arity 0.15.0

A language server, formatter, and linter for R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292

use super::*;

/// Format `text` off the snapshot's cached parse when the db's tracked buffer
/// for `path` still matches it; otherwise re-parse. A write racing the read
/// trips [`salsa::Cancelled`], which also falls back to a fresh parse.
pub(crate) fn format_edits_via_db(
    snapshot: &Analysis,
    path: &Path,
    text: &str,
    style: FormatStyle,
    encoding: PositionEncoding,
) -> Option<Vec<TextEdit>> {
    let cached = salsa::Cancelled::catch(AssertUnwindSafe(|| {
        let file = snapshot.lookup_file(path)?;
        if snapshot.file_text(file) != text {
            // The tracked input lags the live buffer; the cached tree is stale.
            return None;
        }
        if !snapshot.parse_diagnostics(file).is_empty() {
            // Parse errors: the formatter refuses, like `compute_format_edits`.
            return Some(None);
        }
        let root = snapshot.parsed_tree(file);
        let formatted = format_node(&root, style, text).ok();
        Some(formatted.map(|formatted| edits_for_formatted(text, formatted, encoding)))
    }));
    match cached {
        Ok(Some(edits)) => edits,
        // Cache miss (`Ok(None)`) or a racing write (`Err`): re-parse from text.
        Ok(None) | Err(_) => compute_format_edits(text, style, encoding),
    }
}

/// Range-format `text` off the snapshot's cached parse when the db's tracked
/// buffer for `path` still matches it; otherwise re-parse. Mirrors
/// [`format_edits_via_db`]'s cache/cancellation handling.
pub(crate) fn format_range_edits_via_db(
    snapshot: &Analysis,
    path: &Path,
    text: &str,
    range: Range,
    style: FormatStyle,
    encoding: PositionEncoding,
) -> Option<Vec<TextEdit>> {
    let cached = salsa::Cancelled::catch(AssertUnwindSafe(|| {
        let file = snapshot.lookup_file(path)?;
        if snapshot.file_text(file) != text {
            // The tracked input lags the live buffer; the cached tree is stale.
            return None;
        }
        if !snapshot.parse_diagnostics(file).is_empty() {
            // Parse errors: the formatter refuses, like the whole-document path.
            return Some(None);
        }
        let root = snapshot.parsed_tree(file);
        let line_index = LineIndex::new(text);
        let text_range = lsp_range_to_text_range(&line_index, range, encoding);
        let edits = match format_range(&root, text_range, style, text) {
            Ok(Some(formatted)) => Some(range_edits(&line_index, text, formatted, encoding)),
            Ok(None) => Some(Vec::new()),
            Err(_) => None,
        };
        Some(edits)
    }));
    match cached {
        Ok(Some(edits)) => edits,
        // Cache miss (`Ok(None)`) or a racing write (`Err`): re-parse from text.
        Ok(None) | Err(_) => compute_format_range_edits(text, range, style, encoding),
    }
}

/// Compute the LSP `TextEdit`s to format `text` with `style`, re-parsing it.
///
/// Returns `None` when the formatter rejects the input (e.g. parse error).
/// An empty `Vec` means the document is already formatted.
pub fn compute_format_edits(
    text: &str,
    style: FormatStyle,
    encoding: PositionEncoding,
) -> Option<Vec<TextEdit>> {
    let formatted = format_with_style(text, style).ok()?;
    Some(edits_for_formatted(text, formatted, encoding))
}

/// Compute the LSP `TextEdit`s to format the selection `range` of `text`,
/// re-parsing it.
///
/// Returns `None` when the formatter rejects the input (e.g. parse error). An
/// empty `Vec` means the selected region is already formatted or covers no
/// statement.
pub fn compute_format_range_edits(
    text: &str,
    range: Range,
    style: FormatStyle,
    encoding: PositionEncoding,
) -> Option<Vec<TextEdit>> {
    let parsed = parse(text);
    if !parsed.diagnostics.is_empty() {
        return None;
    }
    let line_index = LineIndex::new(text);
    let text_range = lsp_range_to_text_range(&line_index, range, encoding);
    match format_range(&parsed.cst, text_range, style, text).ok()? {
        Some(formatted) => Some(range_edits(&line_index, text, formatted, encoding)),
        None => Some(Vec::new()),
    }
}

/// Convert a byte `TextRange` to an LSP `Range` via `line_index` (built over the
/// text the range indexes).
pub(crate) fn text_range_to_lsp_range(
    line_index: &LineIndex,
    range: TextRange,
    encoding: PositionEncoding,
) -> Range {
    Range {
        start: line_index.byte_to_position(u32::from(range.start()) as usize, encoding),
        end: line_index.byte_to_position(u32::from(range.end()) as usize, encoding),
    }
}

/// Convert an LSP `Range` to a byte `TextRange`. `position_to_byte` already
/// clamps to the text length; we only ensure `start <= end`.
pub(crate) fn lsp_range_to_text_range(
    line_index: &LineIndex,
    range: Range,
    encoding: PositionEncoding,
) -> TextRange {
    let start = line_index.position_to_byte(range.start, encoding);
    let end = line_index.position_to_byte(range.end, encoding);
    TextRange::new(
        TextSize::new(start as u32),
        TextSize::new(start.max(end) as u32),
    )
}

