panache_parser/parser/utils/

attributes.rs

//! Parsing for Pandoc-style attributes: {#id .class key=value}
//!
//! Attributes can appear after headings, fenced code blocks, fenced divs, etc.
//! Syntax: {#identifier .class1 .class2 key1=val1 key2="val2"}
//!
//! Rules:
//! - Surrounded by { }
//! - Identifier: #id (optional, only first one counts)
//! - Classes: .class (can have multiple)
//! - Key-value pairs: key=value or key="value" or key='value' (can have multiple)
//! - Whitespace flexible between items

use crate::syntax::SyntaxKind;
use rowan::GreenNodeBuilder;

#[derive(Debug, PartialEq)]
pub struct AttributeBlock {
    pub identifier: Option<String>,
    pub classes: Vec<String>,
    pub key_values: Vec<(String, String)>,
}

/// Try to parse an attribute block from the end of a string
/// Returns: (attribute_block, text_before_attributes)
pub fn try_parse_trailing_attributes(text: &str) -> Option<(AttributeBlock, &str)> {
    let (attrs, before, _) = try_parse_trailing_attributes_with_pos(text)?;
    Some((attrs, before))
}

/// Try to parse an attribute block from the end of a string.
/// Returns: (attribute_block, text_before_attributes, open_brace_position_in_trimmed_text)
pub fn try_parse_trailing_attributes_with_pos(text: &str) -> Option<(AttributeBlock, &str, usize)> {
    let trimmed = text.trim_end();

    // Must end with }
    if !trimmed.ends_with('}') {
        return None;
    }

    // Find matching opening brace for the trailing attribute block, accounting
    // for braces inside quoted attribute values.
    let open_brace = find_matching_open_brace_for_trailing_block(trimmed)?;

    // Check if this is a bracketed span like [text]{.class} rather than a heading attribute
    // If the { is immediately after ] (with optional whitespace), this should be parsed as a span
    let before_brace = &trimmed[..open_brace];
    if before_brace.trim_end().ends_with(']') {
        log::trace!("Skipping attribute parsing for bracketed span: {}", text);
        return None;
    }

    // Parse the content between { and }
    let attr_content = &trimmed[open_brace + 1..trimmed.len() - 1];
    let attr_block = parse_attribute_content(attr_content)?;

    // Get text before attributes (trim trailing whitespace)
    let before_attrs = trimmed[..open_brace].trim_end();

    Some((attr_block, before_attrs, open_brace))
}

fn find_matching_open_brace_for_trailing_block(text: &str) -> Option<usize> {
    if !text.ends_with('}') {
        return None;
    }

    let mut stack: Vec<usize> = Vec::new();
    let mut in_quote: Option<char> = None;
    let mut escaped = false;
    let mut end_brace_open = None;

    for (idx, ch) in text.char_indices() {
        if let Some(q) = in_quote {
            if escaped {
                escaped = false;
                continue;
            }
            if ch == '\\' {
                escaped = true;
                continue;
            }
            if ch == q {
                in_quote = None;
            }
            continue;
        }

        match ch {
            '\'' | '"' => in_quote = Some(ch),
            '{' => stack.push(idx),
            '}' => {
                let open = stack.pop()?;
                if idx == text.len() - 1 {
                    end_brace_open = Some(open);
                }
            }
            _ => {}
        }
    }

    if in_quote.is_some() || !stack.is_empty() {
        return None;
    }

    end_brace_open
}

/// Parse the content inside the attribute braces
pub fn parse_attribute_content(content: &str) -> Option<AttributeBlock> {
    let mut identifier = None;
    let mut classes = Vec::new();
    let mut key_values = Vec::new();

    let content = content.trim();
    if content.is_empty() {
        return None; // Empty {} is not valid
    }

    let mut pos = 0;
    let bytes = content.as_bytes();

    while pos < bytes.len() {
        // Skip whitespace
        while pos < bytes.len() && bytes[pos].is_ascii_whitespace() {
            pos += 1;
        }

