matching a postix regex in bash

casperdaghost · 05-29-2012, 09:32 PM

I am trying to teach myself the postix regex in bash
If I ls this directory, I do have a few instances of casper1234, casper4325, etc.
For eachone it finds i want it to say horay.

Code:

     1  #!/bin/bash
     2
     3  regex='\bcasper[[:alpha:]]{4}'
     4
     5  for i in $(ls -ltr /var/ftp/)
     6  do
     7  if [ $i =~ $regex ] ; then
     8  echo "horay"
     9  sleep 1
    10  fi
    11  done
    12

error is:

Code:

 
./get_postfix: line 7: [: =~: binary operator expected
./get_postfix: line 7: [: =~: binary operator expected
./get_postfix: line 7: [: =~: binary operator expected
./get_postfix: line 7: [: =~: binary operator expected
./get_postfix: line 7: [: =~: binary operator expected

catkin · 05-29-2012, 11:33 PM

What is the error message if you change from [ ... ] to [[ ... ]] ?

EDIT: it might also be informative to print $i before the if test.

lykwydchykyn · 05-29-2012, 11:36 PM

Try double brackets around your test expression to prevent the regex from being expanded by the shell.

EDIT: beaten by minutes...

Nominal Animal · 05-30-2012, 12:46 AM

Quote:

Originally Posted by casperdaghost

Code:

for i in $(ls -ltr /var/ftp/)

Do not use ls this way to obtain file lists; the output varies by locale. If you are only using it to produce interesting strings for testing, then that is okay, but if you intend to parse the output in any way, DON'T.

If you want a list of files in a directory (or directory tree) sorted by last modification time, oldest last, use something like the following in Bash:

Code:

export LANG=C LC_ALL=C
IFS=$'\t'
while read -rd "" year month day hour min sec filename ; do
    # To get integer seconds, use "${sec%%.*}" instead of "$sec".

    echo "$year-$month-$day+$hour:$min:$sec" "$filename"

done < <( find DIR(s) -maxdepth 1 -type f -printf '%TY\t%Tm\t%Td\t%TH\t%TM\t%TS\t%p\0' | sort -rz )

If you want files in subdirectories too, just drop the -maxdepth 1 option. To get oldest first, use sort -z . If you want the times in UTC instead of local time, add export TZ=UTC at the start. You can also access other metadata on the files, like owner, group, size, access mode and so on; see find format directives for details.

This construct uses ASCII NULs as separators. It means all possible file names will work correctly; even the truly strange ones that have newlines or whatnot. Using the C locale as shown, this is guaranteed to handle all file names correctly in Linux.

grail · 05-30-2012, 03:37 AM

SO I see a few problems with your initial attempt:

1. As NA pointed out, please don't use ls (reasons here)

2. Regex operator (=~) requires the use of [[]]

3. You say you have files of the format casper1234, yet your regex says:

Code:

regex='\bcasper[[:alpha:]]{4}'

Here you are looking for casper followed by 4 alpha characters. This would mean it will never match any of your examples.