/// Turn a [`RangeFormatted`] region into the LSP edit list, dropping the edit
/// when it would not change the buffer.
pub(crate) fn range_edits(
    line_index: &LineIndex,
    text: &str,
    formatted: crate::formatter::RangeFormatted,
    encoding: PositionEncoding,
) -> Vec<TextEdit> {
    let start = usize::from(formatted.range.start());
    let end = usize::from(formatted.range.end());
    if text.get(start..end) == Some(formatted.text.as_str()) {
        return Vec::new();
    }
    vec![TextEdit {
        range: Range {
            start: line_index.byte_to_position(start, encoding),
            end: line_index.byte_to_position(end, encoding),
        },
        new_text: formatted.text,
    }]
}

/// The whole-document edit replacing `text` with its formatted form (empty when
/// already formatted). The single source of the edit geometry shared by the
/// re-parse path ([`compute_format_edits`]) and the cached-tree path.
pub(crate) fn edits_for_formatted(
    text: &str,
    formatted: String,
    encoding: PositionEncoding,
) -> Vec<TextEdit> {
    if formatted == text {
        return Vec::new();
    }
    let line_index = LineIndex::new(text);
    let end = line_index.byte_to_position(text.len(), encoding);
    vec![TextEdit {
        range: Range {
            start: Position::new(0, 0),
            end,
        },
        new_text: formatted,
    }]
}

pub(crate) fn to_lsp_diagnostic(
    d: &Diagnostic,
    idx: &LineIndex,
    encoding: PositionEncoding,
) -> LspDiagnostic {
    let start = idx.byte_to_position(u32::from(d.range.start()) as usize, encoding);
    let end = idx.byte_to_position(u32::from(d.range.end()) as usize, encoding);
    let severity = match d.severity {
        Severity::Error => DiagnosticSeverity::ERROR,
        Severity::Warning => DiagnosticSeverity::WARNING,
        Severity::Info => DiagnosticSeverity::INFORMATION,
        Severity::Hint => DiagnosticSeverity::HINT,
    };
    LspDiagnostic {
        range: Range { start, end },
        severity: Some(severity),
        code: Some(NumberOrString::String(d.rule.to_string())),
        source: Some("arity".to_string()),
        message: d.message.body.clone(),
        ..Default::default()
    }
}

/// Convert a lint's findings into LSP diagnostics against `text` (the source the
/// findings' byte ranges index). Used by the pull-diagnostic path; the push path
/// maps the same way inline in the lint thread.
pub(crate) fn findings_to_items(
    findings: &[Diagnostic],
    text: &str,
    encoding: PositionEncoding,
) -> Vec<LspDiagnostic> {
    let idx = LineIndex::new(text);
    findings
        .iter()
        .map(|d| to_lsp_diagnostic(d, &idx, encoding))
        .collect()
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn findings_to_items_maps_range_severity_and_code() {
        use crate::linter::ViolationData;
        // "line0\nWARN\n": the finding spans bytes 6..10 (the second line).
        let text = "line0\nWARN\n";
        let findings = vec![Diagnostic {
            rule: "demo-rule",
            severity: Severity::Warning,
            path: test_path().to_path_buf(),
            range: TextRange::new(TextSize::from(6), TextSize::from(10)),
            message: ViolationData::new("demo-rule", "a demo finding"),
            fix: None,
        }];
        let items = findings_to_items(&findings, text, PositionEncoding::Utf16);
        assert_eq!(items.len(), 1);
        let item = &items[0];
        assert_eq!(item.range.start, Position::new(1, 0));
        assert_eq!(item.range.end, Position::new(1, 4));
        assert_eq!(item.severity, Some(DiagnosticSeverity::WARNING));
        assert_eq!(item.source.as_deref(), Some("arity"));
        assert_eq!(item.message, "a demo finding");
        assert!(matches!(&item.code, Some(NumberOrString::String(c)) if c == "demo-rule"));
    }

    // --- db read path -----------------------------------------------------

    /// The cached-tree format path matches the re-parse path when the db's
    /// tracked buffer is the live text, and falls back (still correctly) when the
    /// db lags the buffer or has never seen the path.
    #[test]
    fn format_via_db_matches_compute_and_falls_back() {
        use crate::incremental::IncrementalDatabase;
        let style = FormatStyle::default();
        let path = test_path();
        let buffer = "x<-f(1 )\n";
        let encoding = PositionEncoding::Utf16;
        let expected = compute_format_edits(buffer, style, encoding);
        assert!(
            matches!(&expected, Some(edits) if !edits.is_empty()),
            "fixture must require reformatting"
        );

        // Cache hit: tracked text == buffer → format off the cached tree.
        let mut db = IncrementalDatabase::default();
        db.upsert_file(path, buffer.to_string());
        let snapshot = db.snapshot();
        assert_eq!(
            format_edits_via_db(&snapshot, path, buffer, style, encoding),
            expected,
            "cached-tree format must match the re-parse path"
        );

        // Stale db (tracked text lags the buffer) → fall back to a fresh parse.
        let mut stale = IncrementalDatabase::default();
        stale.upsert_file(path, "y <- 1\n".to_string());
        assert_eq!(
            format_edits_via_db(&stale.snapshot(), path, buffer, style, encoding),
            expected,
            "version skew must fall back to the buffer text"
        );

        // Untracked path → fall back as well.
        let empty = IncrementalDatabase::default();
        assert_eq!(
            format_edits_via_db(&empty.snapshot(), path, buffer, style, encoding),
            expected,
            "untracked path must fall back to the buffer text"
        );
    }
}