        if pos >= bytes.len() {
            break;
        }

        // Check what kind of attribute this is
        if bytes[pos] == b'=' {
            // Special case: {=format} for raw attributes
            // This is treated as a class ".=format" for compatibility
            pos += 1; // Skip =
            let start = pos;
            while pos < bytes.len() && !bytes[pos].is_ascii_whitespace() && bytes[pos] != b'}' {
                pos += 1;
            }
            if pos > start {
                // Store as "=format" class (with the = prefix)
                classes.push(format!("={}", &content[start..pos]));
            }
        } else if bytes[pos] == b'#' {
            // Identifier (only take first one)
            if identifier.is_none() {
                pos += 1; // Skip #
                let start = pos;
                while pos < bytes.len() && !bytes[pos].is_ascii_whitespace() && bytes[pos] != b'}' {
                    pos += 1;
                }
                if pos > start {
                    identifier = Some(content[start..pos].to_string());
                }
            } else {
                // Skip duplicate identifiers
                pos += 1;
                while pos < bytes.len() && !bytes[pos].is_ascii_whitespace() && bytes[pos] != b'}' {
                    pos += 1;
                }
            }
        } else if bytes[pos] == b'.' {
            // Class
            pos += 1; // Skip .
            let start = pos;
            while pos < bytes.len() && !bytes[pos].is_ascii_whitespace() && bytes[pos] != b'}' {
                pos += 1;
            }
            if pos > start {
                classes.push(content[start..pos].to_string());
            }
        } else {
            // Key-value pair
            let key_start = pos;
            while pos < bytes.len() && bytes[pos] != b'=' && !bytes[pos].is_ascii_whitespace() {
                pos += 1;
            }

            if pos >= bytes.len() || bytes[pos] != b'=' {
                // Not a valid key=value, skip this token
                while pos < bytes.len() && !bytes[pos].is_ascii_whitespace() {
                    pos += 1;
                }
                continue;
            }

            let key = content[key_start..pos].to_string();
            pos += 1; // Skip =

            // Parse value (may be quoted)
            let value = if pos < bytes.len() && (bytes[pos] == b'"' || bytes[pos] == b'\'') {
                let quote = bytes[pos];
                pos += 1; // Skip opening quote
                let val_start = pos;
                while pos < bytes.len() && bytes[pos] != quote {
                    pos += 1;
                }
                let val = content[val_start..pos].to_string();
                if pos < bytes.len() {
                    pos += 1; // Skip closing quote
                }
                val
            } else {
                // Unquoted value
                let val_start = pos;
                while pos < bytes.len() && !bytes[pos].is_ascii_whitespace() && bytes[pos] != b'}' {
                    pos += 1;
                }
                content[val_start..pos].to_string()
            };

            if !key.is_empty() {
                key_values.push((key, value));
            }
        }
    }

    // At least one attribute must be present
    if identifier.is_none() && classes.is_empty() && key_values.is_empty() {
        return None;
    }

    Some(AttributeBlock {
        identifier,
        classes,
        key_values,
    })
}

/// Parse HTML-style attributes from a raw HTML opening tag text such as
/// `<div id="x" class="a b" data-key="v">`, returning the same
/// `AttributeBlock` shape as Pandoc-style brace attributes. Whitespace-
/// separated `class="..."` is split into individual classes; `id="..."`
/// becomes the identifier; everything else becomes a key/value pair.
/// Returns `None` if the tag has no recognized attributes.
///
/// Self-closing slashes (`<div .../>`) and trailing whitespace are tolerated.
/// The leading `<TAG` and trailing `>` are stripped; this routine does not
/// validate the tag name.
pub fn parse_html_tag_attributes(tag_text: &str) -> Option<AttributeBlock> {
    let trimmed = tag_text.trim_start();
    let after_lt = trimmed.strip_prefix('<')?;
    // Find the end of the opening tag at the first `>` not inside a quoted
    // attribute value. Anything after that `>` (e.g. inline content + close
    // tag for a same-line `<div id="x">Content</div>`) is irrelevant.
    let bytes = after_lt.as_bytes();
    let mut tag_end = None;
    let mut quote: Option<u8> = None;
    for (i, &b) in bytes.iter().enumerate() {
        match (quote, b) {
            (None, b'"') | (None, b'\'') => quote = Some(b),
            (Some(q), b2) if b2 == q => quote = None,
            (None, b'>') => {
                tag_end = Some(i);
                break;
            }
            _ => {}
        }
    }
    let tag_end = tag_end?;
    let inner = &after_lt[..tag_end];
    // Drop any trailing self-closing slash.
    let inner = inner.trim_end().trim_end_matches('/').trim_end();
    // Drop the tag name (alphanumeric run after `<`).
    let bytes = inner.as_bytes();
    let mut name_end = 0usize;
    while name_end < bytes.len()
        && !bytes[name_end].is_ascii_whitespace()
        && bytes[name_end] != b'/'
    {
        name_end += 1;
    }
    let attrs_text = &inner[name_end..];
    parse_html_attribute_list(attrs_text)
}

