klammertext/sks/document/document_latex.cpp

#include "section_link.h"
#include "util.h"
#include "kutil.h"
#include "log.h"
#include "document_class.h"
#include "reference.h"
#include "file.h"
#include "show.h"

std::string Document_class::create_tex_output_directories()
{
    (void)K::log(3);
    std::string output_directory =
        absolute_pathname(get("K_output_dir")); // + "/" + get("K_output_basename");
    m_cache_dir = cache_directory("_latex_files");
    if (!m_toc_only) { 
        if (!file_exists(output_directory, false, true)) {
            fs::create_directory(output_directory);
        }
    }
    return output_directory;
}

// @link :section (section_link.h).  Section.tex() announces every heading
// in a comment line; the links are resolved against them, every link on a
// line.  The resolver this replaces (2026-10-12) read ONE link per line --
// std::regex_match once per line -- so a second @link ... :section @ on a
// line reached the pdf as a dead \hyperlink{__seclink__...}; it ignored
// :basename, and it wrote \hyperlink{} for an unknown title.
std::string Document_class::resolve_tex_section_links(
    const std::vector<std::pair<std::string, std::string>>& pieces)
{
    static const std::regex announce_rgx(
        R"(%__sectionlink__(.*?)__KTSL__(.*?)__KTSL__(.*)__\n?)");
    static const std::regex link_rgx(R"(__seclink__(.*?)__KTSL__(.*?)__)");
    std::vector<Section_target> table {};
    strings_t files {};
    for (const auto& [basename, text] : pieces) {
        if (!basename.empty()) files.push_back(basename);
        for (const std::string& line : find_all(text, announce_rgx, 0)) {
            std::smatch m {};
            std::regex_match(line, m, announce_rgx);
            table.push_back({normalize_section_title(m[3]), m[2], m[1]});
        }
    }
    std::string result {};
    for (const auto& [basename, text] : pieces) {
        std::string stripped = std::regex_replace(text, announce_rgx, "");
        auto begin = stripped.cbegin();
        for (std::sregex_iterator it(stripped.cbegin(), stripped.cend(), link_rgx), end; it != end; ++it) {
            const std::smatch& m = *it;
            result.append(begin, m[0].first);
            result += resolve_section_target(table, m[1], m[2], basename, files).id;
            begin = m[0].second;
        }
        result.append(begin, stripped.cend());
    }
    return result;
}

std::string Document_class::tex()
{
    (void)K::log(3);
    create_tex_output_directories();
    // :text content is already processed (klammers applied by the outer Machine).
    // Use it directly; don't re-process through a sub-Machine.
    // The pieces stay apart until the :section links are resolved, so an
    // error can name the chapter that holds the link; "" is the :text.
    std::vector<std::pair<std::string, std::string>> pieces {};
    if (!m_text.empty()) {
        pieces.push_back({"", add_latex_caption_numbers(m_text)});
    }
    if (m_files.size() > 0) {
        for (auto input_filename : m_files) {
            std::string basename = file_basename(input_filename);
            std::string cached_filename = m_cache_dir + "/" + basename;
            fs::path inpath = parse_input_filename(input_filename, m_input_dir);
            // if (!in_modification_order(input_filename, cached_filename, m_no_cache)) {
            if (m_no_cache || cache_requires_update(m_cache_dir, inpath.string(), basename)) {
                Machine M = m_machine;
                M.m_state.set("document_landscape", landscape ? "true" : "false"); // , "Landscape mode");
                // The chapter's basename: Section.tex() builds stable
                // \hypertarget names from it (a counter alone collided
                // when the cache served some chapters and not others), and
                // phases.check_section_links reads its presence as "inside
                // a @document".  Same in the html path.
                M.m_state.set("K_document_file", basename);
                M.read(inpath);
                std::string processed = trim(M.apply("tex", false));
                processed = add_latex_caption_numbers(processed);
                write_to_cache(m_cache_dir, basename, processed);
                pieces.push_back({basename, processed});
            } else {
                pieces.push_back({basename, read_from_cache(m_cache_dir, basename)});
            }
        }
    }
    std::string latex_text = resolve_tex_section_links(pieces);
    auto split_lines = regex_split(latex_text, std::regex(R"(\n)"), false);
    latex_text = resolve_caption_references("latex", split_lines, latex_line_states(split_lines));
    latex_text = string_replace(latex_text, "___AT___", "^@");
    // @typeface: collect what the document asked for, declare each as a
    // \newfontfamily switch command in the preamble, and replace every
    // marker with its command.  The declarations are appended to the
    // prolog, which latex::page emits AFTER \setmainfont -- so fontspec is
    // loaded and \newfontfamily is valid there.
    collect_typefaces(latex_text);
    latex_text = resolve_typeface_markers(latex_text);
    prolog += typeface_preamble();
    std::string structure_name =
        m_structure == docstruct_t::book    ? "book" :
        m_structure == docstruct_t::article ? "article" : "plain";
    return latex::page(structure_name,
                       m_title, m_subtitle, m_author, m_date, m_version, m_copyright,
                       m_bottom, prolog, paper_size, m_two_column, leading,
                       point_size, ragged_right,
                       cover, landscape, latex_text,
                       m_resolved_serif, m_resolved_sans, m_resolved_mono,
                       m_text_scale);
}