.editor_config
+            .tab_width
+            .map(|n| n.get() as usize)
+            .unwrap_or_else(|| {
+                self.language_config()
+                    .and_then(|config| config.indent.as_ref())
+                    .map_or(DEFAULT_TAB_WIDTH, |config| config.tab_width)
+            })
     }
 
     // The width (in spaces) of a level of indentation.
@@ -1829,6 +1864,20 @@ impl Document {
         self.indent_style.indent_width(self.tab_width())
     }
 
+    /// Whether the document should have a trailing line ending appended on save.
+    pub fn insert_final_newline(&self) -> bool {
+        self.editor_config
+            .insert_final_newline
+            .unwrap_or_else(|| self.config.load().insert_final_newline)
+    }
+
+    /// Whether the document should trim whitespace preceding line endings on save.
+    pub fn trim_trailing_whitespace(&self) -> bool {
+        self.editor_config
+            .trim_trailing_whitespace
+            .unwrap_or_else(|| self.config.load().trim_trailing_whitespace)
+    }
+
     pub fn changes(&self) -> &ChangeSet {
         &self.changes
     }
@@ -2087,12 +2136,17 @@ impl Document {
         }
     }
 
+    pub fn text_width(&self) -> usize {
+        self.editor_config
+            .max_line_length
+            .map(|n| n.get() as usize)
+            .or_else(|| self.language_config().and_then(|config| config.text_width))
+            .unwrap_or_else(|| self.config.load().text_width)
+    }
+
     pub fn text_format(&self, mut viewport_width: u16, theme: Option<&Theme>) -> TextFormat {
         let config = self.config.load();
-        let text_width = self
-            .language_config()
-            .and_then(|config| config.text_width)
-            .unwrap_or(config.text_width);
+        let text_width = self.text_width();
         let mut soft_wrap_at_text_width = self
             .language_config()
             .and_then(|config| {

helix-view/src/editor.rs

@@ -366,6 +366,9 @@ pub struct Config {
     pub end_of_line_diagnostics: DiagnosticFilter,
     // Set to override the default clipboard provider
     pub clipboard_provider: ClipboardProvider,
+    /// Whether to read settings from [EditorConfig](https://editorconfig.org) files. Defaults to
+    /// `true`.
+    pub editor_config: bool,
 }
 
 #[derive(Debug, Clone, PartialEq, Deserialize, Serialize, Eq, PartialOrd, Ord)]
@@ -1009,6 +1012,7 @@ impl Default for Config {
             inline_diagnostics: InlineDiagnosticsConfig::default(),
             end_of_line_diagnostics: DiagnosticFilter::Disable,
             clipboard_provider: ClipboardProvider::default(),
+            editor_config: true,
         }
     }
 }
@@ -1456,6 +1460,7 @@ impl Editor {
         // we have fully unregistered this document from its LS
         doc.language_servers.clear();
         doc.set_path(Some(path));
+        doc.detect_editor_config();
         self.refresh_doc_language(doc_id)
     }
 
@@ -1463,6 +1468,7 @@ impl Editor {
         let loader = self.syn_loader.clone();
         let doc = doc_mut!(self, &doc_id);
         doc.detect_language(loader);
+        doc.detect_editor_config();
         doc.detect_indent_and_line_ending();
         self.refresh_language_servers(doc_id);
         let doc = doc_mut!(self, &doc_id);