@perspective-dev/viewer 5.1.0 → 5.2.0

package/src/rust/agent/markdown/inline.rs

@@ -0,0 +1,497 @@
+// ┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
+// ┃ ██████ ██████ ██████       █      █      █      █      █ █▄  ▀███ █       ┃
+// ┃ ▄▄▄▄▄█ █▄▄▄▄▄ ▄▄▄▄▄█  ▀▀▀▀▀█▀▀▀▀▀ █ ▀▀▀▀▀█ ████████▌▐███ ███▄  ▀█ █ ▀▀▀▀▀ ┃
+// ┃ █▀▀▀▀▀ █▀▀▀▀▀ █▀██▀▀ ▄▄▄▄▄ █ ▄▄▄▄▄█ ▄▄▄▄▄█ ████████▌▐███ █████▄   █ ▄▄▄▄▄ ┃
+// ┃ █      ██████ █  ▀█▄       █ ██████      █      ███▌▐███ ███████▄ █       ┃
+// ┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┫
+// ┃ Copyright (c) 2017, the Perspective Authors.                              ┃
+// ┃ ╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌ ┃
+// ┃ This file is part of the Perspective library, distributed under the terms ┃
+// ┃ of the [Apache License 2.0](https://www.apache.org/licenses/LICENSE-2.0). ┃
+// ┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛
+
+//! Inline markdown: code spans, emphasis/strong/strikethrough, links,
+//! images, autolinks, entities and breaks. A hand-rolled scanner (delimiter
+//! matching needs lookbehind and run-length state that combinators fit
+//! poorly) over a simplified flanking rule: a run may open iff its next
+//! char is non-whitespace and close iff its previous char is
+//! non-whitespace, with `_` additionally barred from opening/closing
+//! against an alphanumeric neighbor so `snake_case` identifiers in prose
+//! stay literal. Unpaired delimiters degrade to their literal text.
+
+use super::entity;
+use super::events::{Event, Tag};
+
+enum Inline {
+    Text(String),
+    Code(String),
+    SoftBreak,
+    HardBreak,
+    Emphasis(Vec<Inline>),
+    Strong(Vec<Inline>),
+    Strikethrough(Vec<Inline>),
+    Link {
+        dest_url: String,
+        title: String,
+        children: Vec<Inline>,
+    },
+    Image {
+        dest_url: String,
+        children: Vec<Inline>,
+    },
+}
+
+enum ScanItem {
+    Node(Inline),
+    Delim {
+        ch: char,
+        len: usize,
+        can_open: bool,
+        can_close: bool,
+    },
+}
+
+/// Parse `text` (one paragraph/heading/cell's worth, newlines = soft break
+/// candidates) and append its events.
+pub fn parse_into(text: &str, events: &mut Vec<Event>) {
+    emit(resolve(scan(text)), events);
+}
+
+fn scan(text: &str) -> Vec<ScanItem> {
+    let chars = text.chars().collect::<Vec<_>>();
+    let mut items = vec![];
+    let mut buf = String::new();
+    let mut i = 0;
+    macro_rules! flush {
+        () => {
+            if !buf.is_empty() {
+                items.push(ScanItem::Node(Inline::Text(std::mem::take(&mut buf))));
+            }
+        };
+    }
+
+    while i < chars.len() {
+        let c = chars[i];
+        match c {
+            '\\' if i + 1 < chars.len() => {
+                let next = chars[i + 1];
+                if next == '\n' {
+                    flush!();
+                    items.push(ScanItem::Node(Inline::HardBreak));
+                } else if next.is_ascii_punctuation() {
+                    buf.push(next);
+                } else {
+                    buf.push('\\');
+                    buf.push(next);
+                }
+
+                i += 2;
+            },
+            '\n' => {
+                let hard = buf.ends_with("  ");
+                while buf.ends_with(' ') {
+                    buf.pop();
+                }
+
+                flush!();
+                items.push(ScanItem::Node(if hard {
+                    Inline::HardBreak
+                } else {
+                    Inline::SoftBreak
+                }));
+
+                i += 1;
+                while chars.get(i) == Some(&' ') {
+                    i += 1;
+                }
+            },
+            '`' => {
+                let run = run_len(&chars, i, '`');
+                if let Some(close) = find_code_close(&chars, i + run, run) {
+                    flush!();
+                    let content = chars[i + run..close]
+                        .iter()
+                        .map(|x| if *x == '\n' { ' ' } else { *x })
+                        .collect::<String>();
