perfectstar2k 0.3.0

//! Standard-manuscript-format (SMF) RTF export (`^KM`).
//!
//! Renders the document to the format literary agents and editors expect for
//! fiction submissions (per William Shunn's canonical guide): 12pt serif,
//! double-spaced, 1-inch margins, first-line-indented paragraphs with no blank
//! line between them, left-aligned, chapters (`#` headings) on fresh pages,
//! `*italic*`/`**bold**` rendered as real emphasis (not literal asterisks),
//! straight quotes/dashes upgraded to typographic ones.
//!
//! RTF is plain-ASCII, brace-delimited control words, so the whole file is
//! generated by hand with no dependencies. Unlike `^KE` (`export_clean`),
//! this strips Markdown markup and produces submission-ready formatting.

use crate::buffer::Buffer;
use crate::markdown::{self, MdKind};

/// The serif body font a manuscript export uses.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum ManuscriptFont {
    TimesNewRoman,
    Courier,
}

impl ManuscriptFont {
    /// The `\fN` index in the RTF font table (see `render`'s header).
    fn font_index(self) -> u8 {
        match self {
            ManuscriptFont::TimesNewRoman => 0,
            ManuscriptFont::Courier => 1,
        }
    }
}

/// Per-character emphasis, tracked so we can emit `\b`/`\i` transitions only
/// at run boundaries.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
struct Emphasis {
    bold: bool,
    italic: bool,
    /// Inline `code`: rendered as plain text, but exempt from smart typography
    /// so literal characters (quotes, dashes) survive verbatim.
    code: bool,
}

impl Emphasis {
    const PLAIN: Emphasis = Emphasis {
        bold: false,
        italic: false,
        code: false,
    };
}

const BLANK_LINES_BEFORE_CHAPTER: usize = 9;

/// Render the whole document as SMF RTF.
pub fn render(buf: &Buffer, font: ManuscriptFont) -> String {
    let f = font.font_index();
    let mut out = String::new();
    out.push_str("{\\rtf1\\ansi\\ansicpg1252\\deff0\\deflang1033\n");
    out.push_str(
        "{\\fonttbl{\\f0\\froman\\fcharset0 Times New Roman;}\
         {\\f1\\fmodern\\fcharset0 Courier New;}}\n",
    );
    out.push_str("\\viewkind4\\uc1\n");
    out.push_str("\\margl1440\\margr1440\\margt1440\\margb1440\n");

    // Full paragraph reset, restated for every paragraph, so nothing depends
    // on inherited state across `\page` breaks.
    let body_para = format!("\\pard\\plain\\f{f}\\fs24\\ql\\sl480\\slmult1\\fi720 ");

    // Whether we've emitted any paragraph yet — the first `#` heading must not
    // be preceded by a `\page` (that would produce a blank leading page).
    let mut first = true;

    for line in 0..buf.len_lines() {
        let text = buf.line_text(line);

        // Note-to-self lines never reach a reader (matches `export_clean`).
        if text.trim_start().starts_with("..") {
            continue;
        }
        // Blank lines are paragraph separators in the source — never emit an
        // empty manuscript paragraph for them.
        if text.trim().is_empty() {
            continue;
        }

        if let Some((level, title)) = markdown::heading_level(&text) {
            if level == 1 {
                if !first {
                    out.push_str("\\page\n");
                }
                for _ in 0..BLANK_LINES_BEFORE_CHAPTER {
                    out.push_str(&format!("\\pard\\plain\\f{f}\\fs24\\sl480\\slmult1\\par\n"));
                }
                out.push_str(&format!("\\pard\\plain\\qc\\b\\f{f}\\fs24 "));
                out.push_str(&escape_rtf(&title));
                out.push_str("\\b0\\par\n");
                first = false;
                continue;
            }
            // Sub-headings (## and deeper): a bold, indented body paragraph —
            // no page break, no centering.
            out.push_str(&body_para);
            out.push_str("\\b ");
            out.push_str(&escape_rtf(&title));
            out.push_str("\\b0\\par\n");
            first = false;
            continue;
        }

        // Ordinary body paragraph.
        out.push_str(&body_para);
        out.push_str(&render_line(&text));
        out.push_str("\\par\n");
        first = false;
    }

    out.push('}');
    out
}

/// Render one body line: strip Markdown markers, apply smart typography, and
/// emit `\b`/`\i` emphasis runs with every character RTF-escaped.
fn render_line(text: &str) -> String {
    let chars: Vec<char> = text.chars().collect();

    // Tag each source char with its emphasis; Marker spans (the literal
    // `**`/`*`/backtick/`#` punctuation) are dropped entirely.
    let mut tagged: Vec<(char, Emphasis)> = Vec::with_capacity(chars.len());
    let spans = markdown::scan_line(text);
    for (i, &c) in chars.iter().enumerate() {
        let mut emph = Emphasis::PLAIN;
        let mut is_marker = false;
        for &(s, e, kind) in &spans {
            if i >= s && i < e {
                match kind {
                    MdKind::Marker => is_marker = true,
                    MdKind::Bold => emph.bold = true,
                    MdKind::Italic => emph.italic = true,
                    MdKind::Code => emph.code = true,
                    MdKind::Heading => {}
                }
            }
        }
        if !is_marker {
            tagged.push((c, emph));
        }
    }

    let typed = smart_typography(&tagged);

