inject language based on file extension & shebang (#3970)

* inject language based on file extension

Nodes can now be captured with "injection.filename". If this capture
contains a valid file extension known to Helix, then the content will
be highlighted as that language.

* inject language by shebang

Nodes can now be captured with "injection.shebang". If this capture
contains a valid shebang line known to Helix, then the content will
be highlighted as the language the shebang calls for.

* add documentation for language injection

* nix: fix highlights

The `@` is now highlighted properly on either side of the function arg.

Also, extending the phases with `buildPhase = prev.buildPhase + ''''`
is now highlighted properly.

Fix highlighting of `''$` style escapes (requires tree-sitter-nix bump)

Fix `inherit` highlighting.

* simplify injection_for_match

Split out injection pair logic into its own method to make the overall
flow easier to follow.

Also transform the top-level function into a method on a
HighlightConfiguration.

* markdown: add shebang injection query

book/src/SUMMARY.md

@@ -16,3 +16,4 @@ # Summary
   - [Adding languages](./guides/adding_languages.md)
   - [Adding textobject queries](./guides/textobject.md)
   - [Adding indent queries](./guides/indent.md)
+  - [Adding injection queries](./guides/injection.md)

book/src/guides/injection.md

@@ -0,0 +1,57 @@
+# Adding Injection Queries
+
+Writing language injection queries allows one to highlight a specific node as a different language.
+In addition to the [standard](upstream-docs) language injection options used by tree-sitter, there
+are a few Helix specific extensions that allow for more control.
+
+And example of a simple query that would highlight all strings as bash in Nix:
+```scm
+((string_expression (string_fragment) @injection.content)
+  (#set! injection.language "bash"))
+```
+
+## Capture Types
+
+- `@injection.language` (standard):
+The captured node may contain the language name used to highlight the node captured by
+`@injection.content`.
+
+- `@injection.content` (standard):
+Marks the content to be highlighted as the language captured with `@injection.language` _et al_.
+
+- `@injection.filename` (extension):
+The captured node may contain a filename with a file-extension known to Helix,
+highlighting `@injection.content` as that language. This uses the language extensions defined in
+both the default languages.toml distributed with Helix, as well as user defined languages.
+
+- `@injection.shebang` (extension):
+The captured node may contain a shebang used to choose a language to highlight as. This also uses
+the shebangs defined in the default and user `languages.toml`.
+
+## Settings
+
+- `injection.combined` (standard):
+Indicates that all the matching nodes in the tree should have their content parsed as one
+nested document.
+
+- `injection.language` (standard):
+Forces the captured content to be highlighted as the given language
+
+- `injection.include-children` (standard):
+Indicates that the content node’s entire text should be re-parsed, including the text of its child
+nodes. By default, child nodes’ text will be excluded from the injected document.
+
+- `injection.include-unnamed-children` (extension):
+Same as `injection.include-children` but only for unnamed child nodes.
+
+## Predicates
+
+- `#eq?` (standard):
+The first argument (a capture) must be equal to the second argument
+(a capture or a string).
+
+- `#match?` (standard):
+The first argument (a capture) must match the regex given in the
+second argument (a string).
+
+[upstream-docs]: http://tree-sitter.github.io/tree-sitter/syntax-highlighting#language-injection

helix-core/src/syntax.rs

@@ -662,9 +662,8 @@ pub fn language_config_for_file_name(&self, path: &Path) -> Option<Arc<LanguageC
 
     pub fn language_config_for_shebang(&self, source: &Rope) -> Option<Arc<LanguageConfiguration>> {
         let line = Cow::from(source.line(0));
-        static SHEBANG_REGEX: Lazy<Regex> = Lazy::new(|| {
+        static SHEBANG_REGEX: Lazy<Regex> =
-            Regex::new(r"^#!\s*(?:\S*[/\\](?:env\s+(?:\-\S+\s+)*)?)?([^\s\.\d]+)").unwrap()
+            Lazy::new(|| Regex::new(&["^", SHEBANG].concat()).unwrap());
-        });
         let configuration_id = SHEBANG_REGEX
             .captures(&line)
             .and_then(|cap| self.language_config_ids_by_shebang.get(&cap[1]));
@@ -686,15 +685,14 @@ pub fn language_config_for_language_id(&self, id: &str) -> Option<Arc<LanguageCo
             .cloned()
     }
 
