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
This commit is contained in:
Timothy DeHerrera 2023-04-28 03:21:34 -06:00 committed by GitHub
parent 204d1eba4b
commit 9c6c63a2be
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
8 changed files with 216 additions and 84 deletions

View File

@ -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)

View File

@ -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 nodes 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

View File

@ -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(|| {
Regex::new(r"^#!\s*(?:\S*[/\\](?:env\s+(?:\-\S+\s+)*)?)?([^\s\.\d]+)").unwrap()
});
static SHEBANG_REGEX: Lazy<Regex> =
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(
&self,
string: &str,
) -> Option<Arc<LanguageConfiguration>> {
/// Unlike language_config_for_language_id, which only returns Some for an exact id, this
/// function will perform a regex match on the given string to find the closest language match.
pub fn language_config_for_name(&self, name: &str) -> 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 {
let configuration = &self.language_configs[i];
return Some(configuration.clone());
best_match_position.map(|i| self.language_configs[i].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(
&layer.config,
&layer.config.injections_query,
&mat,
source_slice,
);
let (injection_capture, content_node, included_children) = layer
.config
.injection_for_match(&layer.config.injections_query, &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(
&layer.config,
combined_injections_query,
&mat,
source_slice,
);
if language_name.is_some() {
entry.0 = language_name;
let (injection_capture, content_node, included_children) = layer
.config
.injection_for_match(combined_injections_query, &mat, source_slice);
if injection_capture.is_some() {
entry.0 = injection_capture;
}
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>(
config: &HighlightConfiguration,
query: &'a Query,
query_match: &QueryMatch<'a, 'a>,
source: RopeSlice<'a>,
) -> (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)
#[derive(Debug, Clone)]
pub enum InjectionLanguageMarker<'a> {
Name(Cow<'a, str>),
Filename(Cow<'a, Path>),
Shebang(String),
}
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>)>>,

View File

@ -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)));

View File

@ -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"

View File

@ -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)

View File

@ -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: (_)

View File

@ -10,9 +10,11 @@
; such as those of stdenv.mkDerivation.
((binding
attrpath: (attrpath (identifier) @_path)
expression: (indented_string_expression
(string_fragment) @injection.content))
(#match? @_path "(^\\w*Phase|(pre|post)\\w*|(.*\\.)?\\w*([sS]cript|[hH]ook)|(.*\\.)?startup)$")
expression: [
(indented_string_expression (string_fragment) @injection.content)
(binary_expression (indented_string_expression (string_fragment) @injection.content))
])
(#match? @_path "(^\\w*Phase|command|(pre|post)\\w*|(.*\\.)?\\w*([sS]cript|[hH]ook)|(.*\\.)?startup)$")
(#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))