Add treesitter textobjects (#728)

* Add treesitter textobject queries

Only for Go, Python and Rust for now.

* Add tree-sitter textobjects

Only has functions and class objects as of now.

* Fix tests

* Add docs for tree-sitter textobjects

* Add guide for creating new textobject queries

* Add parameter textobject

Only parameter.inside is implemented now, parameter.around
will probably require custom predicates akin to nvim' `make-range`
since we want to select a trailing comma too (a comma will be
an anonymous node and matching against them doesn't work similar
to named nodes)

* Simplify TextObject cell init

book/src/SUMMARY.md

@@ -8,3 +8,5 @@
   - [Keymap](./keymap.md)
   - [Key Remapping](./remapping.md)
   - [Hooks](./hooks.md)
+- [Guides](./guides/README.md)
+  - [Adding Textobject Queries](./guides/textobject.md)

book/src/guides/README.md

@@ -0,0 +1,4 @@
+# Guides
+
+This section contains guides for adding new language server configurations,
+tree-sitter grammers, textobject queries, etc.

book/src/guides/textobject.md

@@ -0,0 +1,30 @@
+# Adding Textobject Queries
+
+Textobjects that are language specific ([like functions, classes, etc][textobjects])
+require an accompanying tree-sitter grammar and a `textobjects.scm` query file
+to work properly. Tree-sitter allows us to query the source code syntax tree
+and capture specific parts of it. The queries are written in a lisp dialect.
+More information on how to write queries can be found in the [official tree-sitter
+documentation](tree-sitter-queries).
+
+Query files should be placed in `runtime/queries/{language}/textobjects.scm`
+when contributing. Note that to test the query files locally you should put
+them under your local runtime directory (`~/.config/helix/runtime` on Linux
+for example).
+
+The following [captures][tree-sitter-captures] are recognized:
+
+| Capture Name       |
+| ---                |
+| `function.inside`  |
+| `function.around`  |
+| `class.inside`     |
+| `class.around`     |
+| `parameter.inside` |
+
+[Example query files][textobject-examples] can be found in the helix GitHub repository.
+
+[textobjects]: ../usage.md#textobjects
+[tree-sitter-queries]: https://tree-sitter.github.io/tree-sitter/using-parsers#query-syntax
+[tree-sitter-captures]: https://tree-sitter.github.io/tree-sitter/using-parsers#capturing-nodes
+[textobject-examples]: https://github.com/search?q=repo%3Ahelix-editor%2Fhelix+filename%3Atextobjects.scm&type=Code&ref=advsearch&l=&l=

book/src/usage.md

@@ -51,9 +51,10 @@ Multiple characters are currently not supported, but planned.
 
 ## Textobjects
 
-Currently supported: `word`, `surround`.
+Currently supported: `word`, `surround`, `function`, `class`, `parameter`.
 
 ![textobject-demo](https://user-images.githubusercontent.com/23398472/124231131-81a4bb00-db2d-11eb-9d10-8e577ca7b177.gif)
+![textobject-treesitter-demo](https://user-images.githubusercontent.com/23398472/132537398-2a2e0a54-582b-44ab-a77f-eb818942203d.gif)
 
 - `ma` - Select around the object (`va` in vim, `<alt-a>` in kakoune)
 - `mi` - Select inside the object (`vi` in vim, `<alt-i>` in kakoune)
@@ -62,5 +63,11 @@ Currently supported: `word`, `surround`.
 | ---                    | ---                      |
 | `w`                    | Word                     |
 | `(`, `[`, `'`, etc     | Specified surround pairs |
+| `f`                    | Function                 |
+| `c`                    | Class                    |
+| `p`                    | Parameter                |
 
-Textobjects based on treesitter, like `function`, `class`, etc are planned.
+Note: `f`, `c`, etc need a tree-sitter grammar active for the current
+document and a special tree-sitter query file to work properly. [Only
+some grammars](https://github.com/search?q=repo%3Ahelix-editor%2Fhelix+filename%3Atextobjects.scm&type=Code&ref=advsearch&l=&l=)
+currently have the query file implemented. Contributions are welcome !

