marco_core/parser/inlines/

text_parser.rs

//! Text parser - handle plain text fallback
//!
//! Parses plain text segments when no inline elements match. Handles special
//! cases like trailing spaces before newlines and consecutive backticks.

use super::shared::{opt_span, GrammarSpan};
use crate::parser::ast::{Node, NodeKind};
use nom::bytes::complete::take;
use nom::IResult;
use nom::Input;
use nom::Parser;

/// Parse plain text up to the next special character
///
/// Consumes text until a special inline character is found (*_`[<!&\\n).
///
/// Note: This also stops at several extension delimiters so they can be parsed
/// in the middle of a line:
/// - `^` (superscript)
/// - `~` (subscript or strikethrough)
/// - `==` (mark)
/// - `--` (dash strikethrough)
/// - `˅` (arrow-style subscript)
///   Handles special cases:
/// - Trailing spaces before newlines (potential hard line break)
/// - Consecutive backticks (consume all together)
///
/// # Arguments
/// * `input` - The input text as a GrammarSpan
///
/// # Returns
/// * `Ok((remaining, node))` - Successfully parsed text node
/// * `Err(_)` - No text to parse (input starts with special character)
pub fn parse_text(input: GrammarSpan) -> IResult<GrammarSpan, Node> {
    let text_fragment = input.fragment();

    // Autolink literals, emoji shortcodes, and platform mentions are all
    // single-line constructs, so bound their lookahead scans to the current
    // line instead of the entire remaining document. Without this bound,
    // each of the three `find_next_*_start` helpers below does an unbounded
    // substring search to the end of `text_fragment` (which shrinks by only
    // one small chunk per `parse_text` call), making a large multi-line
    // paragraph with no matches anywhere cost O(n^2) instead of O(n).
    let line_end = text_fragment.find('\n').unwrap_or(text_fragment.len());
    let line = &text_fragment[..line_end];

    // GFM autolink literals can appear in the middle of a text node.
    // If we can see a valid autolink literal starting at some offset, we must
    // stop before it so the dedicated parser can run.
    let next_autolink_literal =
        super::gfm_autolink_literal_parser::find_next_autolink_literal_start(line)
            .unwrap_or(text_fragment.len());

    // Emoji shortcodes (extended syntax) can appear in the middle of a text node.
    // Only stop for *recognized* shortcodes; unknown ones remain literal.
    let next_emoji_shortcode =
        super::marco_emoji_shortcode_parser::find_next_emoji_shortcode_start(line)
            .unwrap_or(text_fragment.len());

    // Platform mentions (extended syntax) can appear in the middle of a text node.
    let next_platform_mention =
        super::marco_platform_mentions_parser::find_next_platform_mention_start(line)
            .unwrap_or(text_fragment.len());

    // Find the next special character / delimiter start.
    //
    // Important: we intentionally do NOT treat a single '-' as special because
    // it's too common in normal prose. Instead we only stop at the start of
    // a *double* dash sequence "--".
    let next_special = text_fragment
        .char_indices()
        .find_map(|(idx, ch)| match ch {
            '*' | '_' | '`' | '[' | '<' | '!' | '&' | '\n' | '\\' | '$' => Some(idx),
            '^' | '~' | '˅' => Some(idx),
            '=' => {
                if text_fragment[idx..].starts_with("==") {
                    Some(idx)
                } else {
                    None
                }
            }
            '-' => {
                if text_fragment[idx..].starts_with("--") {
                    Some(idx)
                } else {
                    None
                }
            }
            _ => None,
        })
        .unwrap_or(text_fragment.len());

    // If an autolink literal begins at offset 0, do not treat it as plain text.
    if next_autolink_literal == 0 {
        return Err(nom::Err::Error(nom::error::Error::new(
            input,
            nom::error::ErrorKind::Verify,
        )));
    }

    // If an emoji shortcode begins at offset 0, do not treat it as plain text.
    if next_emoji_shortcode == 0 {
        return Err(nom::Err::Error(nom::error::Error::new(
            input,
            nom::error::ErrorKind::Verify,
        )));
    }

    // If a platform mention begins at offset 0, do not treat it as plain text.
    if next_platform_mention == 0 {
        return Err(nom::Err::Error(nom::error::Error::new(
            input,
            nom::error::ErrorKind::Verify,
        )));
    }

    let next_special = next_special
        .min(next_autolink_literal)
        .min(next_emoji_shortcode)
        .min(next_platform_mention);

    if next_special == 0 {
        // No text - input starts with special character
        return Err(nom::Err::Error(nom::error::Error::new(
            input,
            nom::error::ErrorKind::Verify,
        )));
    }

    // Check if the upcoming character is a newline and the text ends with spaces
    // If so, don't consume trailing spaces (they might be part of a hard line break)
    let mut text_len = next_special;
    if next_special < text_fragment.len() && text_fragment[next_special..].starts_with('\n') {
        // Check for trailing spaces
        let mut trailing_spaces = 0;
        for ch in text_fragment[..next_special].chars().rev() {
            if ch == ' ' {
                trailing_spaces += 1;
            } else {
                break;
            }
        }

