Unexpected behaviour of /x Regexp modifier?

jvector has asked for the wisdom of the Perl Monks concerning the following question:

Hello Monks,

What am I doing wrong here? I read a line of text and do a simple match. Because I want to Do It Right, I use /x. And it Doesn't Work. The lines look like this:

000.000.002.249-2-P  09-11-2007 15:27:30         102_Low_Alarm
000.000.002.249-2-P  09-11-2007 16:44:18         102_Low_Alarm
000.000.002.249-2-P  09-11-2007 16:44:18         202_Low_Repeat
[download]

and the code is below :

while (<>) {
  chomp;
   my ($meter,$dt,$type) =
     ($_ =~ (\S+)   #ser no
             \s+
            (\d\d-\d\d-\d\d\d\d \d\d:\d\d:\d\d)  # 12-11-2007 13:51:30
             \s+
            (\S+)  #102_Low_Alarm
             $/x);
  ....
}
[download]

Running under -d, I get this:

  DB<15> ($a,$b,$c)= ($line =~ m/(\S+)\s+(\d\d-\d\d-\d\d\d\d \d\d:\d\d
+:\d\d)\s+(\S+)$/x ) ;

  DB<16> x ($a,$b,$c)
0  undef
1  undef
2  undef

  DB<17> ($a,$b,$c)= ($line =~ m/(\S+)\s+(\d\d-\d\d-\d\d\d\d \d\d:\d\d
+:\d\d)\s+(\S+)$/ ) ;
  
  DB<18> x ($a,$b,$c)
0  '000.000.002.249-2-P'
1  '09-11-2007 15:27:30'
2  '102_Low_Alarm'

  DB<19> x $line
0  '000.000.002.249-2-P  09-11-2007 15:27:30  102_Low_Alarm'
[download]

I've just checked and it's not an artefact of running in the debugger; I made 2 versions with and without the /x, and they run differently. I'm obviously missing something very ... obvious.

This signature no verb

Comment on Unexpected behaviour of /x Regexp modifier? Select or Download Code

Replies are listed 'Best First'.
Re: Unexpected behaviour of /x Regexp modifier? by JavaFan (Canon) on Mar 16, 2009 at 19:32 UTC
Because you use `/x` unescaped whitespace is ignored. Including the space between `\d\d\d\d \d\d`. I'm not the biggest fan of `/x`, for this reason. It's so easy to miss a space that needs to be escaped.	[reply] [d/l] [select]
Re: Unexpected behaviour of /x Regexp modifier? by repellent (Priest) on Mar 16, 2009 at 19:34 UTC
You have a literal whitespace in: `(\d\d-\d\d-\d\d\d\d \d\d:\d\d:\d\d) ^ over here` [download] that is ignored when `/x` is used.	[reply] [d/l] [select]
Re: Unexpected behaviour of /x Regexp modifier? by jvector (Friar) on Mar 16, 2009 at 19:53 UTC
Ah. That nicety of the included space had escaped me. For some reason I had imagined that Perl would dwim and not apply the /x within the captures. D'oh. Thank you, as ever! This signature is but a foretaste of The Great Signature to come	[reply]
Re^2: Unexpected behaviour of /x Regexp modifier? by codeacrobat (Chaplain) on Mar 16, 2009 at 20:27 UTC
Hi, you can easily capture spaces literally in \Q ... \E expressions, they are unaffected by /x, see http://perldoc.perl.org/perlre.html#Modifiers `perl -le ' print scalar localtime; $_=scalar localtime; /^\QMon Mar \E (\d+) \s (\d+) : (\d+) : (\d+) \s (\d+)/x; print "$1 $2-$3-$4 $5" ' Mon Mar 16 21:22:30 2009 16 21-19-13 2009` [download] `print+qq(\L@{[ref\&@]}@{['@'x7^'!#2/"!4']});`	[reply] [d/l] [select]
Re^3: Unexpected behaviour of /x Regexp modifier? by johngg (Canon) on Mar 17, 2009 at 10:47 UTC
Another way to do that is to turn `x` off for part of the pattern, either by toggling it, `/ ... (?-x)Mon Mar (?x) ... /x`, or by localising the effect inside parentheses, `/ ... (?-x:Mon Mar ) ... /x`. The technique can also be used for the `i` flag and, I think though I've never tried it, for `s` and `m` as well. I hope this is of interest. Cheers, JohnGG	[reply] [d/l] [select]
Re^2: Unexpected behaviour of /x Regexp modifier? by ysth (Canon) on Mar 17, 2009 at 01:59 UTC
Nope, though that does work in character classes, so you often will see `[ ]` in /x regexes. -- Online Fortune Cookie Search Office Space merchandise	[reply] [d/l]
Re: Unexpected behaviour of /x Regexp modifier? by Marshall (Canon) on Mar 16, 2009 at 23:03 UTC
It appears to me that the basic file format would indicate that a split using white space (/\s+/) and then a sub-parsing using more splits and regex's would work out better rather doing the whole job with one regex from the start. Regex can do just about anything, but I recommend to separate the job into smaller pieces where possible. It looks like you have some sort of an ID, then a date/time combo and then an error_code/description combo. As a general hint for dealing with what looks to be a log file of some sort, convert the date/time into what is called "epoch time". This is a time expressed in seconds, +-seconds from approx. January 1, 1970. This simplifies "date/time" math because everything is now an integer number of seconds, i.e. the time 24 hours earlier than now is -(246060) seconds. The code below shows how to do that. The Perl array of pointers to hash is close to a traditional C 2-D array and the code below shows how to do that too (Perl data structures don't have a one<->one mapping with C). Have fun, hope this helps. Read more... (2 kB)	[reply] [d/l]