Markdown to PDF

ATX and underlined headings, bold, italic, inline code, fenced blocks opened with ``` or ~~~, unordered and ordered lists nested up to three levels, blockquotes, horizontal rules, links and backslash escapes. Tables, raw HTML, footnotes, strikethrough and task lists are not implemented, the parser is a small subset written for this page rather than a full CommonMark engine, so anything it does not recognise is typeset as ordinary text. An image reference becomes an [image: alt] marker where the picture would have gone.

The PDF is built with the standard fonts every reader already carries, Helvetica or Times for the body, Courier for code, and those encode Latin-1 plus a small set of extras. Curly quotes, dashes, bullets and accented Latin letters come through fine. A few symbols are substituted on the way in, so an arrow prints as -> and ≤ prints as <=. What is left over, which in practice means Chinese, Japanese, Cyrillic, Greek and emoji, becomes a question mark, and the counter under the options tells you how many were replaced.

No, a link is drawn as text, with its address printed after the label if you leave that option on.

It is hard-wrapped at the width of the tinted band, character by character, so nothing runs off the page. There is no syntax highlighting: code is set in Courier at 92% of the body size.

Three things get a file turned away, an extension that is not one of the Markdown or text ones, a size over 8 MB, and a NUL byte in the first 4 KB, which means the file is binary whatever its name claims. Each of those is reported by name rather than failing quietly.