|
03:36
lucasb left
08:38
Ven`` joined
10:34
Ven`` left
10:41
Ven`` joined
10:48
Ven`` left
10:57
Ven`` joined
11:17
lucasb joined
12:35
Ven`` left
12:49
Ven`` joined
13:28
Ven`` left
13:54
Ven`` joined
|
|||
| masak | Ven``: I've been thinking about regexes lately. | 13:54 | |
| Ven`` | hi! | ||
| do tell :) | |||
| masak | briefly, | 13:55 | |
| (a) they're absolutely _essential_ for the whole `is parsed` story -- of course. now, `is parsed` isn't currently on the shortlist of things to do Soon, partly because it's a big thing to tackle | |||
| (b) I'm more and more in the mind of sitting down and designing what we need from scratch | 13:56 | ||
| (c) I think we need four things, roughly: (i) strings, (ii) "references"/calls, more on which below, (iii) quantifiers and alternations, and (iv) capture groups | 13:57 | ||
| (d) I want to approach this design by identifying the broad use cases regexes will have. something like "match true/false", "substitute string", "match and then query capture groups", "`is parsed`" -- those are the ones I've been able to think of | 13:58 | ||
| (e) the "references"/calls will do multiple things. the obvious one being refer to other rules/regexes. <®ex> is the syntax to do it in the lexical scope; <.regex> is the syntax to do it in an ambient "context", typically the current 007 parser | 14:00 | ||
| (f) the <whatever> syntax uses something lexical if it can, otherwise the late-bound thing against the context. it also does a named capture, just like () does a positional one | 14:01 | ||
| (g) but the <whatever> syntax can also work with Str (to literal matching), Match objects (and we'll probably want a <$0> syntax or similar for the match objects already captured), and maybe some other types | 14:02 | ||
| (h) in the fullness of time, some regexes could be constructed at compile time. basically if they're "well-behaved enough" and don't return to anything else that's not available statically | 14:03 | ||
| . | |||
| one of the big unanswered questions in the `is parsed` story is "how does something like <EXPR> end up putting a Q.Expr in the Match object?" | 14:05 | ||
| in Perl 6, the answer is actions -- but I've always felt that answer makes sense for grammars, but not so much for `is parsed` | 14:06 | ||
| I think I'd just like for *something* to be there -- in the worst case, we can iterate that into something better | |||
|
15:56
Ven`` left
17:36
lucasb left
18:04
lucasb joined
|
|||