Add hidden --extension to override inference of source type from file extension (#8373)

## Summary

This PR addresses the incompatibility with `jupyterlab-lsp` +
`python-lsp-ruff` arising from the inference of source type from file
extension, raised in #6847.

In particular it follows the suggestion in
https://github.com/astral-sh/ruff/issues/6847#issuecomment-1765724679 to
specify a mapping from file extension to source type.

The source types are

- python
- pyi
- ipynb

Usage:

```sh
ruff check --no-cache --stdin-filename Untitled.ipynb --extension ipynb:python
```

Unlike the original suggestion, `:` instead of `=` is used to associate
file extensions to language since that is what is used with
`--per-file-ignores` which is an existing option that accepts a mapping.

## Test Plan

2 tests added to `integration_test.rs` to ensure the override works as
expected

---------

Co-authored-by: Charlie Marsh <charlie.r.marsh@gmail.com>

crates/ruff_cli/src/args.rs

@@ -8,8 +8,8 @@ use ruff_linter::line_width::LineLength;
 use ruff_linter::logging::LogLevel;
 use ruff_linter::registry::Rule;
 use ruff_linter::settings::types::{
-    FilePattern, PatternPrefixPair, PerFileIgnore, PreviewMode, PythonVersion, SerializationFormat,
-    UnsafeFixes,
+    ExtensionPair, FilePattern, PatternPrefixPair, PerFileIgnore, PreviewMode, PythonVersion,
+    SerializationFormat, UnsafeFixes,
 };
 use ruff_linter::{RuleParser, RuleSelector, RuleSelectorParser};
 use ruff_workspace::configuration::{Configuration, RuleSelection};
@@ -351,6 +351,9 @@ pub struct CheckCommand {
         conflicts_with = "watch",
     )]
     pub show_settings: bool,
+    /// List of mappings from file extension to language (one of ["python", "ipynb", "pyi"]).
+    #[arg(long, value_delimiter = ',', hide = true)]
+    pub extension: Option<Vec<ExtensionPair>>,
     /// Dev-only argument to show fixes
     #[arg(long, hide = true)]
     pub ecosystem_ci: bool,
@@ -535,6 +538,7 @@ impl CheckCommand {
                 force_exclude: resolve_bool_arg(self.force_exclude, self.no_force_exclude),
                 output_format: self.output_format,
                 show_fixes: resolve_bool_arg(self.show_fixes, self.no_show_fixes),
+                extension: self.extension,
             },
         )
     }
@@ -647,6 +651,7 @@ pub struct CliOverrides {
     pub force_exclude: Option<bool>,
     pub output_format: Option<SerializationFormat>,
     pub show_fixes: Option<bool>,
+    pub extension: Option<Vec<ExtensionPair>>,
 }
 
 impl ConfigurationTransformer for CliOverrides {
@@ -731,6 +736,9 @@ impl ConfigurationTransformer for CliOverrides {
         if let Some(target_version) = &self.target_version {
             config.target_version = Some(*target_version);
         }
+        if let Some(extension) = &self.extension {
+            config.lint.extension = Some(extension.clone().into_iter().collect());
+        }
 
         config
     }

crates/ruff_cli/src/commands/check.rs

@@ -30,6 +30,7 @@ use crate::diagnostics::Diagnostics;
 use crate::panic::catch_unwind;
 
 /// Run the linter over a collection of files.
+#[allow(clippy::too_many_arguments)]
 pub(crate) fn check(
     files: &[PathBuf],
     pyproject_config: &PyprojectConfig,
@@ -184,6 +185,7 @@ pub(crate) fn check(
 
