DB & Dev Tools · Data formats: YAML lists at their key's indent, XML entities

YAML reads a list written at its key's own indent, as manifests do, and
a list at a map's indent ends at the next key. A `#` inside a quoted
scalar is no comment. XML reads entity references in text and
attributes, failing on one it does not know, and a text's path counts
text nodes from one. A tree's note takes at most half its row, cut with
an ellipsis, so a long value or attribute leaves the key its name.
Format and Minify fail loud if pressed while the JSON does not read.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

TASKS.md

@@ -387,3 +387,4 @@ Per-component lines live in `tasks/`. Tags there:
 | T33a | 1 | FAIL | the plan, the Redis browser and the structure editor drew past a 280px box; the form's rules and a row button's press went untested |
 | T33a | 2 | FAIL | the structure editor's cells kept the narrow box's first measure, and its Tab stops past the sideways fold stayed out of view |
 | T33a | 3 | PASS | |
+| T33b | 1 | FAIL | YAML lists at their key's indent and quoted `#` misread; a long note hid the key at 280px; XML text paths counted from 0 and entities stayed raw; odd-key paths and Format at rest untested |

src/devtools/datum.rs

@@ -69,16 +69,41 @@ fn scalar(text: &str) -> Datum {
     }
 }
 
+/// Where a line's comment starts: a `#` at its start or after a space, outside a quoted scalar.
+fn comment(line: &str) -> Option<usize> {
+    let mut chars = line.char_indices().peekable();
+    let (mut quote, mut last) = (None, ' ');
+    while let Some((at, ch)) = chars.next() {
+        match quote {
+            Some('"') if ch == '\\' => {
+                chars.next();
+            }
+            Some('\'') if ch == '\'' && chars.peek().is_some_and(|(_, next)| *next == '\'') => {
+                chars.next();
+            }
+            Some(open) if ch == open => quote = None,
+            Some(_) => {}
+            None if ch == '#' && last.is_whitespace() => return Some(at),
+            None if matches!(ch, '"' | '\'') && last.is_whitespace() => quote = Some(ch),
+            None => {}
+        }
+        last = ch;
+    }
+    None
+}
+
+/// Whether a line starts a list item.
+fn dashed(words: &str) -> bool {
+    let words = words.trim_start();
+    words.starts_with("- ") || words == "-"
+}
+
 /// Lines of a YAML block: each one's number, indent and words, without comments and blank lines.
 fn lines(text: &str) -> Vec<(usize, usize, &str)> {
     text.lines()
         .enumerate()
         .filter_map(|(ix, line)| {
-            let words = match line.find(" #") {
+            let words = &line[..comment(line).unwrap_or(line.len())];
-                Some(at) => &line[..at],
-                None if line.trim_start().starts_with('#') => "",
-                None => line,
-            };
             let indent = words.len() - words.trim_start().len();
             (!words.trim().is_empty() && words.trim() != "---")
                 .then(|| (ix + 1, indent, words.trim_end()))
@@ -88,18 +113,11 @@ fn lines(text: &str) -> Vec<(usize, usize, &str)> {
 
 /// The block that starts at `at` and holds the lines at `indent`.
 fn block(lines: &[(usize, usize, &str)], at: &mut usize, indent: usize) -> Result<Datum, Unread> {
-    let list = lines[*at].2.trim_start().starts_with("- ") || lines[*at].2.trim() == "-";
+    if dashed(lines[*at].2) {
-    if list {
         let mut items = Vec::new();
-        while *at < lines.len() && lines[*at].1 == indent {
+        while *at < lines.len() && lines[*at].1 == indent && dashed(lines[*at].2) {
             let (number, _, words) = lines[*at];
-            let Some(rest) = words.trim_start().strip_prefix('-') else {
+            let rest = words.trim_start()[1..].trim_start();
-                return Err(Unread {
-                    line: number,
-                    why: "a list holds only items starting with -".into(),
-                });
-            };
-            let rest = rest.trim_start();
             *at += 1;
             items.push(if rest.is_empty() {
                 nested(lines, at, indent)?
@@ -146,10 +164,10 @@ fn entries(
             },
         };
         let key = SharedString::from(key.trim().trim_matches('"').to_string());
-        let value = if value.trim().is_empty() {
+        let value = match lines.get(*at) {
-            nested(lines, at, indent)?
+            _ if !value.trim().is_empty() => scalar(value),
-        } else {
+            Some((_, same, next)) if *same == indent && dashed(next) => block(lines, at, indent)?,
-            scalar(value)
+            _ => nested(lines, at, indent)?,
         };
         map.push((key, value));
         Ok(())
@@ -244,6 +262,44 @@ mod tests {
         );
     }
 