helix-core/src/indent.rs

@@ -464,6 +464,7 @@ where
                     unit: String::from("    "),
                 }),
                 indent_query: OnceCell::new(),
+                textobject_query: OnceCell::new(),
             }],
         });
 

helix-core/src/syntax.rs

@@ -49,7 +49,7 @@ pub struct Configuration {
 #[serde(rename_all = "kebab-case")]
 pub struct LanguageConfiguration {
     #[serde(rename = "name")]
-    pub(crate) language_id: String,
+    pub language_id: String,
     pub scope: String,           // source.rust
     pub file_types: Vec<String>, // filename ends_with? <Gemfile, rb, etc>
     pub roots: Vec<String>,      // these indicate project roots <.git, Cargo.toml>
@@ -76,6 +76,8 @@ pub struct LanguageConfiguration {
 
     #[serde(skip)]
     pub(crate) indent_query: OnceCell<Option<IndentQuery>>,
+    #[serde(skip)]
+    pub(crate) textobject_query: OnceCell<Option<TextObjectQuery>>,
 }
 
 #[derive(Debug, Serialize, Deserialize)]
@@ -105,6 +107,32 @@ pub struct IndentQuery {
     pub outdent: HashSet<String>,
 }
 
+#[derive(Debug)]
+pub struct TextObjectQuery {
+    pub query: Query,
+}
+
+impl TextObjectQuery {
+    /// Run the query on the given node and return sub nodes which match given
+    /// capture ("function.inside", "class.around", etc).
+    pub fn capture_nodes<'a>(
+        &'a self,
+        capture_name: &str,
+        node: Node<'a>,
+        slice: RopeSlice<'a>,
+        cursor: &'a mut QueryCursor,
+    ) -> Option<impl Iterator<Item = Node<'a>>> {
+        let capture_idx = self.query.capture_index_for_name(capture_name)?;
+        let captures = cursor.captures(&self.query, node, RopeProvider(slice));
+
+        captures
+            .filter_map(move |(mat, idx)| {
+                (mat.captures[idx].index == capture_idx).then(|| mat.captures[idx].node)
+            })
+            .into()
+    }
+}
+
 fn load_runtime_file(language: &str, filename: &str) -> Result<String, std::io::Error> {
     let path = crate::RUNTIME_DIR
         .join("queries")
@@ -153,7 +181,6 @@ impl LanguageConfiguration {
         // highlights_query += "\n(ERROR) @error";
 
         let injections_query = read_query(&language, "injections.scm");
-
         let locals_query = read_query(&language, "locals.scm");
 
         if highlights_query.is_empty() {
@@ -203,6 +230,18 @@ impl LanguageConfiguration {
             .as_ref()
     }
 
+    pub fn textobject_query(&self) -> Option<&TextObjectQuery> {
+        self.textobject_query
+            .get_or_init(|| -> Option<TextObjectQuery> {
+                let lang_name = self.language_id.to_ascii_lowercase();
+                let query_text = read_query(&lang_name, "textobjects.scm");
+                let lang = self.highlight_config.get()?.as_ref()?.language;
+                let query = Query::new(lang, &query_text).ok()?;
+                Some(TextObjectQuery { query })
+            })
+            .as_ref()
+    }
+
     pub fn scope(&self) -> &str {
         &self.scope
     }

helix-core/src/textobject.rs

@@ -1,9 +1,13 @@
+use std::fmt::Display;
+
 use ropey::RopeSlice;
+use tree_sitter::{Node, QueryCursor};
 
 use crate::chars::{categorize_char, char_is_whitespace, CharCategory};
 use crate::graphemes::next_grapheme_boundary;
 use crate::movement::Direction;
 use crate::surround;
+use crate::syntax::LanguageConfiguration;
 use crate::Range;
 
 fn find_word_boundary(slice: RopeSlice, mut pos: usize, direction: Direction) -> usize {
@@ -51,6 +55,15 @@ pub enum TextObject {
     Inside,
 }
 
+impl Display for TextObject {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        f.write_str(match self {
+            Self::Around => "around",
+            Self::Inside => "inside",
+        })
+    }
+}
+
 // count doesn't do anything yet
 pub fn textobject_word(
     slice: RopeSlice,
@@ -108,6 +121,44 @@ pub fn textobject_surround(
         .unwrap_or(range)
 }
 
