Literate Haskell seemed really promising (it starts in a comment by default and has very complex end-comment markers), but unfortunately, the end-comment markers are so complex I think they break several other languages merely by existing; unfortunately, I don't think you can use non-default options like -q in polyglot challenges

@ais523 Bash nice. It sounds like Black would require a major refactor to Whitespace. Maybe push it down to start at line 5.

just finishing the explanation now

and explanation posted

hmm, what would be nice would be a syntax highlighter, which highlighted for each language: code that's never parsed; code that's parsed, but never executed; string literals whose value is discarded; and code that actually runs

when we finally do a big blog-post style explanation of one of the polyglots (I guess 50 languages is a milestone in its own right…), it'll be a good visual way to explain how each of the polyglots works

I'm not sure how automated it could be though

"50. … 1024 bytes"

thanks to Mistah Figgins for pointing out that my 1025 byte submission was a byte too large :-)

more language ideas: 7 is probably doable (and predates the challenge by just a couple of days!), but will be painful (and likely lock the length of the program mod 3, but that at least is probably easy to work around)

the main issue is that 7 doesn't like unmatched closing parentheses, but its definition of a "closing parenthesis" is the 3-bit sequence 110 (aligned to a 3-bit boundary)

thus at the start of the program, you'd need a long string of opening parentheses to balance them

exiting the program cleanly might also be a little nontrivial, but hopefully not too hard

@ais523 regarding ␠␠ token issues, did you look at converting an ␠␠ to a ␠␉? I know Chance changed some unneeded tabs to spaces recently for legibility. Might buy us a couple characters.

oh, that probably does indeed save bytes

despite likely making the program harder to read

I haven't really been concentrating on golfing, more on fitting the languages in

the golfing's typically been done by other entries

That's how most of mine have been. Seems easier to notice things to golf when I'm not fighting with compatibility. Last couple answers of mine I've tried to do a round of golfing before even working on the inserting the next language. Once I get started adding something new all thoughts of golfing tend to go out the window

fitting into byte limits used to be a notable problem, although not so much any more, as we're mostly finding plenty of short languages

although there are languages like COW which are a) easily added, but b) very verbose, and thus haven't really fit in yet

I know people have been looking at it; we should try to get it a number that's terse to express

66, perhaps

although 55 is almost as good

(I'm assuming it only has character-at-a-time output; I can't quite remember)

COW uses non-wrapping integers (I just verified that by reading the source of and/or running the implementations), 54 and 56 thus seem like the best known options.

Cow will let you output just the raw integer, so it's better than most bf variants in that regard. Even with that it's 63 characters though, which is one more than we have to not regress VIP-wise. If it doesn't break too many things, that might actually work out this time. I'll have to give it a shot.