mdprint

Markdown → print-quality, self-contained HTML. One static binary.

Česká verze dokumentace: README.cs.md

Live demos — real documents produced by mdprint --toc --lof --lot, with numbered equations and cross-references, captioned tables, code, figures, typography and dark mode: čeština · English · Deutsch. Try printing them (Ctrl+P), or open the printed PDFs with page numbers and a table of contents, list of figures and list of tables with page references (mdprint --toc-pages --lof --lot --pdf=all): CS · EN · DE (also in small and large type, linked from the demo pages).

Why

Pandoc is a cannon for this job — a universal converter with a LaTeX toolchain behind it when you want beautiful PDF. Printing straight from the browser is the opposite extreme: zero setup, but no typesetting — no widow control, no non-breaking spaces, tables split mid-row. And no existing tool cares about Czech microtypography at all. mdprint sits in the middle: one binary turns a Markdown file into a single HTML file that reads well on the web and prints beautifully from Chrome (Ctrl+P → PDF or paper), with Czech typographic rules applied by default.

Install

cargo install mdprint

or grab a prebuilt binary from Releases.

Quick start

mdprint dokument.md

This produces dokument.html — a single file with all CSS, fonts and build-time-rendered KaTeX inlined — plus an img_dokument/ folder with the referenced images (deduplicated by content hash). Open the HTML anywhere, print it from Chrome, or publish it as-is.

Features

• Single static binary — no runtime dependencies (a Chrome/Edge only for the optional --pdf and --toc-pages), no JavaScript in the output, fonts (Libertinus, JetBrains Mono) embedded as WOFF2
• KaTeX math rendered at build time ($…$, $$…$$) via a native Rust port — invalid TeX fails the build with a line number. KaTeX cannot line-break display math, so in print mdprint allows breaks at top-level operators and shrinks over-wide formulas in steps so nothing gets clipped; for full control use aligned/split with explicit \\
• Syntax highlighting — static <span>s (syntect); light theme on screen and in print, a dark theme (Base16 Ocean Dark) in dark mode
• Czech, German and English typography applied on the AST — see below
• Hyphenation with embedded Czech, German and English patterns (soft hyphens, words ≥ 6 chars, ≥ 3 chars around each break, headings exempt)
• Print CSS — A4 page setup, page numbers „3 / 8“ in the footer, orphans/widows control, table headers repeated across pages, unbreakable figures and table rows, long code lines wrapped with hanging indent, URLs printed after external links
• Page references in the table of contents (--toc-pages) — mdprint lays the document out in headless Chrome/Edge in all three font sizes, reads the page of every heading from the PDF and writes the numbers into the TOC; print picks the set matching the font size you chose
• PDF export (--pdf, --pdf=all) through the same headless browser
• Equation numbering and cross-references with LaTeX syntax — \label{eq:x} in a display formula, \eqref{eq:x} / \ref{x} / \autoref{x} in text; labels on figures ({#fig:x}) and captioned tables (Table: … {#tbl:x}); numbers are resolved at build time, so the author never types one. See Cross-references and lists
• List of figures and list of tables (--lof, --lot) at the end of the document, with page numbers when --toc-pages is on
• Image handling — paths resolved relative to the input, copies named <slug>-<hash>, cleanup driven by a manifest (mdprint never deletes anything it didn't create), optional for remote images with a disk cache

Typography

The distinguishing feature. Rules are applied only to text nodes (never code, math or URLs), each rule is a separately tested function, and the rule set follows the document language (--lang, front matter lang:).

Czech (--lang cs, default)

Rules follow ČSN 01 6910 and Pecina's Knihy a typografie.

Rule	Before	After
Non-breaking space after single-letter prepositions	šel k lesu	šel k lesu (k U+00A0 lesu)
Number–unit binding	25 °C, 10 kg, 100 %	bound with U+00A0
Abbreviations	č. 12, např. zde	bound with U+00A0
Initials	J. K. Tyl	bound with U+00A0
Dates	1. 1. 2026, 5. července	bound with U+00A0
Czech quotes	"slovo", 'vnořené'	„slovo“, ‚vnořené‘
Dash with spaces	text - vsuvka	text – vsuvka (NBSP before dash)
Number ranges	10-20, 1918-1938	10–20 (en dash)
Ellipsis	atd...	atd…
Multiplication	2 x 3, 40x60	2 × 3, 40×60 (hex 0x1F untouched)
Thousands groups	1 000 000	narrow NBSP U+202F
Apostrophes	d'Artagnan	d’Artagnan

