Talk:Brace expansion

Task is ready for accepting solutions!

As far as I'm concerned, the task description is stable. Feel free to add solutions for other languages, and if you have any problems with the task then drop a line here! --Smls (talk) 17:49, 25 January 2014 (UTC)

There appears to be a bug in the task. Your third test case should have backslashes on the comma by the "always leave in the backslash" rule. See the Perl 6 output. --TimToady (talk) 02:02, 26 January 2014 (UTC)

I edited the task to add the backslashes, after verifying that the Perl code actually does include the backslashes before comma. --TimToady (talk) 02:53, 26 January 2014 (UTC)

Also, with the toy output, it's not at all clear that the current Perl or Python implementations are doing this according to spec. The Perl code claims to follow the spec, but the spec is missing backslashes on the commas. The Python code talks about barfing on code that the specification says it should accept. Maybe we should require a bit more rigor in proving solutions actually pass the spec for backslashed and not-really-meta characters. --TimToady (talk) 02:22, 26 January 2014 (UTC)

Regarding the python code: the commas not in bracers can be parsed as literal chars, but the unmatched bracers as specified by task is not workable. How do you parse "{a,{b,c}" ? As "{a", "b", "c", or as "a", "{b", "c"? Same for closing bracers. This is not just a problem with descending parsers, it's simply ambiguous and counterintuitive, so best treated as a syntax error IMO. The part of spec about "{a}" parsed literally is also not done in python, which can take some workaround but is again not intuitive. --Ledrug (talk) 02:37, 26 January 2014 (UTC)

I understand that Python culture tends more toward throwing an exception if in doubt, but I think the intent of the task was to emulate the shell's notion of how to do this, hence the failsoft aspects, which are pretty easy to do with a decent backtracking engine. It's certainly straighforward in the Perl 6 solution, once I figured out the top level parses with slightly different rules than sublevels do. And it was certainly specced how it was supposed to treat weird characters. --TimToady (talk) 02:49, 26 January 2014 (UTC)