+    #[test]
+    fn yaml_reads_a_list_at_its_key_indent() {
+        let steps = |run: &str| Datum::Map(vec![("run".into(), text(run))]);
+        assert_eq!(
+            read_yaml("steps:\n- run: build\n- run: test\nname: ci\n").expect("yaml"),
+            Datum::Map(vec![
+                (
+                    "steps".into(),
+                    Datum::List(vec![steps("build"), steps("test")])
+                ),
+                ("name".into(), text("ci"))
+            ])
+        );
+        assert_eq!(
+            read_yaml("ports:\n- 80\n- 443\n").expect("yaml"),
+            Datum::Map(vec![(
+                "ports".into(),
+                Datum::List(vec![
+                    Datum::Number("80".into()),
+                    Datum::Number("443".into())
+                ])
+            )])
+        );
+    }
+
+    #[test]
+    fn a_hash_inside_quotes_is_no_comment() {
+        assert_eq!(
+            read_yaml("title: \"Issue #12\" # a note\nmessage: 'Fix #42'\nnote: it's #1\n")
+                .expect("yaml"),
+            Datum::Map(vec![
+                ("title".into(), text("Issue #12")),
+                ("message".into(), text("Fix #42")),
+                ("note".into(), text("it's"))
+            ])
+        );
+    }
+
     #[test]
     fn yaml_says_which_line_it_could_not_read() {
         assert_eq!(read_yaml("a: 1\njust words\n").expect_err("no key").line, 2);

src/devtools/document.rs

@@ -218,6 +218,10 @@ mod tests {
         let mut all = Vec::new();
         let every = nodes(&datum, "$", "", &mut all);
         assert_eq!(every.len(), 3);
+        assert_eq!(
+            every[2].key, "$[\"odd key\"]",
+            "a key beyond letters takes brackets"
+        );
         assert!(all.is_empty(), "without a query nothing is opened");
     }
 }

src/devtools/json_editor.rs

@@ -80,9 +80,8 @@ impl RenderOnce for JsonEditor {
                 .disabled(read.is_err())
                 .on_click(move |_, _, cx| {
                     let text = editor.read(cx).text().to_string();
-                    let Some(out) = reformat(&text, pretty) else {
+                    let out = reformat(&text, pretty)
-                        return;
+                        .expect("Format and Minify rest until the JSON reads");
-                    };
                     log::info!("json editor: {label}");
                     editor.update(cx, |editor, cx| editor.set_text(out, cx));
                 })

src/devtools/tests/formats.rs

@@ -1,4 +1,7 @@
-use gpui::{AnyElement, App, AppContext as _, Entity, IntoElement, TestAppContext, Window, div};
+use gpui::{
+    AnyElement, App, AppContext as _, Entity, Focusable as _, IntoElement, TestAppContext, Window,
+    div,
+};
 
 use super::{Bench, bench, settle, tab, tap};
 use crate::{
@@ -61,3 +64,20 @@ fn format_indents_the_json_and_minify_folds_it(cx: &mut TestAppContext) {
         r#"{"a":[1,2]}"#
     );
 }
+
+/// Stops: the editor alone, the buttons resting while the JSON does not read.
+#[gpui::test]
+fn format_and_minify_rest_while_the_json_does_not_read(cx: &mut TestAppContext) {
+    let (host, cx) = bench(editor, cx);
+    let editor = cx.update(|window, cx| cx.new(|cx| CodeEditor::new("{", window, cx)));
+    host.update(cx, |bench, cx| {
+        bench.editor = Some(editor.clone());
+        cx.notify();
+    });
+    settle(cx);
+    tab(1, cx);
+    assert!(
+        cx.update(|window, cx| editor.focus_handle(cx).is_focused(window)),
+        "Tab passes Format and Minify"
+    );
+}

