snippets: Strip path separators from language snippet file names (#61421)

Closes #59620

## Problem

A language whose name contains a `/`, such as a custom `PL/X` extension
(`lsp_id` → `pl/x`), broke snippets end to end:

- The `snippets: configure snippets` action wrote the file to
`~/.config/zed/snippets/pl/x.json`, i.e. inside a `pl/` subdirectory.
- The snippet scanner reads `snippets/` non-recursively and skips
directories, so that file was never loaded and the snippet could never
be used.
- The completion lookup keyed off the raw `lsp_id` (`pl/x`), which
wouldn't have matched the file-stem key even if the file had been
scanned.

So Zed's own UI created a snippet file it could never read back.

## Fix

Add `LanguageName::snippet_scope_id()` (the `lsp_id` with `/` and `\`
removed) and use it everywhere a language maps to its snippet file name
or lookup key:

- the Configure Snippets writer and its "already configured" label, and
- the two completion lookups in `editor`.

`PL/X` now maps to a flat `plx.json`, as suggested in the issue. The
`editor::InsertSnippet` action is intentionally left unchanged: its
`language` field is documented to be the snippet file name stem, which
is already separator-free.

Note: files created under the old behavior (nested `foo/bar.json`)
aren't migrated; re-running Configure Snippets writes the corrected flat
file.

## Testing

- Added a `language_core` unit test asserting `snippet_scope_id()`
strips `/` and `\` (e.g. `PL/X` → `plx`).
- `cargo test -p language_core` passes.
- `./script/clippy -p language_core -p language` passes.
- Docs Prettier passes.

Release Notes:

- Fixed snippets being unusable for languages whose name contains a `/`
character

crates/editor/src/completions.rs

@@ -1140,7 +1140,7 @@ fn has_strong_snippet_prefix_match(
 
     languages.iter().any(|language| {
         snippet_store
-            .snippets_for(Some(language.lsp_id()), cx)
+            .snippets_for(Some(language.snippet_scope_id()), cx)
             .iter()
             .flat_map(|snippet| snippet.prefix.iter())
             .flat_map(|prefix| snippet_candidate_suffixes(prefix, &is_word_char))
@@ -1161,7 +1161,7 @@ fn snippet_completions(
     let scopes: Vec<_> = languages
         .iter()
         .filter_map(|language| {
-            let language_name = language.lsp_id();
+            let language_name = language.snippet_scope_id();
             let snippets = snippet_store.snippets_for(Some(language_name), cx);
 
             if snippets.is_empty() {

crates/language/src/language.rs

@@ -1165,6 +1165,10 @@ impl Language {
         self.config.name.lsp_id()
     }
 
+    pub fn snippet_scope_id(&self) -> String {
+        self.config.name.snippet_scope_id()
+    }
+
     pub fn prettier_parser_name(&self) -> Option<&str> {
         self.config.prettier_parser_name.as_deref()
     }

crates/language_core/src/language_name.rs

@@ -51,6 +51,16 @@ impl LanguageName {
             language_name => language_name.to_lowercase(),
         }
     }
+
+    /// Identifier used to name and look up this language's snippet file.
+    ///
+    /// Path separators are stripped from the `lsp_id` because the snippet file
+    /// is stored as a flat file directly under the snippets directory; a `/`
+    /// (or `\`) would otherwise place it in a subdirectory that is never
+    /// scanned, making the snippet impossible to use.
+    pub fn snippet_scope_id(&self) -> String {
+        self.lsp_id().replace(['/', '\\'], "")
+    }
 }
 
 impl From<LanguageName> for SharedString {
@@ -107,3 +117,26 @@ impl From<LanguageName> for String {
         Self::from(value)
     }
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn snippet_scope_id_strips_path_separators() {
+        // A `/` or `\` in the language name would route the snippet file into a
+        // subdirectory that is never scanned (see issue #59620).
+        assert_eq!(LanguageName::new("PL/X").snippet_scope_id(), "plx");
+        assert_eq!(LanguageName::new("a\\b").snippet_scope_id(), "ab");
+        // Names without separators are unchanged beyond `lsp_id`'s lowercasing.
+        assert_eq!(LanguageName::new("Rust").snippet_scope_id(), "rust");
+        assert_eq!(
+            LanguageName::new("Shell Script").snippet_scope_id(),
+            "shell script"
+        );
+        assert_eq!(
+            LanguageName::new("Plain Text").snippet_scope_id(),
+            "plaintext"
+        );
+    }
+}

crates/snippets_ui/src/snippets_ui.rs

@@ -220,7 +220,7 @@ impl PickerDelegate for ScopeSelectorDelegate {
                 cx.spawn_in(window, async move |_, cx| {
                     let scope_file_name = ScopeFileName(match scope_name.to_lowercase().as_str() {
                         GLOBAL_SCOPE_NAME => Cow::Borrowed(GLOBAL_SCOPE_FILE_NAME),
-                        _ => Cow::Owned(language.await?.lsp_id()),
+                        _ => Cow::Owned(language.await?.snippet_scope_id()),
                     });
 
                     workspace.update_in(cx, |workspace, window, cx| {
@@ -322,7 +322,7 @@ impl PickerDelegate for ScopeSelectorDelegate {
         let name_label = mat.string.clone();
 
         let scope_name = ScopeName(Cow::Owned(
-            LanguageName::new(&self.candidates[mat.candidate_id].string).lsp_id(),
+            LanguageName::new(&self.candidates[mat.candidate_id].string).snippet_scope_id(),
         ));
         let file_label = if self.existing_scopes.contains(&scope_name) {
             Some(ScopeFileName::from(scope_name).with_extension())

docs/src/snippets.md

@@ -39,6 +39,8 @@ The scope is determined by the language name in lowercase e.g. `python.json` for
 
 To create JSX snippets you have to use `javascript.json` snippets file, instead of `jsx.json`, but this does not apply to TSX and TypeScript which follow the above rule.
 
+Path separators (`/` and `\`) are removed from the file name, so a language named `PL/X` uses `plx.json`.
+
 ## Known Limitations
 
 - Only the first prefix is used when a list of prefixes is passed in.