-    pub fn language_configuration_for_injection_string(
+    /// Unlike language_config_for_language_id, which only returns Some for an exact id, this
-        &self,
+    /// function will perform a regex match on the given string to find the closest language match.
-        string: &str,
+    pub fn language_config_for_name(&self, name: &str) -> Option<Arc<LanguageConfiguration>> {
-    ) -> Option<Arc<LanguageConfiguration>> {
         let mut best_match_length = 0;
         let mut best_match_position = None;
         for (i, configuration) in self.language_configs.iter().enumerate() {
             if let Some(injection_regex) = &configuration.injection_regex {
-                if let Some(mat) = injection_regex.find(string) {
+                if let Some(mat) = injection_regex.find(name) {
                     let length = mat.end() - mat.start();
                     if length > best_match_length {
                         best_match_position = Some(i);
@@ -704,11 +702,20 @@ pub fn language_configuration_for_injection_string(
             }
         }
 
-        if let Some(i) = best_match_position {
+        best_match_position.map(|i| self.language_configs[i].clone())
-            let configuration = &self.language_configs[i];
+    }
-            return Some(configuration.clone());
+
+    pub fn language_configuration_for_injection_string(
+        &self,
+        capture: &InjectionLanguageMarker,
+    ) -> Option<Arc<LanguageConfiguration>> {
+        match capture {
+            InjectionLanguageMarker::Name(string) => self.language_config_for_name(string),
+            InjectionLanguageMarker::Filename(file) => self.language_config_for_file_name(file),
+            InjectionLanguageMarker::Shebang(shebang) => {
+                self.language_config_for_language_id(shebang)
+            }
         }
-        None
     }
 
     pub fn language_configs(&self) -> impl Iterator<Item = &Arc<LanguageConfiguration>> {
@@ -800,7 +807,7 @@ pub fn update(
         queue.push_back(self.root);
 
         let scopes = self.loader.scopes.load();
-        let injection_callback = |language: &str| {
+        let injection_callback = |language: &InjectionLanguageMarker| {
             self.loader
                 .language_configuration_for_injection_string(language)
                 .and_then(|language_config| language_config.highlight_config(&scopes))
@@ -961,12 +968,9 @@ fn point_sub(a: Point, b: Point) -> Point {
                 );
                 let mut injections = Vec::new();
                 for mat in matches {
-                    let (language_name, content_node, included_children) = injection_for_match(
+                    let (injection_capture, content_node, included_children) = layer
-                        &layer.config,
+                        .config
-                        &layer.config.injections_query,
+                        .injection_for_match(&layer.config.injections_query, &mat, source_slice);
-                        &mat,
-                        source_slice,
-                    );
 
                     // Explicitly remove this match so that none of its other captures will remain
                     // in the stream of captures.
@@ -974,9 +978,10 @@ fn point_sub(a: Point, b: Point) -> Point {
 
                     // If a language is found with the given name, then add a new language layer
                     // to the highlighted document.
-                    if let (Some(language_name), Some(content_node)) = (language_name, content_node)
+                    if let (Some(injection_capture), Some(content_node)) =
+                        (injection_capture, content_node)
                     {
-                        if let Some(config) = (injection_callback)(&language_name) {
+                        if let Some(config) = (injection_callback)(&injection_capture) {
                             let ranges =
                                 intersect_ranges(&layer.ranges, &[content_node], included_children);
 
