Yehowshua/ZIm
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

language: Add context-aware decrease indent for Python (#33370)

Browse source 
Closes #33238, follow-up to
https://github.com/zed-industries/zed/pull/29625.

Changes:

- Removed `significant_indentation`, which was the way to introduce
indentation scoping in languages like Python. However, it turned out to
be unnecessarily complicated to define and maintain.
- Introduced `decrease_indent_patterns`, which takes a `pattern` keyword
to automatically outdent and `valid_after` keywords to treat as valid
code points to snap to. The outdent happens to the most recent
`valid_after` keyword that also has less or equal indentation than the
currently typed keyword.

Fixes:

1. In Python, typing `except`, `finally`, `else`, and so on now
automatically indents intelligently based on the context in which it
appears. For instance:

```py
try:
    if a == 1:
        try:
             b = 2
             ^  # <-- typing "except:" here would indent it to inner try block
```

but,

```py
try:
    if a == 1:
        try:
             b = 2
    ^  # <-- typing "except:" here would indent it to outer try block
```

2. Fixes comments not maintaining indent.

Release Notes:

- Improved auto outdent for Python while typing keywords like `except`,
`else`, `finally`, etc.
- Fixed the issue where comments in Python would not maintain their
indentation.
This commit is contained in:
Smit Barmase 2025-06-26 11:11:03 +05:30 • committed by GitHub
parent 1753432406
commit d09c7eb317
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
6 changed files with 211 additions and 181 deletions
Download patch file Download diff file Expand all files Collapse all files

102
crates/language/src/buffer.rs
View file

@ -2913,7 +2913,12 @@ impl BufferSnapshot {
) -> Option<impl Iterator<Item = Option<IndentSuggestion>> + '_> {
let config = &self.language.as_ref()?.config;
let prev_non_blank_row = self.prev_non_blank_row(row_range.start);
let significant_indentation = config.significant_indentation;
#[derive(Debug, Clone)]
struct StartPosition {
start: Point,
suffix: SharedString,
}
// Find the suggested indentation ranges based on the syntax tree.
let start = Point::new(prev_non_blank_row.unwrap_or(row_range.start), 0);
@ -2929,13 +2934,13 @@ impl BufferSnapshot {
.collect::<Vec<_>>();
let mut indent_ranges = Vec::<Range<Point>>::new();
let mut start_positions = Vec::<StartPosition>::new();
let mut outdent_positions = Vec::<Point>::new();
while let Some(mat) = matches.peek() {
let mut start: Option<Point> = None;
let mut end: Option<Point> = None;
let mut outdent: Option<Point> = None;
let config = &indent_configs[mat.grammar_index];
let config = indent_configs[mat.grammar_index];
for capture in mat.captures {
if capture.index == config.indent_capture_ix {
start.get_or_insert(Point::from_ts_point(capture.node.start_position()));
@ -2945,21 +2950,18 @@ impl BufferSnapshot {
} else if Some(capture.index) == config.end_capture_ix {
end = Some(Point::from_ts_point(capture.node.start_position()));
} else if Some(capture.index) == config.outdent_capture_ix {
let point = Point::from_ts_point(capture.node.start_position());
outdent.get_or_insert(point);
outdent_positions.push(point);
outdent_positions.push(Point::from_ts_point(capture.node.start_position()));
} else if let Some(suffix) = config.suffixed_start_captures.get(&capture.index) {
start_positions.push(StartPosition {
start: Point::from_ts_point(capture.node.start_position()),
suffix: suffix.clone(),
});
}
}
matches.advance();
// in case of significant indentation expand end to outdent position
let end = if significant_indentation {
outdent.or(end)
} else {
end
};
if let Some((start, end)) = start.zip(end) {
if start.row == end.row && (!significant_indentation || start.column < end.column) {
if start.row == end.row {
continue;
}
let range = start..end;
@ -2997,24 +2999,26 @@ impl BufferSnapshot {
matches.advance();
}
// we don't use outdent positions to truncate in case of significant indentation
// rather we use them to expand (handled above)
if !significant_indentation {
outdent_positions.sort();
for outdent_position in outdent_positions {
// find the innermost indent range containing this outdent_position
// set its end to the outdent position
if let Some(range_to_truncate) = indent_ranges
.iter_mut()
.filter(|indent_range| indent_range.contains(&outdent_position))
.next_back()
{
range_to_truncate.end = outdent_position;
}
outdent_positions.sort();
for outdent_position in outdent_positions {
// find the innermost indent range containing this outdent_position
// set its end to the outdent position
if let Some(range_to_truncate) = indent_ranges
.iter_mut()
.filter(|indent_range| indent_range.contains(&outdent_position))
.next_back()
{
range_to_truncate.end = outdent_position;
}
}
start_positions.sort_by_key(|b| b.start);
// Find the suggested indentation increases and decreased based on regexes.
let mut regex_outdent_map = HashMap::default();
let mut last_seen_suffix: HashMap<String, Vec<Point>> = HashMap::default();
let mut start_positions_iter = start_positions.iter().peekable();
let mut indent_change_rows = Vec::<(u32, Ordering)>::new();
self.for_each_line(
Point::new(prev_non_blank_row.unwrap_or(row_range.start), 0)
@ -3034,6 +3038,33 @@ impl BufferSnapshot {
{
indent_change_rows.push((row + 1, Ordering::Greater));
}
while let Some(pos) = start_positions_iter.peek() {
if pos.start.row < row {
let pos = start_positions_iter.next().unwrap();
last_seen_suffix
.entry(pos.suffix.to_string())
.or_default()
.push(pos.start);
} else {
break;
}
}
for rule in &config.decrease_indent_patterns {
if rule.pattern.as_ref().map_or(false, |r| r.is_match(line)) {
let row_start_column = self.indent_size_for_line(row).len;
let basis_row = rule
.valid_after
.iter()
.filter_map(|valid_suffix| last_seen_suffix.get(valid_suffix))
.flatten()
.filter(|start_point| start_point.column <= row_start_column)
.max_by_key(|start_point| start_point.row);
if let Some(outdent_to_row) = basis_row {
regex_outdent_map.insert(row, outdent_to_row.row);
}
break;
}
}
},
);
@ -3043,6 +3074,7 @@ impl BufferSnapshot {
} else {
row_range.start.saturating_sub(1)
};
let mut prev_row_start = Point::new(prev_row, self.indent_size_for_line(prev_row).len);
Some(row_range.map(move |row| {
let row_start = Point::new(row, self.indent_size_for_line(row).len);
@ -3080,17 +3112,17 @@ impl BufferSnapshot {
if range.start.row == prev_row && range.end > row_start {
indent_from_prev_row = true;
}
if significant_indentation && self.is_line_blank(row) && range.start.row == prev_row
{
indent_from_prev_row = true;
}
if !significant_indentation || !self.is_line_blank(row) {
if range.end > prev_row_start && range.end <= row_start {
outdent_to_row = outdent_to_row.min(range.start.row);
}
if range.end > prev_row_start && range.end <= row_start {
outdent_to_row = outdent_to_row.min(range.start.row);
}
}
if let Some(basis_row) = regex_outdent_map.get(&row) {
indent_from_prev_row = false;
outdent_to_row = *basis_row;
from_regex = true;
}
let within_error = error_ranges
.iter()
.any(|e| e.start.row < row && e.end > row_start);