German (--lang de)

Rules follow DIN 5008: „Gänsefüßchen“ quotes (‚nested‘ too), protected spaces inside multi-part abbreviations (z. B., d. h., i. d. R.), references bound to their numbers (Nr. 5, S. 12, Abb. 3), Gedankenstrich ( – with a protected space before it), en-dash ranges, ×, number–unit binding, dates (1. 1. 2026, 3. Oktober), narrow no-break thousands groups, apostrophes (geht's → geht’s). Generated texts are localized (Inhalt, Abb., Version) and the byline date follows DIN (04.09.2026).

English (--lang en)

Chicago-style em dashes (-, --, -- → — closed up), “curly” quotes with ‘nested’ ones, apostrophes in contractions (don't → don’t), en-dash number ranges, ×, number–unit binding (SI), ellipsis. Thousands keep their commas; generated texts use Contents and Fig.

Configuration

Optional mdprint.toml next to the input file (or --config <path>). CLI flags override it; every key is optional.

lang = "cs"                     # cs | en | de; front matter `lang:` overrides this
template = "../my-pack"         # "default" or path to a template pack folder

[paper]
size = "A4"                     # anything valid for CSS @page size
margin = "2.5cm 2cm"

[figures]
prefix = "Obr."                 # figure caption prefix: "Obr. 1: …"
numbering = true
list = false                    # list of figures at the end (--lof)

[tables]
prefix = "Tab."                 # caption prefix for "Table: …" paragraphs
list = false                    # list of tables at the end (--lot)

[equations]
numbering = "labeled"           # labeled | all | none — which $$…$$ get a number

[fonts]                         # primary family overrides (embedded stay as fallback)
serif = "Libertinus Serif"
sans = "Libertinus Sans"
mono = "JetBrains Mono"

[print]
page_numbers = true             # "3 / 8" bottom right (Chrome/Edge 131+)
toc_pages = false               # page numbers in the TOC (needs Chrome/Edge)
font_size = "normal"            # small | normal | large: initial A/A/A and --pdf size
browser = "C:/path/chrome.exe"  # optional; else MDPRINT_BROWSER or auto-detect

[toc]
numbering = true                # "2.1" before entries; off when headings carry own numbers
depth = 3                       # deepest heading level in the TOC (1–6)

[math]                          # math size in em relative to surrounding text
inline = 0.95                   # allowed range 0.5–2.0; defaults calibrated
display = 1.0                   # for Libertinus (page-fit shrink adapts)

[company]                       # footer of a branded template pack;
name = "…"                      # overrides the pack's own defaults
address = "…"
ico = "…"                       # company registration number
dic = "…"                       # VAT number
web = "…"
email = "…"

Documents may carry YAML front matter (title, author, date, version, lang, template, page_numbers, toc_pages, font_size, lof, lot) — title renders a document header, ISO dates are typeset the Czech way (2. 9. 2026), version shows in the byline.

A template pack is a folder with pack.toml, template.css and optional logos (logo-light.png, logo-dark.png), fonts and footer links ([[links]]: label, URL and an optional SVG/PNG icon from the pack — monochrome icons invert to white in dark mode); pass it via --template <folder>. The output stays a single HTML file — pack assets are inlined like the built-in ones. See the pack section in README.cs.md for the manifest schema, or start from the forkable starter pack: mdprint-pack-starter.

CLI reference

Usage: mdprint [OPTIONS] <INPUT>

Arguments:
  <INPUT>  Input Markdown file

Options:
  -o, --output <OUTPUT>      Output HTML file (default: next to input, .html)
      --lang <LANG>          Document language [possible values: cs, en, de]
      --toc                  Generate a table of contents (headings 1-3,
                             see [toc] depth)
      --toc-pages            Add print page numbers to the TOC and lists
                             (implies --toc; needs Chrome or Edge)
      --lof                  List of figures at the end of the document
      --lot                  List of tables (captioned ones) at the end
      --font-size <SIZE>     Initial font size: small | normal | large
      --no-page-numbers      No page numbers in print
      --pdf[=all]            Also write PDF (all = one per font size;
                             needs Chrome or Edge)
      --browser <PATH>       Browser for --pdf/--toc-pages (else
                             MDPRINT_BROWSER or auto-detect)
      --fetch                Download remote images (with a disk cache)
      --config <CONFIG>      Config path (default: mdprint.toml next to input)
      --template <TEMPLATE>  "default" or path to a template pack folder
  -h, --help                 Print help
  -V, --version              Print version

Exit code is non-zero on any error; missing images and invalid TeX report the source line.