/// Parse a raw HTML attribute list (the bytes between a tag name and the
/// closing `>`, exclusive). Accepts inputs like `id="x" class="a b"
/// data-key=v` and produces an [`AttributeBlock`]. Returns `None` if no
/// recognized attributes are present.
///
/// Used by [`parse_html_tag_attributes`] (which strips `<TAG ...>`
/// surrounding chrome before delegating here) and by
/// `AttributeNode::id` for the structural `HTML_ATTRS` CST node, whose
/// text holds JUST the attribute region.
pub fn parse_html_attribute_list(attrs_text: &str) -> Option<AttributeBlock> {
    let mut identifier: Option<String> = None;
    let mut classes: Vec<String> = Vec::new();
    let mut key_values: Vec<(String, String)> = Vec::new();

    let bytes = attrs_text.as_bytes();
    let mut i = 0usize;
    while i < bytes.len() {
        match bytes[i] {
            b' ' | b'\t' | b'\n' | b'\r' | b'/' => {
                i += 1;
            }
            _ => {
                let key_start = i;
                while i < bytes.len()
                    && !matches!(bytes[i], b' ' | b'\t' | b'\n' | b'\r' | b'=' | b'/')
                {
                    i += 1;
                }
                let key = &attrs_text[key_start..i];
                let value = if i < bytes.len() && bytes[i] == b'=' {
                    i += 1;
                    if i < bytes.len() && (bytes[i] == b'"' || bytes[i] == b'\'') {
                        let quote = bytes[i];
                        i += 1;
                        let v_start = i;
                        while i < bytes.len() && bytes[i] != quote {
                            i += 1;
                        }
                        let v = attrs_text[v_start..i].to_string();
                        if i < bytes.len() {
                            i += 1;
                        }
                        v
                    } else {
                        let v_start = i;
                        while i < bytes.len()
                            && !matches!(bytes[i], b' ' | b'\t' | b'\n' | b'\r' | b'/')
                        {
                            i += 1;
                        }
                        attrs_text[v_start..i].to_string()
                    }
                } else {
                    String::new()
                };
                if key.is_empty() {
                    continue;
                }
                match key {
                    "id" => {
                        if identifier.is_none() && !value.is_empty() {
                            identifier = Some(value);
                        }
                    }
                    "class" => {
                        for c in value.split_ascii_whitespace() {
                            classes.push(c.to_string());
                        }
                    }
                    _ => key_values.push((key.to_string(), value)),
                }
            }
        }
    }

    if identifier.is_none() && classes.is_empty() && key_values.is_empty() {
        return None;
    }
    Some(AttributeBlock {
        identifier,
        classes,
        key_values,
    })
}

/// Emit attribute block as AST nodes
pub fn emit_attributes(builder: &mut GreenNodeBuilder, attrs: &AttributeBlock) {
    builder.start_node(SyntaxKind::ATTRIBUTE.into());

    // Build the attribute string to emit
    let mut attr_str = String::from("{");

    if let Some(ref id) = attrs.identifier {
        attr_str.push('#');
        attr_str.push_str(id);
    }

    for class in &attrs.classes {
        if attr_str.len() > 1 {
            attr_str.push(' ');
        }
        // Special case: if class starts with =, it's a raw format specifier
        // Emit as {=format} not {.=format}
        if class.starts_with('=') {
            attr_str.push_str(class);
        } else {
            attr_str.push('.');
            attr_str.push_str(class);
        }
    }

    for (key, value) in &attrs.key_values {
        if attr_str.len() > 1 {
            attr_str.push(' ');
        }
        attr_str.push_str(key);
        attr_str.push('=');

        // Always quote attribute values to match Pandoc's behavior
        attr_str.push('"');
        attr_str.push_str(&value.replace('"', "\\\""));
        attr_str.push('"');
    }

    attr_str.push('}');

    builder.token(SyntaxKind::ATTRIBUTE.into(), &attr_str);
    builder.finish_node();
}

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

    #[test]
    fn test_simple_id() {
        let result = try_parse_trailing_attributes("Heading {#my-id}");
        assert!(result.is_some());
        let (attrs, before) = result.unwrap();
        assert_eq!(before, "Heading");
        assert_eq!(attrs.identifier, Some("my-id".to_string()));
        assert!(attrs.classes.is_empty());
        assert!(attrs.key_values.is_empty());
    }

    #[test]
    fn test_single_class() {
        let result = try_parse_trailing_attributes("Text {.myclass}");
        assert!(result.is_some());
        let (attrs, _) = result.unwrap();
        assert_eq!(attrs.classes, vec!["myclass"]);
    }

    #[test]
    fn test_multiple_classes() {
        let result = try_parse_trailing_attributes("Text {.class1 .class2 .class3}");
        assert!(result.is_some());
        let (attrs, _) = result.unwrap();
        assert_eq!(attrs.classes, vec!["class1", "class2", "class3"]);
    }

