Re: grepp -- Perl version of grep

[Abigail-II]

Interesting, but unfortunally, a bit slow. grep rocks when it comes to speed:

#!/bin/sh
         
FILE=/home/abigail/Words/enable.lst
PATTERNS="
abc
sch[ae]
.*a.*e.*i.*o.*u.*
^[[:xdigit:]]+$
(a|b)(c|d)(e|f)
"   

PROGRAMS=("/bin/grep -E" ./grepp)
OUTPUT=(/tmp/grep.out /tmp/grepp.out)
 
i=0
while [ $i -lt ${#PROGRAMS[*]} ]
do  echo -n ${PROGRAMS[$i]}
    time for pat in $PATTERNS
         do  ${PROGRAMS[$i]} $pat $FILE
         done > ${OUTPUT[$i]}
    i=$((i + 1))
done

i=1
while [ $i -lt ${#OUTPUT[*]} ]
do diff ${OUTPUT[0]} ${OUTPUT[$i]}
   i=$((i + 1))  
done

rm -f ${OUTPUT[*]}



/bin/grep -E
real    0m0.070s
user    0m0.060s
sys     0m0.010s
./grepp
real    0m2.774s
user    0m2.610s
sys     0m0.030s

Abigail

[jdporter]

Such a benchmark is only valid for tests which can be run by both implementations. Depending on what you're trying to do, slow is better than impossible.

[Abigail-II]

Do you have common regular expressions you'd use 'grepp' for, and for which there's no 'grep' equivalent?

Abigail

[graff]

Well, that bit in the pod about matching GB characters was based on personal experience. Access to Perl 5.8's character-set transliteration and unicode-based character semantics for matches (not to mention the very handy "\p" unicode-based character classes) was the main reason I had to write this tool in the first place. The other bells and whistles (handling compressed data, controlling the input record separator) were after-thoughts -- once I started using this thing on real (multi-language, multi-coded) text data, those other things were just handy and easy to add. But the Encode module was crucial.

Maybe some people have tricks to search for a specific GB (or Big5, or Shif-JIS) character using plain-old "grep", but I couldn't figure a way to make it trustworhty for such things, and doing it with Perl just made sense.

[Anonymous Monk]

agreed. grep does not use regex when matching against simple patterns. most grep implementations use a skip-search aglo ala Boyer-Moore for simple patterns.

[Abigail-II]

grep does not use regex when matching against simple patterns.

And neither does Perl. Unless you want to call '(a|b)(c|d)(e|f)' "simple" - this is a pattern where grep wins big time over Perl.

Abigail