Add support for reading air.toml in the CLI and LSP (#122)

* Add support for reading `air.toml` in the CLI and LSP

* Split discovery from resolver, make workspace open/close handle their own errors

* Use fixed name for toml watcher

Technically you can use this to dynamically unregister at any time

* Remove empty line

* Fix small documentation issue with `crlf`

* Ensure `document_settings()` can determine if fallback settings are in use or not

* Use more obvious enum variant names

* Add in some copyright headers

* One more header

* Rename `LineLength` to `LineWidth` to better match `IndentWidth`

* Nest `line_width` and `indent_width` under `[format]`

Cargo.toml

@@ -32,6 +32,7 @@ line_ending = { path = "./crates/line_ending" }
 lsp = { path = "./crates/lsp" }
 lsp_test = { path = "./crates/lsp_test" }
 tests_macros = { path = "./crates/tests_macros" }
+workspace = { path = "./crates/workspace" }
 
 anyhow = "1.0.89"
 assert_matches = "1.5.0"
@@ -59,14 +60,17 @@ line-index = "0.1.2"
 memchr = "2.7.4"
 path-absolutize = "3.1.1"
 proc-macro2 = "1.0.86"
-serde = { version = "1.0.215", features = ["derive"] }
+rustc-hash = "2.1.0"
+serde = "1.0.215"
 serde_json = "1.0.132"
 struct-field-names-as-array = "0.3.0"
 strum = "0.26"
+tempfile = "3.9.0"
 time = "0.3.37"
 thiserror = "2.0.5"
 tokio = { version = "1.41.1" }
 tokio-util = "0.7.12"
+toml = "0.8.19"
 # For https://github.com/ebkalderon/tower-lsp/pull/428
 tower-lsp = { git = "https://github.com/lionel-/tower-lsp", branch = "bugfix/patches" }
 tracing = { version = "0.1.40", default-features = false, features = ["std"] }
@@ -124,7 +128,6 @@ unnecessary_join = "warn"
 unnested_or_patterns = "warn"
 unreadable_literal = "warn"
 verbose_bit_mask = "warn"
-zero_sized_map_values = "warn"
 
 # restriction
 cfg_not_test = "warn"

crates/air/Cargo.toml

@@ -28,9 +28,10 @@ lsp = { workspace = true }
 thiserror = { workspace = true }
 tokio = "1.41.1"
 tracing = { workspace = true }
+workspace = { workspace = true }
 
 [dev-dependencies]
-tempfile = "3.9.0"
+tempfile = { workspace = true }
 
 [lints]
 workspace = true

crates/air/src/commands/format.rs

@@ -8,23 +8,41 @@ use std::path::PathBuf;
 use air_r_formatter::context::RFormatOptions;
 use air_r_parser::RParserOptions;
 use fs::relativize_path;
-use ignore::DirEntry;
 use itertools::Either;
 use itertools::Itertools;
-use line_ending::LineEnding;
 use thiserror::Error;
+use workspace::discovery::discover_r_file_paths;
+use workspace::discovery::discover_settings;
+use workspace::discovery::DiscoveredSettings;
+use workspace::resolve::PathResolver;
+use workspace::settings::FormatSettings;
+use workspace::settings::Settings;
 
 use crate::args::FormatCommand;
 use crate::ExitStatus;
 
 pub(crate) fn format(command: FormatCommand) -> anyhow::Result<ExitStatus> {
     let mode = FormatMode::from_command(&command);
-    let paths = resolve_paths(&command.paths);
+
+    let paths = discover_r_file_paths(&command.paths);
+
+    let mut resolver = PathResolver::new(Settings::default());
+
+    for DiscoveredSettings {
+        directory,
+        settings,
+    } in discover_settings(&command.paths)?
+    {
+        resolver.add(&directory, settings);
+    }
 
     let (actions, errors): (Vec<_>, Vec<_>) = paths
         .into_iter()
         .map(|path| match path {
-            Ok(path) => format_file(path, mode),
+            Ok(path) => {
+                let settings = resolver.resolve_or_fallback(&path);
+                format_file(path, mode, &settings.format)
+            }
             Err(err) => Err(err.into()),
         })
         .partition_map(|result| match result {
@@ -99,62 +117,6 @@ fn write_changed(actions: &[FormatFileAction], f: &mut impl Write) -> io::Result
     Ok(())
 }
 