src/devtools/xml.rs

@@ -32,15 +32,51 @@ struct Reader<'a> {
 }
 
 impl<'a> Reader<'a> {
-    fn line(&self) -> usize {
+    /// Fails on the line `ahead` bytes past where the reader stands.
-        self.text[..self.at].matches('\n').count() + 1
+    fn fail_at<T>(&self, ahead: usize, why: impl Into<String>) -> Result<T, Unread> {
+        Err(Unread {
+            line: self.text[..self.at + ahead].matches('\n').count() + 1,
+            why: why.into(),
+        })
     }
 
     fn fail<T>(&self, why: impl Into<String>) -> Result<T, Unread> {
-        Err(Unread {
+        self.fail_at(0, why)
-            line: self.line(),
+    }
-            why: why.into(),
+
-        })
+    /// The `len` bytes where the reader stands, their entity references read.
+    fn decoded(&self, len: usize) -> Result<String, Unread> {
+        let raw = &self.rest()[..len];
+        let mut out = String::with_capacity(len);
+        let mut done = 0;
+        while let Some(amp) = raw[done..].find('&').map(|at| done + at) {
+            out.push_str(&raw[done..amp]);
+            let Some(end) = raw[amp..].find(';').map(|at| amp + at) else {
+                return self.fail_at(amp, "& starts no entity");
+            };
+            let name = &raw[amp + 1..end];
+            if name.is_empty() || name.contains(|c: char| c.is_whitespace() || c == '&') {
+                return self.fail_at(amp, "& starts no entity");
+            }
+            let number = match name.strip_prefix("#x") {
+                Some(hex) => u32::from_str_radix(hex, 16).ok(),
+                None => name.strip_prefix('#').and_then(|dec| dec.parse().ok()),
+            };
+            out.push(match name {
+                "amp" => '&',
+                "lt" => '<',
+                "gt" => '>',
+                "quot" => '"',
+                "apos" => '\'',
+                _ => match number.and_then(char::from_u32) {
+                    Some(ch) => ch,
+                    None => return self.fail_at(amp, format!("&{name}; is no entity")),
+                },
+            });
+            done = end + 1;
+        }
+        out.push_str(&raw[done..]);
+        Ok(out)
     }
 
     fn rest(&self) -> &'a str {
@@ -112,7 +148,7 @@ impl<'a> Reader<'a> {
             let Some(end) = self.rest().find(quote) else {
                 return self.fail(format!("{key} in <{name}> is never closed"));
             };
-            attributes.push((key, self.rest()[..end].to_string().into()));
+            attributes.push((key, self.decoded(end)?.into()));
             self.at += end + 1;
         }
         let children = self.content()?;
@@ -152,9 +188,9 @@ impl<'a> Reader<'a> {
                 children.push(XmlNode::Element(self.element()?));
             } else {
                 let end = rest.find('<').unwrap_or(rest.len());
-                let words = rest[..end].trim();
+                let words = self.decoded(end)?;
-                if !words.is_empty() {
+                if !words.trim().is_empty() {
-                    children.push(XmlNode::Text(words.to_string().into()));
+                    children.push(XmlNode::Text(words.trim().to_string().into()));
                 }
                 self.at += end;
             }
@@ -189,12 +225,15 @@ pub fn read_xml(text: &str) -> Result<XmlElement, Unread> {
 /// Rows for what `element` holds, each element's path listed in `open`.
 fn nodes(element: &XmlElement, path: &str, open: &mut Vec<SharedString>) -> Vec<TreeNode> {
     let mut counts: Vec<(SharedString, usize)> = Vec::new();
+    let mut texts = 0;
     element
         .children
         .iter()
-        .enumerate()
+        .map(|child| match child {
-        .map(|(ix, child)| match child {
+            XmlNode::Text(text) => {
-            XmlNode::Text(text) => TreeNode::new(format!("{path}/text()[{ix}]"), text.clone()),
+                texts += 1;
+                TreeNode::new(format!("{path}/text()[{texts}]"), text.clone())
+            }
             XmlNode::Element(child) => {
                 let seen = match counts.iter_mut().find(|(name, _)| *name == child.name) {
                     Some((_, count)) => {
@@ -289,7 +328,35 @@ impl RenderOnce for XmlViewer {
 