+/// Transform the given range to select text objects based on tree-sitter.
+/// `object_name` is a query capture base name like "function", "class", etc.
+/// `slice_tree` is the tree-sitter node corresponding to given text slice.
+pub fn textobject_treesitter(
+    slice: RopeSlice,
+    range: Range,
+    textobject: TextObject,
+    object_name: &str,
+    slice_tree: Node,
+    lang_config: &LanguageConfiguration,
+    _count: usize,
+) -> Range {
+    let get_range = move || -> Option<Range> {
+        let byte_pos = slice.char_to_byte(range.cursor(slice));
+
+        let capture_name = format!("{}.{}", object_name, textobject); // eg. function.inner
+        let mut cursor = QueryCursor::new();
+        let node = lang_config
+            .textobject_query()?
+            .capture_nodes(&capture_name, slice_tree, slice, &mut cursor)?
+            .filter(|node| node.byte_range().contains(&byte_pos))
+            .min_by_key(|node| node.byte_range().len())?;
+
+        let len = slice.len_bytes();
+        let start_byte = node.start_byte();
+        let end_byte = node.end_byte();
+        if start_byte >= len || end_byte >= len {
+            return None;
+        }
+
+        let start_char = slice.byte_to_char(start_byte);
+        let end_char = slice.byte_to_char(end_byte);
+
+        Some(Range::new(start_char, end_char))
+    };
+    get_range().unwrap_or(range)
+}
+
 #[cfg(test)]
 mod test {
     use super::TextObject::*;

helix-term/src/commands.rs

@@ -4465,9 +4465,28 @@ fn select_textobject(cx: &mut Context, objtype: textobject::TextObject) {
             let (view, doc) = current!(cx.editor);
             let text = doc.text().slice(..);
 
+            let textobject_treesitter = |obj_name: &str, range: Range| -> Range {
+                let (lang_config, syntax) = match doc.language_config().zip(doc.syntax()) {
+                    Some(t) => t,
+                    None => return range,
+                };
+                textobject::textobject_treesitter(
+                    text,
+                    range,
+                    objtype,
+                    obj_name,
+                    syntax.tree().root_node(),
+                    lang_config,
+                    count,
+                )
+            };
+
             let selection = doc.selection(view.id).clone().transform(|range| {
                 match ch {
                     'w' => textobject::textobject_word(text, range, objtype, count),
+                    'c' => textobject_treesitter("class", range),
+                    'f' => textobject_treesitter("function", range),
+                    'p' => textobject_treesitter("parameter", range),
                     // TODO: cancel new ranges if inconsistent surround matches across lines
                     ch if !ch.is_ascii_alphanumeric() => {
                         textobject::textobject_surround(text, range, objtype, ch, count)

runtime/queries/go/textobjects.scm

@@ -0,0 +1,21 @@
+(function_declaration
+  body: (block)? @function.inside) @function.around
+
+(func_literal
+  (_)? @function.inside) @function.around
+
+(method_declaration
+  body: (block)? @function.inside) @function.around
+
+;; struct and interface declaration as class textobject?
+(type_declaration
+  (type_spec (type_identifier) (struct_type (field_declaration_list (_)?) @class.inside))) @class.around
+
+(type_declaration
+  (type_spec (type_identifier) (interface_type (method_spec_list (_)?) @class.inside))) @class.around
+
+(parameter_list
+  (_) @parameter.inside)
+
+(argument_list
+  (_) @parameter.inside)

runtime/queries/python/textobjects.scm

@@ -0,0 +1,14 @@
+(function_definition
+  body: (block)? @function.inside) @function.around
+
+(class_definition
+  body: (block)? @class.inside) @class.around
+
+(parameters
+  (_) @parameter.inside)
+  
+(lambda_parameters
+  (_) @parameter.inside)
+
+(argument_list
+  (_) @parameter.inside)

runtime/queries/rust/textobjects.scm

@@ -0,0 +1,26 @@
+(function_item
+  body: (_) @function.inside) @function.around
+
+(struct_item
+  body: (_) @class.inside) @class.around
+
+(enum_item
+  body: (_) @class.inside) @class.around
+
+(union_item
+  body: (_) @class.inside) @class.around
+
+(trait_item
+  body: (_) @class.inside) @class.around
+
+(impl_item
+  body: (_) @class.inside) @class.around
+
+(parameters
+  (_) @parameter.inside)
+
+(closure_parameters
+  (_) @parameter.inside)
+
+(arguments
+  (_) @parameter.inside)