    #[test]
    fn test_key_value_unquoted() {
        let result = try_parse_trailing_attributes("Text {key=value}");
        assert!(result.is_some());
        let (attrs, _) = result.unwrap();
        assert_eq!(
            attrs.key_values,
            vec![("key".to_string(), "value".to_string())]
        );
    }

    #[test]
    fn test_key_value_quoted() {
        let result = try_parse_trailing_attributes("Text {key=\"value with spaces\"}");
        assert!(result.is_some());
        let (attrs, _) = result.unwrap();
        assert_eq!(
            attrs.key_values,
            vec![("key".to_string(), "value with spaces".to_string())]
        );
    }

    #[test]
    fn test_full_attributes() {
        let result =
            try_parse_trailing_attributes("Heading {#id .class1 .class2 key1=val1 key2=\"val 2\"}");
        assert!(result.is_some());
        let (attrs, before) = result.unwrap();
        assert_eq!(before, "Heading");
        assert_eq!(attrs.identifier, Some("id".to_string()));
        assert_eq!(attrs.classes, vec!["class1", "class2"]);
        assert_eq!(attrs.key_values.len(), 2);
        assert_eq!(
            attrs.key_values[0],
            ("key1".to_string(), "val1".to_string())
        );
        assert_eq!(
            attrs.key_values[1],
            ("key2".to_string(), "val 2".to_string())
        );
    }

    #[test]
    fn test_trailing_attributes_with_shortcode_in_quoted_value() {
        let text = "Slide Title {background-image='{{< placeholder 100 100 >}}' background-size=\"100px\"}";
        let result = try_parse_trailing_attributes(text);
        assert!(result.is_some());
        let (attrs, before) = result.unwrap();
        assert_eq!(before, "Slide Title");
        assert_eq!(attrs.key_values.len(), 2);
        assert_eq!(
            attrs.key_values[0],
            (
                "background-image".to_string(),
                "{{< placeholder 100 100 >}}".to_string()
            )
        );
        assert_eq!(
            attrs.key_values[1],
            ("background-size".to_string(), "100px".to_string())
        );
    }

    #[test]
    fn test_no_attributes() {
        let result = try_parse_trailing_attributes("Heading with no attributes");
        assert!(result.is_none());
    }

    #[test]
    fn test_empty_braces() {
        let result = try_parse_trailing_attributes("Heading {}");
        assert!(result.is_none());
    }

    #[test]
    fn test_only_first_id_counts() {
        let result = try_parse_trailing_attributes("Text {#id1 #id2}");
        assert!(result.is_some());
        let (attrs, _) = result.unwrap();
        assert_eq!(attrs.identifier, Some("id1".to_string()));
    }

    #[test]
    fn test_whitespace_handling() {
        let result = try_parse_trailing_attributes("Text {  #id   .class   key=val  }");
        assert!(result.is_some());
        let (attrs, _) = result.unwrap();
        assert_eq!(attrs.identifier, Some("id".to_string()));
        assert_eq!(attrs.classes, vec!["class"]);
        assert_eq!(
            attrs.key_values,
            vec![("key".to_string(), "val".to_string())]
        );
    }

    #[test]
    fn test_parse_html_tag_attributes_id_only() {
        let attrs = parse_html_tag_attributes(r#"<div id="anchor-c">"#).unwrap();
        assert_eq!(attrs.identifier.as_deref(), Some("anchor-c"));
        assert!(attrs.classes.is_empty());
        assert!(attrs.key_values.is_empty());
    }

    #[test]
    fn test_parse_html_tag_attributes_inline_content_after_open() {
        // For a same-line block `<div id="x">Content</div>`, the entire
        // line is in the HTML_BLOCK_TAG. The parser must terminate at the
        // first unquoted `>` and ignore the trailing content + close tag.
        let attrs = parse_html_tag_attributes(r#"<div id="anchor-c">Content.</div>"#).unwrap();
        assert_eq!(attrs.identifier.as_deref(), Some("anchor-c"));
    }

    #[test]
    fn test_parse_html_tag_attributes_class_and_kv() {
        let attrs = parse_html_tag_attributes(r#"<div id="x" class="a b" data-key="v">"#).unwrap();
        assert_eq!(attrs.identifier.as_deref(), Some("x"));
        assert_eq!(attrs.classes, vec!["a", "b"]);
        assert_eq!(
            attrs.key_values,
            vec![("data-key".to_string(), "v".to_string())]
        );
    }

    #[test]
    fn test_parse_html_tag_attributes_no_attrs() {
        assert!(parse_html_tag_attributes("<div>").is_none());
    }

    #[test]
    fn test_trailing_whitespace_before_attrs() {
        let result = try_parse_trailing_attributes("Heading   {#id}");
        assert!(result.is_some());
        let (_, before) = result.unwrap();
        assert_eq!(before, "Heading");
    }
}