    // Emit, tracking the active emphasis so control words appear only at run
    // boundaries.
    let mut out = String::new();
    let mut active = Emphasis::PLAIN;
    for &(c, emph) in &typed {
        if emph.bold != active.bold {
            out.push_str(if emph.bold { "\\b " } else { "\\b0 " });
        }
        if emph.italic != active.italic {
            out.push_str(if emph.italic { "\\i " } else { "\\i0 " });
        }
        active = emph;
        out.push_str(&escape_char(c));
    }
    // Close any still-open emphasis at end of paragraph.
    if active.bold {
        out.push_str("\\b0 ");
    }
    if active.italic {
        out.push_str("\\i0 ");
    }
    out
}

/// Upgrade straight ASCII punctuation to typographic characters, carrying each
/// replacement's emphasis from the first source char it consumes. Characters
/// tagged `code` are passed through untouched.
fn smart_typography(tagged: &[(char, Emphasis)]) -> Vec<(char, Emphasis)> {
    let mut out: Vec<(char, Emphasis)> = Vec::with_capacity(tagged.len());
    let mut i = 0;
    while i < tagged.len() {
        let (c, emph) = tagged[i];

        if emph.code {
            out.push((c, emph));
            i += 1;
            continue;
        }

        match c {
            '-' if run_len(tagged, i, '-') >= 2 => {
                // `--` (or more) → em dash.
                out.push(('\u{2014}', emph));
                i += run_len(tagged, i, '-');
            }
            '.' if run_len(tagged, i, '.') >= 3 => {
                // `...` → ellipsis (exactly three; extra dots stay literal).
                out.push(('\u{2026}', emph));
                i += 3;
            }
            '"' => {
                let open = i == 0 || opens_quote(tagged[i - 1].0);
                out.push((if open { '\u{201C}' } else { '\u{201D}' }, emph));
                i += 1;
            }
            '\'' => {
                let open = i == 0 || opens_quote(tagged[i - 1].0);
                out.push((if open { '\u{2018}' } else { '\u{2019}' }, emph));
                i += 1;
            }
            _ => {
                out.push((c, emph));
                i += 1;
            }
        }
    }
    out
}

/// Whether a preceding character means the next quote opens (rather than
/// closes / is an apostrophe).
fn opens_quote(prev: char) -> bool {
    prev.is_whitespace() || matches!(prev, '(' | '[' | '{' | '\u{2014}' | '\u{2013}')
}

/// Length of the run of `target` starting at `i`.
fn run_len(tagged: &[(char, Emphasis)], i: usize, target: char) -> usize {
    let mut n = 0;
    while i + n < tagged.len() && tagged[i + n].0 == target {
        n += 1;
    }
    n
}

/// RTF-escape one already-typography-processed character.
fn escape_char(c: char) -> String {
    match c {
        '\\' => "\\\\".to_string(),
        '{' => "\\{".to_string(),
        '}' => "\\}".to_string(),
        '\t' => "\\tab ".to_string(),
        c if (c as u32) < 0x80 => c.to_string(),
        // NOTE: astral characters (>= U+10000) would require a UTF-16
        // surrogate pair of `\uN` escapes. They never occur in English prose
        // manuscripts, so we degrade to a single `?` rather than implement
        // surrogate emission.
        c if (c as u32) >= 0x10000 => "?".to_string(),
        c => {
            let cp = c as u32;
            // `\uN` takes a *signed* 16-bit integer; values above 0x7FFF are
            // written as their negative two's-complement form.
            let signed = if cp > 0x7FFF {
                cp as i32 - 0x10000
            } else {
                cp as i32
            };
            // Exactly one ASCII fallback char follows (matches `\uc1`).
            format!("\\u{signed} {}", ascii_fallback(c))
        }
    }
}

/// A single ASCII stand-in for a non-ASCII char, for readers that ignore
/// `\uN`. One character only, to keep `\uc1` skip counting correct.
fn ascii_fallback(c: char) -> char {
    match c {
        '\u{2018}' | '\u{2019}' => '\'',
        '\u{201C}' | '\u{201D}' => '"',
        '\u{2014}' | '\u{2013}' => '-',
        '\u{2026}' => '.',
        _ => '?',
    }
}

/// Escape a whole plain string (used for heading titles, which carry no
/// inline emphasis or smart-typography of their own — a title is verbatim).
fn escape_rtf(text: &str) -> String {
    let mut out = String::new();
    for c in text.chars() {
        out.push_str(&escape_char(c));
    }
    out
}