+
+                    items.push(ScanItem::Node(Inline::Code(trim_code_span(content))));
+                    i = close + run;
+                } else {
+                    buf.extend(std::iter::repeat_n('`', run));
+                    i += run;
+                }
+            },
+            '*' | '_' | '~' => {
+                let run = run_len(&chars, i, c);
+                let prev = if i > 0 { Some(chars[i - 1]) } else { None };
+                let next = chars.get(i + run).copied();
+                let mut can_open = next.is_some_and(|x| !x.is_whitespace());
+                let mut can_close = prev.is_some_and(|x| !x.is_whitespace());
+                if c == '_' {
+                    can_open &= !prev.is_some_and(|x| x.is_alphanumeric());
+                    can_close &= !next.is_some_and(|x| x.is_alphanumeric());
+                }
+
+                if c == '~' && run < 2 || !can_open && !can_close {
+                    buf.extend(std::iter::repeat_n(c, run));
+                } else {
+                    flush!();
+                    items.push(ScanItem::Delim {
+                        ch: c,
+                        len: run,
+                        can_open,
+                        can_close,
+                    });
+                }
+
+                i += run;
+            },
+            '[' => match parse_link(&chars, i, false) {
+                Some((node, next)) => {
+                    flush!();
+                    items.push(ScanItem::Node(node));
+                    i = next;
+                },
+                None => {
+                    buf.push('[');
+                    i += 1;
+                },
+            },
+            '!' if chars.get(i + 1) == Some(&'[') => match parse_link(&chars, i + 1, true) {
+                Some((node, next)) => {
+                    flush!();
+                    items.push(ScanItem::Node(node));
+                    i = next;
+                },
+                None => {
+                    buf.push('!');
+                    i += 1;
+                },
+            },
+            '<' => match parse_autolink(&chars, i) {
+                Some((node, next)) => {
+                    flush!();
+                    items.push(ScanItem::Node(node));
+                    i = next;
+                },
+                None => {
+                    buf.push('<');
+                    i += 1;
+                },
+            },
+            '&' => {
+                let rest = chars[i + 1..].iter().collect::<String>();
+                match entity::decode(&rest) {
+                    Some((decoded, used)) => {
+                        buf.push(decoded);
+                        i += 1 + used;
+                    },
+                    None => {
+                        buf.push('&');
+                        i += 1;
+                    },
+                }
+            },
+            c => {
+                buf.push(c);
+                i += 1;
+            },
+        }
+    }
+
+    flush!();
+    items
+}
+
+fn run_len(chars: &[char], start: usize, ch: char) -> usize {
+    chars[start..].iter().take_while(|x| **x == ch).count()
+}
+
+/// Find the next backtick run of EXACTLY `n` (longer/shorter runs are
+/// skipped whole, per CommonMark).
+fn find_code_close(chars: &[char], mut i: usize, n: usize) -> Option<usize> {
+    while i < chars.len() {
+        if chars[i] == '`' {
+            let run = run_len(chars, i, '`');
+            if run == n {
+                return Some(i);
+            }
+
+            i += run;
+        } else {
+            i += 1;
+        }
+    }
+
+    None
+}
+
+fn trim_code_span(content: String) -> String {
+    if content.starts_with(' ')
+        && content.ends_with(' ')
+        && content.len() > 1
+        && !content.chars().all(|x| x == ' ')
+    {
+        content[1..content.len() - 1].to_owned()
+    } else {
+        content
+    }
+}
+
+/// `[text](dest "title")` at `i` (pointing at the `[`). The bracket text is
+/// parsed recursively; URL trust policy stays in the renderer.