 /// Wraps [`lint_path`](crate::diagnostics::lint_path) in a [`catch_unwind`](std::panic::catch_unwind) and emits
 /// a diagnostic if the linting the file panics.
+#[allow(clippy::too_many_arguments)]
 fn lint_path(
     path: &Path,
     package: Option<&Path>,

crates/ruff_cli/src/diagnostics.rs

@@ -17,13 +17,13 @@ use ruff_linter::logging::DisplayParseError;
 use ruff_linter::message::Message;
 use ruff_linter::pyproject_toml::lint_pyproject_toml;
 use ruff_linter::registry::AsRule;
-use ruff_linter::settings::types::UnsafeFixes;
+use ruff_linter::settings::types::{ExtensionMapping, UnsafeFixes};
 use ruff_linter::settings::{flags, LinterSettings};
 use ruff_linter::source_kind::{SourceError, SourceKind};
 use ruff_linter::{fs, IOError, SyntaxError};
 use ruff_notebook::{Notebook, NotebookError, NotebookIndex};
 use ruff_python_ast::imports::ImportMap;
-use ruff_python_ast::{SourceType, TomlSourceType};
+use ruff_python_ast::{PySourceType, SourceType, TomlSourceType};
 use ruff_source_file::{LineIndex, SourceCode, SourceFileBuilder};
 use ruff_text_size::{TextRange, TextSize};
 use ruff_workspace::Settings;
@@ -177,6 +177,11 @@ impl AddAssign for FixMap {
     }
 }
 
+fn override_source_type(path: Option<&Path>, extension: &ExtensionMapping) -> Option<PySourceType> {
+    let ext = path?.extension()?.to_str()?;
+    extension.get(ext).map(PySourceType::from)
+}
+
 /// Lint the source code at the given `Path`.
 pub(crate) fn lint_path(
     path: &Path,
@@ -221,31 +226,35 @@ pub(crate) fn lint_path(
 
     debug!("Checking: {}", path.display());
 
-    let source_type = match SourceType::from(path) {
-        SourceType::Toml(TomlSourceType::Pyproject) => {
-            let messages = if settings
-                .rules
-                .iter_enabled()
-                .any(|rule_code| rule_code.lint_source().is_pyproject_toml())
-            {
-                let contents = match std::fs::read_to_string(path).map_err(SourceError::from) {
-                    Ok(contents) => contents,
-                    Err(err) => {
-                        return Ok(Diagnostics::from_source_error(&err, Some(path), settings));
-                    }
+    let source_type = match override_source_type(Some(path), &settings.extension) {
+        Some(source_type) => source_type,
+        None => match SourceType::from(path) {
+            SourceType::Toml(TomlSourceType::Pyproject) => {
+                let messages = if settings
+                    .rules
+                    .iter_enabled()
+                    .any(|rule_code| rule_code.lint_source().is_pyproject_toml())
+                {
+                    let contents = match std::fs::read_to_string(path).map_err(SourceError::from) {
+                        Ok(contents) => contents,
+                        Err(err) => {
+                            return Ok(Diagnostics::from_source_error(&err, Some(path), settings));
+                        }
+                    };
+                    let source_file =
+                        SourceFileBuilder::new(path.to_string_lossy(), contents).finish();
+                    lint_pyproject_toml(source_file, settings)
+                } else {
+                    vec![]
                 };
-                let source_file = SourceFileBuilder::new(path.to_string_lossy(), contents).finish();
-                lint_pyproject_toml(source_file, settings)
-            } else {
-                vec![]
-            };
-            return Ok(Diagnostics {
-                messages,
-                ..Diagnostics::default()
-            });
-        }
-        SourceType::Toml(_) => return Ok(Diagnostics::default()),
-        SourceType::Python(source_type) => source_type,
+                return Ok(Diagnostics {
+                    messages,
+                    ..Diagnostics::default()
+                });
+            }
+            SourceType::Toml(_) => return Ok(Diagnostics::default()),
+            SourceType::Python(source_type) => source_type,
+        },
     };
 
     // Extract the sources from the file.
@@ -370,8 +379,15 @@ pub(crate) fn lint_stdin(
     fix_mode: flags::FixMode,
 ) -> Result<Diagnostics> {
     // TODO(charlie): Support `pyproject.toml`.
-    let SourceType::Python(source_type) = path.map(SourceType::from).unwrap_or_default() else {
-        return Ok(Diagnostics::default());
+    let source_type = if let Some(source_type) =
+        override_source_type(path, &settings.linter.extension)
+    {
+        source_type
+    } else {
+        let SourceType::Python(source_type) = path.map(SourceType::from).unwrap_or_default() else {
+            return Ok(Diagnostics::default());
+        };
+        source_type
     };
 
     // Extract the sources from the file.