Cross-references and lists

Markdown has no equation numbers, and KaTeX only knows a manual \tag. mdprint resolves LaTeX-style labels at build time, so numbers are never typed by hand:

$$ w_{\max} = \frac{5 q L^4}{384 E I} \label{eq:deflection} $$

Formula \eqref{eq:deflection} rests on … (see \autoref{fig:beam} and \autoref{tbl:data}).

![Static scheme of the beam](beam.png){#fig:beam}

Table: Input data {#tbl:data}

| L | q |
|---|---|

• Equations: \label{…} anywhere in a display formula ($$…$$); the label is stripped and \tag{N} added, typeset by KaTeX on the right. By default only labelled equations are numbered (like LaTeX equation vs. equation*); [equations] numbering = "all" numbers every display formula, with \notag / \nonumber as opt-out. A formula with its own \tag keeps it and references show that text. Numbers run through the whole document (1, 2, 3 …).
• Figures: {#fig:x} right after the image (any label name works). Figures without a label still get an anchor fig-N.
• Tables: a paragraph starting with Table:, Tabulka: or Tabelle: (accepted in every language) immediately before a table is its caption, typeset above the table as "Table 1: …"; a trailing {#tbl:x} is the label. Tables without a caption are not numbered and not listed.
• References in text (outside math): \ref{x} → bare number, \eqref{eq:x} → "(3)", \autoref{x} → localized prefix with a non-breaking space: Eq. (3), Fig. 2, Table 1 (cs: rovnice (3), Obr. 2, Tab. 1; de: Gl. (3), Abb. 2, Tab. 1). All become links. An unknown or duplicate label, or \label inside inline math, fails the build with the source line. Labels: letters, digits and : _ . -, starting with a letter.
• Lists: --lof / --lot (front matter lof: true, lot: true; [figures] list, [tables] list) append List of Figures / List of Tables (Seznam obrázků / Seznam tabulek, Abbildungsverzeichnis / Tabellenverzeichnis) after the body — on a new page in print. With --toc-pages they get page numbers from the same measurement as the TOC.

Page numbers — what to expect

• The footer number is a CSS @page margin box: Chrome and Edge 131+ print it, Safari and Firefox currently omit it (nothing breaks).
• TOC page numbers are measured, not estimated: three headless renders (one per font size) take a few seconds. They are exact for printing from Chrome/Edge with the document's paper size and margins at 100 % scale. Another browser or a changed scale in the print dialog may shift them; --pdf output always matches.
• Turn off the browser's own Headers and footers in the print dialog — Chrome then leaves the footer to mdprint.
• An entry whose heading has no anchor in the measured PDF (e.g. a template pushes it off the printable area) is listed in a warning instead of silently losing its number.
• Without a browser, --toc-pages warns and writes the TOC without numbers; --pdf fails. Everything else still needs no runtime dependency.

Build from source

cargo build --release
cargo test

Pure Rust, no C dependencies — syntect runs with the default-fancy feature (fancy-regex instead of oniguruma), so cross-platform builds need nothing beyond a Rust toolchain (MSRV 1.88).

Security

mdprint is a build tool for your own documents and trusts its inputs: raw HTML in Markdown passes through to the output, image paths (including absolute ones and ..) are read from your filesystem and copied next to the output, and template packs are code-equivalent (their CSS and fonts land in every page). Do not run it on untrusted Markdown, configs or packs — and see SECURITY.md for reporting vulnerabilities.

Supply chain: release binaries carry GitHub build provenance attestations (gh attestation verify mdprint-<version>-<target>.zip --repo filip-cokesh/mdprint), crates.io releases are published through Trusted Publishing from the tagged CI run (no long-lived tokens), and every workflow action is pinned to a commit SHA with Dependabot keeping the pins current.

License

Licensed under either of MIT or Apache-2.0, at your option. Bundled fonts and KaTeX assets keep their own licenses — see THIRD-PARTY-LICENSES.