+fn parse_link(chars: &[char], i: usize, image: bool) -> Option<(Inline, usize)> {
+    let mut j = i + 1;
+    let mut depth = 1;
+    while j < chars.len() {
+        match chars[j] {
+            '\\' => j += 1,
+            '[' => depth += 1,
+            ']' => {
+                depth -= 1;
+                if depth == 0 {
+                    break;
+                }
+            },
+            _ => (),
+        }
+
+        j += 1;
+    }
+
+    if depth != 0 || chars.get(j + 1) != Some(&'(') {
+        return None;
+    }
+
+    let mut k = j + 2;
+    while chars.get(k) == Some(&' ') {
+        k += 1;
+    }
+
+    let mut dest_url = String::new();
+    let mut paren_depth = 0;
+    while k < chars.len() {
+        match chars[k] {
+            '\\' if k + 1 < chars.len() => {
+                dest_url.push(chars[k + 1]);
+                k += 1;
+            },
+            '(' => {
+                paren_depth += 1;
+                dest_url.push('(');
+            },
+            ')' if paren_depth == 0 => break,
+            ')' => {
+                paren_depth -= 1;
+                dest_url.push(')');
+            },
+            ' ' => break,
+            c => dest_url.push(c),
+        }
+
+        k += 1;
+    }
+
+    while chars.get(k) == Some(&' ') {
+        k += 1;
+    }
+
+    let mut title = String::new();
+    if let Some(quote @ ('"' | '\'')) = chars.get(k).copied() {
+        k += 1;
+        while k < chars.len() && chars[k] != quote {
+            title.push(chars[k]);
+            k += 1;
+        }
+
+        if k >= chars.len() {
+            return None;
+        }
+
+        k += 1;
+        while chars.get(k) == Some(&' ') {
+            k += 1;
+        }
+    }
+
+    if chars.get(k) != Some(&')') {
+        return None;
+    }
+
+    let inner = chars[i + 1..j].iter().collect::<String>();
+    let children = resolve(scan(&inner));
+    let node = if image {
+        Inline::Image { dest_url, children }
+    } else {
+        Inline::Link {
+            dest_url,
+            title,
+            children,
+        }
+    };
+
+    Some((node, k + 1))
+}
+
+/// `<https://…>` / `<mailto:…>` autolinks.
+fn parse_autolink(chars: &[char], i: usize) -> Option<(Inline, usize)> {
+    let end = i
+        + chars[i + 1..]
+            .iter()
+            .position(|x| *x == '>' || x.is_whitespace() || *x == '<')?
+        + 1;
+
+    if chars[end] != '>' {
+        return None;
+    }
+
+    let dest = chars[i + 1..end].iter().collect::<String>();
+    let lower = dest.to_ascii_lowercase();
+    if !(lower.starts_with("http://")
+        || lower.starts_with("https://")
+        || lower.starts_with("mailto:"))
+    {
+        return None;
+    }
+
+    Some((
+        Inline::Link {
+            dest_url: dest.clone(),
+            title: String::new(),
+            children: vec![Inline::Text(dest)],
+        },
+        end + 1,
+    ))
+}
+
+struct Frame {
+    /// `Some((char, run length))` for a pending opener; `None` for the
+    /// bottom output frame.
+    delim: Option<(char, usize)>,
+    children: Vec<Inline>,
+}
+
+/// Pair delimiter runs into emphasis nodes: strict nesting against the
+/// frame stack, with non-matching intervening openers demoted to literal
+/// text, and greedy `2 → Strong / 1 → Emphasis` consumption of partial
+/// runs (`***x***` → strong + em).