@@ -1001,14 +1006,11 @@ fn point_sub(a: Point, b: Point) -> Point {
                     );
                     for mat in matches {
                         let entry = &mut injections_by_pattern_index[mat.pattern_index];
-                        let (language_name, content_node, included_children) = injection_for_match(
+                        let (injection_capture, content_node, included_children) = layer
-                            &layer.config,
+                            .config
-                            combined_injections_query,
+                            .injection_for_match(combined_injections_query, &mat, source_slice);
-                            &mat,
+                        if injection_capture.is_some() {
-                            source_slice,
+                            entry.0 = injection_capture;
-                        );
-                        if language_name.is_some() {
-                            entry.0 = language_name;
                         }
                         if let Some(content_node) = content_node {
                             entry.1.push(content_node);
@@ -1395,6 +1397,8 @@ pub struct HighlightConfiguration {
     non_local_variable_patterns: Vec<bool>,
     injection_content_capture_index: Option<u32>,
     injection_language_capture_index: Option<u32>,
+    injection_filename_capture_index: Option<u32>,
+    injection_shebang_capture_index: Option<u32>,
     local_scope_capture_index: Option<u32>,
     local_def_capture_index: Option<u32>,
     local_def_value_capture_index: Option<u32>,
@@ -1538,6 +1542,8 @@ pub fn new(
         // Store the numeric ids for all of the special captures.
         let mut injection_content_capture_index = None;
         let mut injection_language_capture_index = None;
+        let mut injection_filename_capture_index = None;
+        let mut injection_shebang_capture_index = None;
         let mut local_def_capture_index = None;
         let mut local_def_value_capture_index = None;
         let mut local_ref_capture_index = None;
@@ -1558,6 +1564,8 @@ pub fn new(
             match name.as_str() {
                 "injection.content" => injection_content_capture_index = i,
                 "injection.language" => injection_language_capture_index = i,
+                "injection.filename" => injection_filename_capture_index = i,
+                "injection.shebang" => injection_shebang_capture_index = i,
                 _ => {}
             }
         }
@@ -1573,6 +1581,8 @@ pub fn new(
             non_local_variable_patterns,
             injection_content_capture_index,
             injection_language_capture_index,
+            injection_filename_capture_index,
+            injection_shebang_capture_index,
             local_scope_capture_index,
             local_def_capture_index,
             local_def_value_capture_index,
@@ -1631,6 +1641,90 @@ pub fn configure(&self, recognized_names: &[String]) {
 
         self.highlight_indices.store(Arc::new(indices));
     }
+
+    fn injection_pair<'a>(
+        &self,
+        query_match: &QueryMatch<'a, 'a>,
+        source: RopeSlice<'a>,
+    ) -> (Option<InjectionLanguageMarker<'a>>, Option<Node<'a>>) {
+        let mut injection_capture = None;
+        let mut content_node = None;
+
+        for capture in query_match.captures {
+            let index = Some(capture.index);
+            if index == self.injection_language_capture_index {
+                let name = byte_range_to_str(capture.node.byte_range(), source);
+                injection_capture = Some(InjectionLanguageMarker::Name(name));
+            } else if index == self.injection_filename_capture_index {
+                let name = byte_range_to_str(capture.node.byte_range(), source);
+                let path = Path::new(name.as_ref()).to_path_buf();
+                injection_capture = Some(InjectionLanguageMarker::Filename(path.into()));
+            } else if index == self.injection_shebang_capture_index {
+                let node_slice = source.byte_slice(capture.node.byte_range());
+
+                // some languages allow space and newlines before the actual string content
+                // so a shebang could be on either the first or second line
+                let lines = if let Ok(end) = node_slice.try_line_to_byte(2) {
+                    node_slice.byte_slice(..end)
+                } else {
+                    node_slice
+                };
+
+                static SHEBANG_REGEX: Lazy<Regex> = Lazy::new(|| Regex::new(SHEBANG).unwrap());
+
+                injection_capture = SHEBANG_REGEX
+                    .captures(&Cow::from(lines))
+                    .map(|cap| InjectionLanguageMarker::Shebang(cap[1].to_owned()))
+            } else if index == self.injection_content_capture_index {
+                content_node = Some(capture.node);
+            }
+        }
+        (injection_capture, content_node)
+    }
+
+    fn injection_for_match<'a>(
+        &self,
+        query: &'a Query,
+        query_match: &QueryMatch<'a, 'a>,
+        source: RopeSlice<'a>,
+    ) -> (
+        Option<InjectionLanguageMarker<'a>>,
+        Option<Node<'a>>,
+        IncludedChildren,
+    ) {
+        let (mut injection_capture, content_node) = self.injection_pair(query_match, source);
+
+        let mut included_children = IncludedChildren::default();
+        for prop in query.property_settings(query_match.pattern_index) {
+            match prop.key.as_ref() {
+                // In addition to specifying the language name via the text of a
+                // captured node, it can also be hard-coded via a `#set!` predicate
+                // that sets the injection.language key.
+                "injection.language" if injection_capture.is_none() => {
+                    injection_capture = prop
+                        .value
+                        .as_ref()
+                        .map(|s| InjectionLanguageMarker::Name(s.as_ref().into()));
+                }
+
+                // By default, injections do not include the *children* of an
+                // `injection.content` node - only the ranges that belong to the
+                // node itself. This can be changed using a `#set!` predicate that
+                // sets the `injection.include-children` key.
+                "injection.include-children" => included_children = IncludedChildren::All,
+
+                // Some queries might only exclude named children but include unnamed
+                // children in their `injection.content` node. This can be enabled using
+                // a `#set!` predicate that sets the `injection.include-unnamed-children` key.
+                "injection.include-unnamed-children" => {
+                    included_children = IncludedChildren::Unnamed
+                }
+                _ => {}
+            }
+        }
+
+        (injection_capture, content_node, included_children)
+    }
 }
 
