Re: Pattern matching when there are exception strings

Replies are listed 'Best First'.
Re^2: Pattern matching when there are exception strings by davidrw (Prior) on Sep 21, 2005 at 13:14 UTC
That wouldn't match the string "ALPHA, I AM" ... also, a character class can't deal with the "XL5 ALPHA" exception.. .. which leads what to the author is looking for -- the zero-width negative look-behind assertion "(?<!pattern)" (see perlre). `use strict; use warnings; while(<DATA>){ print /(?<!XL5 )(?<![#_])ALPHA/ ? "OK\n" : "NOT OK\n"; } __DATA__ I AM ALPHA AND GOOD ALPHA I AM ALPHA AND BETA I AM ALPHA AND #ALPHA I AM ALPHA AND _ALPHA I AM ALPHA AND XL5 ALPHA I AM #ALPHA I AM _ALPHA I AM XL5 ALPHA __OUTPUT__ OK OK OK OK OK NOT OK NOT OK NOT OK` [download]	[reply] [d/l]
Re^3: Pattern matching when there are exception strings by pboin (Deacon) on Sep 21, 2005 at 13:27 UTC
You're right, and I'll take my lumps for being wrong. But what I haven't quite figured out is why `[^_#]ALPHA` wouldn't match on 'ALPHA, I AM'. Is it because the character class has to match _some_ character? My train of thought was that the negation would be primary. "Is there an underscore or a hash leading?" "No."	[reply] [d/l]
Re^4: Pattern matching when there are exception strings by chester (Hermit) on Sep 21, 2005 at 13:41 UTC
The `(?<!pattern)` construct is called "zero-width" because it matches without consuming any characters. `[]` is not zero-width. It has to match one or more characters.	[reply] [d/l] [select]
Re^5: Pattern matching when there are exception strings by sauoq (Abbot) on Sep 21, 2005 at 16:21 UTC
Re^4: Pattern matching when there are exception strings by davidrw (Prior) on Sep 21, 2005 at 14:45 UTC
But what I haven't quite figured out is why* ^_#ALPHA wouldn't match on 'ALPHA, I AM'. Is it because the character class has to match _some_ character?* Yes -- it has to match exactly one character ... you can, however, use the quantifiers: `/[A]B/; # must be 'AB' /[A]?B/; # 0 or 1, so 'AB' or 'B' or 'CB' /[A]+B/; # 1 or more, so 'AB', or 'AAAB' /[A]B/; # 0 or more, so AB' or 'AAAB' or 'B' or 'CB'` [download] My train of thought was that the negation would be primary. "Is there an underscore or a hash leading?" "No."* "Is there a character that is (NOT) an underscore or a hash" The "NOT" being there iff there's the leading carat.	[reply] [d/l]
Re^2: Pattern matching when there are exception strings by chester (Hermit) on Sep 21, 2005 at 13:19 UTC
That wouldn't match "ALPHA" on a line by itself; it demands a non-# non-_ character before the ALPHA. If you do it this way you probably want a zero-width assertion. My first thought is to do a global replacement of all the exceptions, to remove them from the data. Then search what's left for the strings you do want to count. Performance-wise I don't know how great this would be, but it's easier to read (and possibly easier to write if it needs to be automated for a large amount of search strings). `while(<>){ s/(?:_\|#\|XL5 )ALPHA//g; print if /ALPHA/; }` [download] edit: davidrw beat me to it.	[reply] [d/l]
Re^3: Pattern matching when there are exception strings by Roy Johnson (Monsignor) on Sep 21, 2005 at 13:45 UTC
Splicing out things as you suggest would leave the possibility (however remote) that a false match would be created by the pasted-together remnants. Like if you removed "XL5 ALPHA" from "ALXL5 ALPHAPHA". See my split-and-grep recommendation for a similar technique that isn't subject to the pasting problem. Caution: Contents may have been coded under pressure.	[reply]