Living on the (b)leading edge

Perl 6 is innovative in many ways, and sometimes we don't fully appreciate all the implications, for good or for bad.

There's one I stumbled upon recently: The use of fancy Unicode symbols for built-in stuff. In this case: the `.gist` output of Match objects. For example

my token word { \w+ }
say 'abc=def' ~~ /<word> '=' <word>/;

｢abc=def｣
 word => ｢abc｣
 word => ｢def｣

And that's where the problems start. In my current quest to write a book on Perl 6 regexes, I noticed that the PDF that LeanPub generates from my Markdown sources don't correctly display those pesky ｢｣ characters, which are

$ uni -c ｢｣
｢ - U+0FF62 - HALFWIDTH LEFT CORNER BRACKET
｣ - U+0FF63 - HALFWIDTH RIGHT CORNER BRACKET

When I copied the text from the PDF and pasted into my editor, they showed up correctly, which indicates that the characters are likely missing from the monospace font.

The toolchain allows control over the font used for displaying code, so I tried all the monospace fonts that were available. I tried them in alphabetical order. Among the earlier fonts I tried was Deja Vu Sans Mono, which I use in my terminal, and which hasn't let me down yet. No dice. I arrived at Noto, a font designed to cover all Unicode codepoints. And it didn't work either. So it turns out these two characters are part of some Noto Sans variants, but not of the monospace font.

My terminal, and even some font viewers, use some kind of fallback where they use glyphs from other fonts to render missing characters. The book generation toolchain does not.

The Google Group for Leanpub was somewhat helpful: if I could recommend an Open Source mono space font that fit my needs, they'd likely include it in their toolchain.

So I searched and searched, learning more about fonts than I wanted to know. My circle of geek friends came up with several suggestions, one of them being Iosevka, which actually contains those characters. So now I wait for others to step up, either for LeanPub to include that font, or for the Noto maintainers to create a monospace variant of those characters (and then LeanPub updating their version of the font).

And all of that because Perl 6 was being innovative, and used two otherwise little-used characters as delimiters, in an attempt to avoid collisions between delimiters and content.

(In the mean time I've replaced the two offending characters with ones that look similar. It means the example output is technically incorrect, but at least it's readable).