31
crates/language/src/language.rs
View file

@ -696,10 +696,6 @@ pub struct LanguageConfig {
#[serde(default)]
#[schemars(schema_with = "bracket_pair_config_json_schema")]
pub brackets: BracketPairConfig,
/// If set to true, indicates the language uses significant whitespace/indentation
/// for syntax structure (like Python) rather than brackets/braces for code blocks.
#[serde(default)]
pub significant_indentation: bool,
/// If set to true, auto indentation uses last non empty line to determine
/// the indentation level for a new line.
#[serde(default = "auto_indent_using_last_non_empty_line_default")]
@ -717,6 +713,12 @@ pub struct LanguageConfig {
#[serde(default, deserialize_with = "deserialize_regex")]
#[schemars(schema_with = "regex_json_schema")]
pub decrease_indent_pattern: Option<Regex>,
/// A list of rules for decreasing indentation. Each rule pairs a regex with a set of valid
/// "block-starting" tokens. When a line matches a pattern, its indentation is aligned with
/// the most recent line that began with a corresponding token. This enables context-aware
/// outdenting, like aligning an `else` with its `if`.
#[serde(default)]
pub decrease_indent_patterns: Vec<DecreaseIndentConfig>,
/// A list of characters that trigger the automatic insertion of a closing
/// bracket when they immediately precede the point where an opening
/// bracket is inserted.
@ -776,6 +778,15 @@ pub struct LanguageConfig {
pub documentation: Option<DocumentationConfig>,
}
#[derive(Clone, Debug, Deserialize, Default, JsonSchema)]
pub struct DecreaseIndentConfig {
#[serde(default, deserialize_with = "deserialize_regex")]
#[schemars(schema_with = "regex_json_schema")]
pub pattern: Option<Regex>,
#[serde(default)]
pub valid_after: Vec<String>,
}
#[derive(Clone, Debug, Serialize, Deserialize, Default, JsonSchema)]
pub struct LanguageMatcher {
/// Given a list of `LanguageConfig`'s, the language of a file can be determined based on the path extension matching any of the `path_suffixes`.
@ -899,6 +910,7 @@ impl Default for LanguageConfig {
auto_indent_on_paste: None,
increase_indent_pattern: Default::default(),
decrease_indent_pattern: Default::default(),
decrease_indent_patterns: Default::default(),
autoclose_before: Default::default(),
line_comments: Default::default(),
block_comment: Default::default(),
@ -914,7 +926,6 @@ impl Default for LanguageConfig {
jsx_tag_auto_close: None,
completion_query_characters: Default::default(),
debuggers: Default::default(),
significant_indentation: Default::default(),
documentation: None,
}
}
@ -1092,6 +1103,7 @@ struct IndentConfig {
start_capture_ix: Option<u32>,
end_capture_ix: Option<u32>,
outdent_capture_ix: Option<u32>,
suffixed_start_captures: HashMap<u32, SharedString>,
}
pub struct OutlineConfig {
@ -1522,6 +1534,14 @@ impl Language {
("outdent", &mut outdent_capture_ix),
],
);
let mut suffixed_start_captures = HashMap::default();
for (ix, name) in query.capture_names().iter().enumerate() {
if let Some(suffix) = name.strip_prefix("start.") {
suffixed_start_captures.insert(ix as u32, suffix.to_owned().into());
}
}
if let Some(indent_capture_ix) = indent_capture_ix {
grammar.indents_config = Some(IndentConfig {
query,
@ -1529,6 +1549,7 @@ impl Language {
start_capture_ix,
end_capture_ix,
outdent_capture_ix,
suffixed_start_captures,
});
}
Ok(self)