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

    fn buf(text: &str) -> Buffer {
        let mut b = Buffer::open(None).expect("empty buffer");
        b.insert(0, text);
        b
    }

    #[test]
    fn escapes_backslash_and_braces() {
        assert_eq!(escape_rtf("a\\b{c}d"), "a\\\\b\\{c\\}d");
    }

    #[test]
    fn escapes_curly_quote_with_single_fallback() {
        // U+2019 = 8217, fallback a single apostrophe.
        assert_eq!(escape_char('\u{2019}'), "\\u8217 '");
        // U+201C = 8220, fallback a single double-quote.
        assert_eq!(escape_char('\u{201C}'), "\\u8220 \"");
        // Em dash U+2014 = 8212, single hyphen fallback.
        assert_eq!(escape_char('\u{2014}'), "\\u8212 -");
    }

    #[test]
    fn astral_char_degrades_to_question_mark() {
        assert_eq!(escape_char('\u{1F600}'), "?");
    }

    #[test]
    fn plain_ascii_passes_through() {
        assert_eq!(escape_rtf("Hello, world."), "Hello, world.");
    }

    #[test]
    fn bold_and_italic_become_control_words_markers_gone() {
        let line = render_line("a **b** and *c*");
        // No literal asterisks survive.
        assert!(!line.contains('*'), "markers leaked: {line}");
        assert!(line.contains("\\b b\\b0"), "bold missing: {line}");
        assert!(line.contains("\\i c\\i0"), "italic missing: {line}");
    }

    #[test]
    fn code_span_is_plain_text_no_font_switch() {
        let line = render_line("run `x` now");
        assert!(!line.contains('`'), "backticks leaked: {line}");
        assert!(!line.contains("\\f1"), "code got a font switch: {line}");
        assert!(line.contains('x'));
    }

    #[test]
    fn smart_quotes_and_em_dash() {
        // Opening double quote, apostrophe, em dash, closing double quote.
        let line = render_line("\"It's fine--really.\"");
        assert!(line.contains("\\u8220 "), "opening quote missing: {line}");
        assert!(line.contains("\\u8217 "), "apostrophe missing: {line}");
        assert!(line.contains("\\u8212 "), "em dash missing: {line}");
        assert!(line.contains("\\u8221 "), "closing quote missing: {line}");
    }

    #[test]
    fn code_span_keeps_straight_quotes() {
        let line = render_line("`it's` literal");
        // The apostrophe inside code stays straight (no \u escape for it).
        assert!(line.contains("it's"), "code apostrophe was smartened: {line}");
    }

    #[test]
    fn level1_heading_page_break_but_not_first() {
        let doc = buf("# Chapter One\nProse.\n# Chapter Two\nMore.\n");
        let rtf = render(&doc, ManuscriptFont::TimesNewRoman);
        // Exactly one \page — before Chapter Two, not Chapter One.
        assert_eq!(rtf.matches("\\page").count(), 1);
        let ch1 = rtf.find("Chapter One").unwrap();
        let page = rtf.find("\\page").unwrap();
        let ch2 = rtf.find("Chapter Two").unwrap();
        assert!(ch1 < page && page < ch2);
    }

    #[test]
    fn sub_heading_is_bold_paragraph_no_page_break() {
        let doc = buf("# Chapter\nBody.\n## A Scene\nMore body.\n");
        let rtf = render(&doc, ManuscriptFont::TimesNewRoman);
        // Only the chapter causes a page break; the scene does not add one.
        assert_eq!(rtf.matches("\\page").count(), 0); // first heading, no page
        assert!(rtf.contains("\\b A Scene\\b0"), "scene not bold: {rtf}");
    }

    #[test]
    fn note_lines_contribute_nothing() {
        let doc = buf(".. fix this later\nReal prose.\n");
        let rtf = render(&doc, ManuscriptFont::TimesNewRoman);
        assert!(!rtf.contains("fix this later"));
        assert!(rtf.contains("Real prose."));
    }

    #[test]
    fn document_is_brace_balanced() {
        let doc = buf("# Title\nText with a \\ and { and } in it.\n");
        let rtf = render(&doc, ManuscriptFont::TimesNewRoman);
        assert!(rtf.starts_with("{\\rtf1"));
        assert!(rtf.ends_with('}'));
        // User braces are escaped as \{ \} — confirm that happened, then count
        // only *structural* braces (drop the escaped literals first).
        assert!(rtf.contains("\\{ and \\}"), "user braces not escaped: {rtf}");
        let structural = rtf.replace("\\{", "").replace("\\}", "");
        let opens = structural.matches('{').count();
        let closes = structural.matches('}').count();
        assert_eq!(opens, closes, "unbalanced structural braces");
    }

    #[test]
    fn courier_selects_font_one() {
        let doc = buf("Plain line.\n");
        let rtf = render(&doc, ManuscriptFont::Courier);
        assert!(rtf.contains("\\f1\\fs24"));
    }
}