-fn resolve_paths(paths: &[PathBuf]) -> Vec<Result<PathBuf, ignore::Error>> {
-    let paths: Vec<PathBuf> = paths.iter().map(fs::normalize_path).collect();
-
-    let (first_path, paths) = paths
-        .split_first()
-        .expect("Clap should ensure at least 1 path is supplied.");
-
-    // TODO: Parallel directory visitor
-    let mut builder = ignore::WalkBuilder::new(first_path);
-
-    for path in paths {
-        builder.add(path);
-    }
-
-    let mut out = Vec::new();
-
-    for path in builder.build() {
-        match path {
-            Ok(entry) => {
-                if let Some(path) = is_valid_path(entry) {
-                    out.push(Ok(path));
-                }
-            }
-            Err(err) => {
-                out.push(Err(err));
-            }
-        }
-    }
-
-    out
-}
-
-// Decide whether or not to accept an `entry` based on include/exclude rules.
-fn is_valid_path(entry: DirEntry) -> Option<PathBuf> {
-    // Ignore directories
-    if entry.file_type().map_or(true, |ft| ft.is_dir()) {
-        return None;
-    }
-
-    // Accept all files that are passed-in directly, even non-R files
-    if entry.depth() == 0 {
-        let path = entry.into_path();
-        return Some(path);
-    }
-
-    // Otherwise check if we should accept this entry
-    // TODO: Many other checks based on user exclude/includes
-    let path = entry.into_path();
-
-    if !fs::has_r_extension(&path) {
-        return None;
-    }
-
-    Some(path)
-}
-
 pub(crate) enum FormatFileAction {
     Formatted(PathBuf),
     Unchanged,
@@ -166,18 +128,15 @@ impl FormatFileAction {
     }
 }
 
-// TODO: Take workspace `FormatOptions` that get resolved to `RFormatOptions`
-// for the formatter here. Respect user specified `LineEnding` option too, and
-// only use inferred endings when `FormatOptions::LineEnding::Auto` is used.
-fn format_file(path: PathBuf, mode: FormatMode) -> Result<FormatFileAction, FormatCommandError> {
+fn format_file(
+    path: PathBuf,
+    mode: FormatMode,
+    settings: &FormatSettings,
+) -> Result<FormatFileAction, FormatCommandError> {
     let source = std::fs::read_to_string(&path)
         .map_err(|err| FormatCommandError::Read(path.clone(), err))?;
 
-    let line_ending = match line_ending::infer(&source) {
-        LineEnding::Lf => biome_formatter::LineEnding::Lf,
-        LineEnding::Crlf => biome_formatter::LineEnding::Crlf,
-    };
-    let options = RFormatOptions::default().with_line_ending(line_ending);
+    let options = settings.to_format_options(&source);
 
     let formatted = match format_source(source.as_str(), options) {
         Ok(formatted) => formatted,

crates/air_r_formatter/src/context.rs

@@ -17,6 +17,7 @@ use biome_formatter::TransformSourceMap;
 use crate::comments::FormatRLeadingComment;
 use crate::comments::RCommentStyle;
 use crate::comments::RComments;
+use crate::options::MagicLineBreak;
 
 pub struct RFormatContext {
     options: RFormatOptions,
@@ -77,6 +78,10 @@ pub struct RFormatOptions {
 
     /// The max width of a line. Defaults to 80.
     line_width: LineWidth,
+
+    // TODO: Actually use this internally!
+    /// The behavior of magic line breaks.
+    magic_line_break: MagicLineBreak,
 }
 
 impl RFormatOptions {
@@ -106,6 +111,11 @@ impl RFormatOptions {
         self
     }
 
+    pub fn with_magic_line_break(mut self, magic_line_break: MagicLineBreak) -> Self {
+        self.magic_line_break = magic_line_break;
+        self
+    }
+
     pub fn set_indent_style(&mut self, indent_style: IndentStyle) {
         self.indent_style = indent_style;
     }
@@ -121,6 +131,10 @@ impl RFormatOptions {
     pub fn set_line_width(&mut self, line_width: LineWidth) {
         self.line_width = line_width;
     }
+
+    pub fn set_magic_line_break(&mut self, magic_line_break: MagicLineBreak) {
+        self.magic_line_break = magic_line_break;
+    }
 }
 
 impl FormatOptions for RFormatOptions {

crates/air_r_formatter/src/lib.rs

@@ -21,6 +21,7 @@ use crate::cst::FormatRSyntaxNode;
 pub mod comments;
 pub mod context;
 mod cst;
+pub mod options;
 mod prelude;
 mod r;
 pub(crate) mod separated;

crates/air_r_formatter/src/options.rs

@@ -0,0 +1,3 @@
+mod magic_line_break;
+
+pub use magic_line_break::*;