 #[cfg(test)]
 mod tests {
-    use super::{XmlNode, read_xml};
+    use super::{XmlNode, nodes, read_xml};
+
+    #[test]
+    fn entities_read_in_text_and_attributes() {
+        let root =
+            read_xml("<title a=\"Tom &amp; Jerry\">Tom &amp; Jerry &lt;3 &#65;&#x42;</title>")
+                .expect("xml");
+        assert_eq!(root.attributes, [("a".into(), "Tom & Jerry".into())]);
+        assert_eq!(root.children, [XmlNode::Text("Tom & Jerry <3 AB".into())]);
+        let unread = read_xml("<a>\n  &nope;</a>").expect_err("an unknown entity");
+        assert_eq!(
+            (unread.line, unread.why.as_str()),
+            (2, "&nope; is no entity")
+        );
+        assert_eq!(
+            read_xml("<a>Q&A</a>").expect_err("a bare ampersand").why,
+            "& starts no entity"
+        );
+    }
+
+    #[test]
+    fn a_text_path_counts_text_nodes_from_one() {
+        let root = read_xml("<a>x<b/>y</a>").expect("xml");
+        let keys: Vec<_> = nodes(&root, "/a", &mut Vec::new())
+            .into_iter()
+            .map(|node| node.key)
+            .collect();
+        assert_eq!(keys, ["/a/text()[1]", "/a/b[1]", "/a/text()[2]"]);
+    }
 
     #[test]
     fn elements_attributes_and_text_read_in_order() {

src/lists/tree/row.rs

@@ -3,7 +3,7 @@ use std::{ops::Range, rc::Rc};
 use gpui::{
     AnyElement, App, DragMoveEvent, ElementId, Entity, EntityId, FocusHandle, InteractiveElement,
     IntoElement, MouseButton, ParentElement, SharedString, StatefulInteractiveElement, Styled,
-    Window, div, prelude::*,
+    Window, div, prelude::*, relative,
 };
 
 use super::{
@@ -210,10 +210,12 @@ fn row(rows: &Rows, ix: usize, window: &mut Window, cx: &mut App) -> AnyElement
         .child(body)
         .children(note.map(|note| {
             div()
+                .debug_selector(|| "tree-note".into())
                 .flex_none()
+                .max_w(relative(0.5))
                 .text_size(theme.text_size(TextSize::Xs))
                 .text_color(ink.unwrap_or(colors.fg_subtle))
-                .child(note)
+                .child(Ellipsis::new(note))
         }))
         .children(landing.filter(|at| *at != DropAt::Inside).map(line))
         .on_click(move |event, window, cx| {

src/lists/tree/tests.rs

@@ -34,7 +34,8 @@ impl Render for Explorer {
                 TreeNode::new("ui", "ui").child(TreeNode::new("button", "button.rs")),
             ]),
             docs,
-            TreeNode::new("readme", "README.md"),
+            TreeNode::new("readme", "README.md")
+                .note("a note long enough to take the whole row and more"),
         ];
         let (select, check, load, rename, moved) = (
             view.clone(),
@@ -95,6 +96,20 @@ fn read<T>(view: &Entity<Explorer>, cx: &mut VisualTestContext, f: impl Fn(&Expl
     view.read_with(cx, |explorer, _| f(explorer))
 }
 
+#[gpui::test]
+fn a_long_note_leaves_the_label_half_the_row(cx: &mut TestAppContext) {
+    let (_, cx) = explorer(false, cx);
+    let note = cx.debug_bounds("tree-note").expect("the readme's note");
+    assert!(
+        note.size.width <= px(150.0),
+        "the note takes at most half the row: {note:?}"
+    );
+    assert!(
+        note.size.width > px(100.0),
+        "the note keeps its half: {note:?}"
+    );
+}
+
 #[gpui::test]
 fn keys_walk_open_climb_and_close(cx: &mut TestAppContext) {
     let (view, cx) = explorer(false, cx);