+fn resolve(items: Vec<ScanItem>) -> Vec<Inline> {
+    let mut stack = vec![Frame {
+        delim: None,
+        children: vec![],
+    }];
+
+    for item in items {
+        match item {
+            ScanItem::Node(node) => stack.last_mut().unwrap().children.push(node),
+            ScanItem::Delim {
+                ch,
+                len,
+                can_open,
+                can_close,
+            } => {
+                let mut remaining = len;
+                while remaining > 0 && can_close {
+                    let Some(fi) = stack.iter().rposition(|x| {
+                        x.delim
+                            .is_some_and(|(oc, ol)| oc == ch && (ch != '~' || ol >= 2))
+                    }) else {
+                        break;
+                    };
+
+                    if ch == '~' && remaining < 2 {
+                        break;
+                    }
+
+                    demote_above(&mut stack, fi);
+                    let frame = stack.pop().unwrap();
+                    let (_, olen) = frame.delim.unwrap();
+                    let take = if ch != '~' && (olen < 2 || remaining < 2) {
+                        1
+                    } else {
+                        2
+                    };
+
+                    let node = match (ch, take) {
+                        ('~', _) => Inline::Strikethrough(frame.children),
+                        (_, 2) => Inline::Strong(frame.children),
+                        _ => Inline::Emphasis(frame.children),
+                    };
+
+                    let leftover = olen - take;
+                    if leftover > 0 {
+                        stack.push(Frame {
+                            delim: Some((ch, leftover)),
+                            children: vec![node],
+                        });
+                    } else {
+                        stack.last_mut().unwrap().children.push(node);
+                    }
+
+                    remaining -= take;
+                }
+
+                if remaining > 0 {
+                    if can_open {
+                        stack.push(Frame {
+                            delim: Some((ch, remaining)),
+                            children: vec![],
+                        });
+                    } else {
+                        stack
+                            .last_mut()
+                            .unwrap()
+                            .children
+                            .push(Inline::Text(std::iter::repeat_n(ch, remaining).collect()));
+                    }
+                }
+            },
+        }
+    }
+
+    demote_above(&mut stack, 0);
+    stack.pop().unwrap().children
+}
+
+/// Collapse every frame above `fi` into its parent: the unclosed delimiter
+/// becomes literal text, its children splice through — content is never
+/// dropped.
+fn demote_above(stack: &mut Vec<Frame>, fi: usize) {
+    while stack.len() - 1 > fi {
+        let frame = stack.pop().unwrap();
+        let parent = stack.last_mut().unwrap();
+        if let Some((ch, len)) = frame.delim {
+            parent
+                .children
+                .push(Inline::Text(std::iter::repeat_n(ch, len).collect()));
+        }
+
+        parent.children.extend(frame.children);
+    }
+}
+
+fn emit(nodes: Vec<Inline>, events: &mut Vec<Event>) {
+    for node in nodes {
+        match node {
+            Inline::Text(text) => events.push(Event::Text(text)),
+            Inline::Code(text) => events.push(Event::Code(text)),
+            Inline::SoftBreak => events.push(Event::SoftBreak),
+            Inline::HardBreak => events.push(Event::HardBreak),
+            Inline::Emphasis(children) => wrap(Tag::Emphasis, children, events),
+            Inline::Strong(children) => wrap(Tag::Strong, children, events),
+            Inline::Strikethrough(children) => wrap(Tag::Strikethrough, children, events),
+            Inline::Link {
+                dest_url,
+                title,
+                children,
+            } => wrap(Tag::Link { dest_url, title }, children, events),
+            Inline::Image { dest_url, children } => wrap(Tag::Image { dest_url }, children, events),
+        }
+    }
+}
+
+fn wrap(tag: Tag, children: Vec<Inline>, events: &mut Vec<Event>) {
+    events.push(Event::Start(tag));
+    emit(children, events);
+    events.push(Event::End);
+}