snapper_fmt/parser/

span.rs

//! Source-byte spans for splice reflow.
//!
//! Native parsers record half-open `[start, end)` ranges into the original
//! input. Structure, blank, and code regions are those slices; only prose
//! (and a configured code-comment body) is rewritten. Output is assembled
//! by copying the gaps between rewrite ranges.

use crate::parser::Region;

/// Half-open byte range `[start, end)` into the parser input.
#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash, Default)]
pub struct ByteSpan {
    pub start: usize,
    pub end: usize,
}

impl ByteSpan {
    pub const fn new(start: usize, end: usize) -> Self {
        Self { start, end }
    }

    pub const fn is_empty(self) -> bool {
        self.start == self.end
    }

    /// Slice `input` if the range is in bounds and on a char boundary.
    pub fn slice(self, input: &str) -> Option<&str> {
        input.get(self.start..self.end)
    }
}

/// Origin recorded by a parser for one [`Region`].
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum RegionOrigin {
    /// One contiguous source range (prose, structure, blank).
    Whole(ByteSpan),
    /// Fenced/env code block with independently recorded parts.
    Code {
        header: ByteSpan,
        body: ByteSpan,
        footer: ByteSpan,
    },
}

impl RegionOrigin {
    pub fn whole(self) -> ByteSpan {
        match self {
            RegionOrigin::Whole(s) => s,
            RegionOrigin::Code { header, footer, .. } => {
                ByteSpan::new(header.start, footer.end.max(header.end))
            }
        }
    }

    pub fn code(self) -> Option<CodeSpans> {
        match self {
            RegionOrigin::Code {
                header,
                body,
                footer,
            } => Some(CodeSpans {
                header,
                body,
                footer,
            }),
            RegionOrigin::Whole(_) => None,
        }
    }
}

/// Header/body/footer spans of a [`Region::Code`].
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub struct CodeSpans {
    pub header: ByteSpan,
    pub body: ByteSpan,
    pub footer: ByteSpan,
}

/// A classified region plus the parser-recorded source origin, if any.
///
/// Pandoc reconstructs regions from an AST and leaves `origin` unset.
/// Native line parsers always set it.
#[derive(Debug, Clone, PartialEq, Eq)]
pub struct SpannedRegion {
    pub region: Region,
    pub origin: Option<RegionOrigin>,
}

impl SpannedRegion {
    pub fn unspanned(region: Region) -> Self {
        Self {
            region,
            origin: None,
        }
    }

    pub fn prose(text: String, span: ByteSpan) -> Self {
        Self {
            region: Region::Prose(text),
            origin: Some(RegionOrigin::Whole(span)),
        }
    }

    pub fn structure(input: &str, span: ByteSpan) -> Self {
        Self {
            region: Region::Structure(input[span.start..span.end].to_string()),
            origin: Some(RegionOrigin::Whole(span)),
        }
    }

    pub fn blank(input: &str, span: ByteSpan) -> Self {
        Self {
            region: Region::BlankLines(input[span.start..span.end].to_string()),
            origin: Some(RegionOrigin::Whole(span)),
        }
    }

    pub fn code(
        input: &str,
        lang: Option<String>,
        header: ByteSpan,
        body: ByteSpan,
        footer: ByteSpan,
    ) -> Self {
        Self {
            region: Region::Code {
                lang,
                header: input[header.start..header.end].to_string(),
                body: input[body.start..body.end].to_string(),
                footer: input[footer.start..footer.end].to_string(),
            },
            origin: Some(RegionOrigin::Code {
                header,
                body,
                footer,
            }),
        }
    }
}

/// One physical line of `input`, with byte offsets.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub struct Line<'a> {
    /// Byte offset of the first character of the line.
    pub start: usize,
    /// Byte offset past the terminator (`\n` or `\r\n`), or EOF.
    pub end: usize,
    /// Line text without the terminator.
    pub text: &'a str,
}

impl Line<'_> {
    pub fn span(self) -> ByteSpan {
        ByteSpan::new(self.start, self.end)
    }

    /// Range of `text` only (no terminator).
    pub fn content_span(self) -> ByteSpan {
        ByteSpan::new(self.start, self.start + self.text.len())
    }

    /// Range of the terminator, empty at EOF with no trailing newline.
    pub fn terminator_span(self) -> ByteSpan {
        ByteSpan::new(self.start + self.text.len(), self.end)
    }
}

/// Split `input` into lines the way [`str::lines`] does, but keep offsets.
///
/// Terminators are `\n` or `\r\n`. A final line without a terminator is
/// included. `""` yields no lines.
pub fn iter_lines(input: &str) -> Vec<Line<'_>> {
    let mut lines = Vec::new();
    let bytes = input.as_bytes();
    let mut start = 0;
    let mut i = 0;
    while i < bytes.len() {
        if bytes[i] == b'\n' {
            let content_end = if i > start && bytes[i - 1] == b'\r' {
                i - 1
            } else {
                i
            };
            lines.push(Line {
                start,
                end: i + 1,
                text: &input[start..content_end],
            });
            start = i + 1;
        }
        i += 1;
    }
    if start < input.len() {
        lines.push(Line {
            start,
            end: input.len(),
            text: &input[start..],
        });
    }
    lines
}

/// Append a physical line to the running prose buffer and extend its span.
///
/// `include_terminator` is true for ordinary paragraphs (the original
/// newline is inside the rewrite range) and false for list-item text
/// (the terminator is a separate Structure slice).
pub fn push_prose_line(
    prose: &mut String,
    prose_span: &mut Option<ByteSpan>,
    line: &Line<'_>,
    join_space: bool,
    include_terminator: bool,
) {
    if !prose.is_empty() && join_space {
        prose.push(' ');
    }
    prose.push_str(line.text.trim());
    let end = if include_terminator {
        line.end
    } else {
        line.start + line.text.len()
    };
    match prose_span {
        None => *prose_span = Some(ByteSpan::new(line.start, end)),
        Some(s) => s.end = end,
    }
}

/// Flush accumulated prose into the region list.
pub fn flush_prose_spanned(
    prose: &mut String,
    prose_span: &mut Option<ByteSpan>,
    regions: &mut Vec<SpannedRegion>,
) {
    if prose.is_empty() {
        return;
    }
    let span = prose_span.take().unwrap_or(ByteSpan::new(0, 0));
    regions.push(SpannedRegion::prose(std::mem::take(prose), span));
}

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

    #[test]
    fn iter_lines_matches_str_lines_and_keeps_newlines() {
        let input = "a\nb\n\nc";
        let lines = iter_lines(input);
        let plain: Vec<&str> = input.lines().collect();
        assert_eq!(lines.iter().map(|l| l.text).collect::<Vec<_>>(), plain);
        assert_eq!(lines[0].span().slice(input), Some("a\n"));
        assert_eq!(lines[1].span().slice(input), Some("b\n"));
        assert_eq!(lines[2].span().slice(input), Some("\n"));
        assert_eq!(lines[3].span().slice(input), Some("c"));
        assert!(lines[3].terminator_span().is_empty());
    }

    #[test]
    fn iter_lines_empty() {
        assert!(iter_lines("").is_empty());
    }

    #[test]
    fn iter_lines_crlf() {
        let input = "a\r\nb\r\n";
        let lines = iter_lines(input);
        assert_eq!(lines.len(), 2);
        assert_eq!(lines[0].text, "a");
        assert_eq!(lines[0].span().slice(input), Some("a\r\n"));
        assert_eq!(lines[1].text, "b");
    }
}