 impl<'a> HighlightIterLayer<'a> {
@@ -2042,56 +2136,15 @@ fn next(&mut self) -> Option<Self::Item> {
     }
 }
 
-fn injection_for_match<'a>(
+#[derive(Debug, Clone)]
-    config: &HighlightConfiguration,
+pub enum InjectionLanguageMarker<'a> {
-    query: &'a Query,
+    Name(Cow<'a, str>),
-    query_match: &QueryMatch<'a, 'a>,
+    Filename(Cow<'a, Path>),
-    source: RopeSlice<'a>,
+    Shebang(String),
-) -> (Option<Cow<'a, str>>, Option<Node<'a>>, IncludedChildren) {
-    let content_capture_index = config.injection_content_capture_index;
-    let language_capture_index = config.injection_language_capture_index;
-
-    let mut language_name = None;
-    let mut content_node = None;
-    for capture in query_match.captures {
-        let index = Some(capture.index);
-        if index == language_capture_index {
-            let name = byte_range_to_str(capture.node.byte_range(), source);
-            language_name = Some(name);
-        } else if index == content_capture_index {
-            content_node = Some(capture.node);
-        }
-    }
-
-    let mut included_children = IncludedChildren::default();
-    for prop in query.property_settings(query_match.pattern_index) {
-        match prop.key.as_ref() {
-            // In addition to specifying the language name via the text of a
-            // captured node, it can also be hard-coded via a `#set!` predicate
-            // that sets the injection.language key.
-            "injection.language" => {
-                if language_name.is_none() {
-                    language_name = prop.value.as_ref().map(|s| s.as_ref().into())
-                }
-            }
-
-            // By default, injections do not include the *children* of an
-            // `injection.content` node - only the ranges that belong to the
-            // node itself. This can be changed using a `#set!` predicate that
-            // sets the `injection.include-children` key.
-            "injection.include-children" => included_children = IncludedChildren::All,
-
-            // Some queries might only exclude named children but include unnamed
-            // children in their `injection.content` node. This can be enabled using
-            // a `#set!` predicate that sets the `injection.include-unnamed-children` key.
-            "injection.include-unnamed-children" => included_children = IncludedChildren::Unnamed,
-            _ => {}
-        }
-    }
-
-    (language_name, content_node, included_children)
 }
 
