DonationCoder.com Software > Find And Run Robot

bug in regex processing? or bug in my regex?

(1/1)

mysteryman:
This is related to the shipmenttracking.alias i updated last night (this morning). Before i posted the alias file, i tried to combine the keyword based, and regex based regex's together into one single regex to simplify future updates (if ups, etc changes their check url)
the regex for ups that i believed would work (and kinda does) is this

^(?:ups (.*)|(1Z.*))-regex
--- End quote ---

my understanding, is that this regex should return either, but not both (.*) or (1Z.*) as $$1. However look at the results of what it does, and feel free to test

$$1=1Z1-ups 1Z1
--- End quote ---
see, this works... it detects the first one works, then skips to the end of the (?:)

$$1= | $$2=1Z1-1Z1
--- End quote ---
this one does not work. It appears that each time it processes a ()  that does not lead with (?: that it counts up one variable. It seems it does this irregardless of matching. This first (.*) gets counted as 1, even though it does not match (seen by the null return of $$1), and should be ignored (correct me if i'm wrong), then it procedes to check the other half of the (?:) . It does return a match, but instead of returning a match on $$1 like it should (again, correct me if i'm wrong), it returns it on $$2. making it very difficult to make one alias work for both of them


if there is another way to do this, or if i am using the incorrect syntax for the regexp, please let me know...

f0dder:
Aren't you supposed to have parens around the whole OR-expression? Christ, my regex-fu is dusty, and I'm too lazy to go into the living-room to pick up my copy of Mastering Regular Expressions. But IIRC it should look something like...
^(?:ups ((.*)|(1Z.*)))
-regex
--- End quote ---

mysteryman:
yes, it should, except you missed the earlier (?: that IS an opening parrenthesis for the OR, but it is a special one to prevent capturing/recording the information found (ie, so you will capture "1Z1", NOT "ups 1Z1" in the string ups 1Z1)

a little easier if you break it up (yes i know it would not work like this, but it does make more sense)

^(?:
  ups (.*)
|
  (1Z.*)
)

make more sense now? it at least does to me, but i could still be wrong

EDIT: oh, and btw, if you reverse the regexp, they reverse... 1Z works, and UPS does not... added colors too

jgpaiva:
I can see your problem, mysteryman! (oh, and thanks for teaching me how to use the ?: after parenthesis, i had never understood how that worked).

The problem here is that both the non-matched and the matched groups get numbers, and you'd like to have only the matched groups to be numbered.
I think this requires someone with a higher "regex-fu" (like f0dder called it) than i have, to tell us how this is supposed to work by the regex rules.

BTW: do you know how would other programs do it? (i couldn't find any program that would do a parsing by groups like farr does :()

Navigation

[0] Message Index