        // If we have 2+ trailing spaces, don't consume them
        // (they might be part of a hard line break pattern)
        if trailing_spaces >= 2 {
            text_len = next_special - trailing_spaces;
        }
    }

    if text_len == 0 {
        // Only trailing spaces - don't consume them
        return Err(nom::Err::Error(nom::error::Error::new(
            input,
            nom::error::ErrorKind::Verify,
        )));
    }

    // Use nom::Input to properly advance by byte count (not character count!)
    let text_content = input.take(text_len);
    let rest = input.take_from(text_len);

    let span = opt_span(text_content);

    let node = Node {
        kind: NodeKind::Text(text_content.fragment().to_string()),
        span,
        children: Vec::new(),
    };

    Ok((rest, node))
}

/// Parse a single special character as text (fallback for unmatched syntax)
///
/// When an inline element parser fails to match, this function consumes the
/// special character as plain text. For backticks, consumes all consecutive
/// backticks together.
///
/// # Arguments
/// * `input` - The input text as a GrammarSpan
///
/// # Returns
/// * `Ok((remaining, node))` - Successfully parsed text node with special character
/// * `Err(_)` - Input is empty
pub fn parse_special_as_text(input: GrammarSpan) -> IResult<GrammarSpan, Node> {
    let text_fragment = input.fragment();

    if text_fragment.is_empty() {
        return Err(nom::Err::Error(nom::error::Error::new(
            input,
            nom::error::ErrorKind::Eof,
        )));
    }

    // Special case: if it's a backtick, consume all consecutive backticks
    // This prevents ```foo`` from being parsed as ` + ``foo``
    let char_len = if text_fragment.starts_with('`') {
        // Count all consecutive backticks
        text_fragment.chars().take_while(|&c| c == '`').count()
    } else {
        text_fragment
            .chars()
            .next()
            .map(|c| c.len_utf8())
            .unwrap_or(1)
    };

    let (rest, text_content) = take(char_len).parse(input)?;

    let span = opt_span(text_content);

    let node = Node {
        kind: NodeKind::Text(text_content.fragment().to_string()),
        span,
        children: Vec::new(),
    };

    Ok((rest, node))
}

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

    #[test]
    fn smoke_test_parse_text_basic() {
        let input = GrammarSpan::new("Hello World*");
        let result = parse_text(input);

        assert!(result.is_ok(), "Failed to parse plain text");
        let (rest, node) = result.unwrap();

        assert_eq!(rest.fragment(), &"*");

        if let NodeKind::Text(text) = &node.kind {
            assert_eq!(text, "Hello World");
        } else {
            panic!("Expected Text node");
        }
    }

    #[test]
    fn smoke_test_parse_text_up_to_special() {
        let input = GrammarSpan::new("text with `code`");
        let result = parse_text(input);

        assert!(result.is_ok());
        let (rest, node) = result.unwrap();

        assert_eq!(rest.fragment(), &"`code`");

        if let NodeKind::Text(text) = &node.kind {
            assert_eq!(text, "text with ");
        }
    }

    #[test]
    fn smoke_test_parse_text_trailing_spaces() {
        let input = GrammarSpan::new("text  \n");
        let result = parse_text(input);

        assert!(result.is_ok());
        let (rest, node) = result.unwrap();

        // Should not consume trailing spaces before newline
        assert_eq!(rest.fragment(), &"  \n");

        if let NodeKind::Text(text) = &node.kind {
            assert_eq!(text, "text");
        }
    }

    #[test]
    fn smoke_test_parse_text_starts_with_special() {
        let input = GrammarSpan::new("*emphasis*");
        let result = parse_text(input);

        assert!(
            result.is_err(),
            "Should not parse text starting with special char"
        );
    }

    #[test]
    fn smoke_test_parse_special_as_text_asterisk() {
        let input = GrammarSpan::new("* not emphasis");
        let result = parse_special_as_text(input);

        assert!(result.is_ok(), "Failed to parse special as text");
        let (rest, node) = result.unwrap();

        assert_eq!(rest.fragment(), &" not emphasis");

        if let NodeKind::Text(text) = &node.kind {
            assert_eq!(text, "*");
        }
    }

    #[test]
    fn smoke_test_parse_special_as_text_backticks() {
        let input = GrammarSpan::new("```not code");
        let result = parse_special_as_text(input);

        assert!(result.is_ok());
        let (rest, node) = result.unwrap();

        assert_eq!(rest.fragment(), &"not code");

        if let NodeKind::Text(text) = &node.kind {
            assert_eq!(text, "```");
        }
    }

    #[test]
    fn smoke_test_parse_text_position() {
        let input = GrammarSpan::new("Hello*");
        let result = parse_text(input);

        assert!(result.is_ok());
        let (_, node) = result.unwrap();

        assert!(node.span.is_some(), "Text should have position info");

        let span = node.span.unwrap();
        assert_eq!(span.start.offset, 0);
        assert_eq!(span.end.offset, 5); // "Hello" is 5 bytes
    }
}