+const SHEBANG: &str = r"#!\s*(?:\S*[/\\](?:env\s+(?:\-\S+\s+)*)?)?([^\s\.\d]+)";
+
 pub struct Merge<I> {
     iter: I,
     spans: Box<dyn Iterator<Item = (usize, std::ops::Range<usize>)>>,

helix-term/src/ui/markdown.rs

@@ -9,7 +9,7 @@
 use pulldown_cmark::{CodeBlockKind, Event, HeadingLevel, Options, Parser, Tag};
 
 use helix_core::{
-    syntax::{self, HighlightEvent, Syntax},
+    syntax::{self, HighlightEvent, InjectionLanguageMarker, Syntax},
     Rope,
 };
 use helix_view::{
@@ -47,7 +47,9 @@ pub fn highlighted_code_block<'a>(
 
     let rope = Rope::from(text.as_ref());
     let syntax = config_loader
-        .language_configuration_for_injection_string(language)
+        .language_configuration_for_injection_string(&InjectionLanguageMarker::Name(
+            language.into(),
+        ))
         .and_then(|config| config.highlight_config(theme.scopes()))
         .map(|config| Syntax::new(&rope, config, Arc::clone(&config_loader)));
 

languages.toml

@@ -591,7 +591,7 @@ indent = { tab-width = 2, unit = "  " }
 
 [[grammar]]
 name = "nix"
-source = { git = "https://github.com/nix-community/tree-sitter-nix", rev = "6b71a810c0acd49b980c50fc79092561f7cee307" }
+source = { git = "https://github.com/nix-community/tree-sitter-nix", rev = "1b69cf1fa92366eefbe6863c184e5d2ece5f187d" }
 
 [[language]]
 name = "ruby"

runtime/queries/markdown/injections.scm

@@ -1,5 +1,9 @@
 ; From nvim-treesitter/nvim-treesitter
 
+(fenced_code_block
+  (code_fence_content) @injection.shebang @injection.content
+  (#set! injection.include-unnamed-children))
+
 (fenced_code_block
   (info_string
     (language) @injection.language)

runtime/queries/nix/highlights.scm

@@ -47,8 +47,10 @@
 (float_expression) @constant.numeric.float
 
 (escape_sequence) @constant.character.escape
+(dollar_escape) @constant.character.escape
 
 (function_expression
+  "@"? @punctuation.delimiter
   universal: (identifier) @variable.parameter
   "@"? @punctuation.delimiter
 )
@@ -82,7 +84,8 @@
 (binding
   attrpath: (attrpath attr: (identifier)) @variable.other.member)
 
-(inherit_from attrs: (inherited_attrs attr: (identifier) @variable))
+(inherit_from attrs: (inherited_attrs attr: (identifier) @variable.other.member))
+(inherited_attrs attr: (identifier) @variable)
 
 (has_attr_expression
   expression: (_)

runtime/queries/nix/injections.scm

@@ -10,9 +10,11 @@
 ; such as those of stdenv.mkDerivation.
 ((binding
    attrpath: (attrpath (identifier) @_path)
-   expression: (indented_string_expression
+   expression: [
-     (string_fragment) @injection.content))
+     (indented_string_expression (string_fragment) @injection.content)
- (#match? @_path "(^\\w*Phase|(pre|post)\\w*|(.*\\.)?\\w*([sS]cript|[hH]ook)|(.*\\.)?startup)$")
+     (binary_expression (indented_string_expression (string_fragment) @injection.content))
+   ])
+ (#match? @_path "(^\\w*Phase|command|(pre|post)\\w*|(.*\\.)?\\w*([sS]cript|[hH]ook)|(.*\\.)?startup)$")
  (#set! injection.language "bash")
  (#set! injection.combined))
 
@@ -150,3 +152,13 @@
 ;  (#match? @_func "(^|\\.)writeFSharp(Bin)?$")
 ;  (#set! injection.language "f-sharp")
 ;  (#set! injection.combined))
+
+((apply_expression
+   function: (apply_expression function: (_) @_func
+     argument: (string_expression (string_fragment) @injection.filename))
+   argument: (indented_string_expression (string_fragment) @injection.content))
+ (#match? @_func "(^|\\.)write(Text|Script(Bin)?)$")
+ (#set! injection.combined))
+
+((indented_string_expression (string_fragment) @injection.shebang @injection.content)
+  (#set! injection.combined))