[rehoot]

This is a minor fix for anyone who uses scripts to capture the output from ls and use that between Mac OS X (10.3 and 10.4). The problem was that the default date format changed from day/month to month/day. I vaguely remember a similar change in the past.

I run a home-made sync between a 10.3 PowerMac and a 10.4 MacPro, and this relies on the output of ls -l. I used sed to parse the columns of the outupt and then rearrange the columns. I created pipe-separated output for my script, but I will present a simpler version here:

ls -lT|sed -e 's/^\([^\ ]*[\ ]*\)\([^\ ]*[\ ]*\)\([^\ ]*[\ ]*\)\([^\ ]*[\ ]*\)\([^\ ]*[\ ]*\)\([^\ ]*[\ ]*\)\([^\ ]*[\ ]*\)\([^\ ]*[\ ]*\)/\1\2\3\4\5\7\6\8/'


The sed command contains many copies of the string: \([^\ ]*[\ ]*\)
The first caret (^) means to find this string at the start of the line. The [^\ ]* is a regular expression (regexp) to find "not a space" one or more times. The [\ ]* finds one or more spaces in a row (the space is preceded by a slash). These two expressions are inside "slash-prefixed parentheses." Each repetition of the stuff in the parentheses can be output at the end of the statement where \1 represents the first original column and \7 represents the seventh original column. You can rearrange or delete the numbers on the right to alter the output. One side effect is that the column alignment might look funny. Note that I reversed 7 and 6.

[drp]

The order that ls and other command line utilities format dates and numbers is based on your current locale settings.

If you type "locale" at the command prompt, it'll show you your current settings. As and example, a typical system will be unset and default to the "C" type. Similar to this:

LANG=
LC_COLLATE="C"
LC_CTYPE="C"
LC_MESSAGES="C"
LC_MONETARY="C"
LC_NUMERIC="C"
LC_TIME="C"
LC_ALL="C"

So an ls will use a format similar to:

-rwxr-xr-x 1 root wheel 1068844 Jul 1 2006 bash

Changing the locale/language will change the date to the correct local format. As an example, Great Britain uses the inverted format:

export LANG=en_GB.UTF-8

Will change the format to:

-rwxr-xr-x 1 root wheel 1068844 1 Jul 2006 bash

As a result, if your scripts are relying on the output from system commands, it's best to set the locale environment variables in the script so you're guaranteed what the output will look like.

[Hal Itosis]

Originally Posted by rehoot 

but I will present a simpler version here:

ls -lT|sed -e 's/^\([^\ ]*[\ ]*\)\([^\ ]*[\ ]*\)\([^\ ]*[\ ]*\)\([^\ ]*[\ ]*\)\([^\ ]*[\ ]*\)\([^\ ]*[\ ]*\)\([^\ ]*[\ ]*\)\([^\ ]*[\ ]*\)/\1\2\3\4\5\7\6\8/'

Goodgawd man! What are you trying to do? (Surely "finding spaces" isn't the ultimate goal).

If that's the "simple" version, I guarantee there's probably an even simpler version that will work.

(try awk maybe... or stat -f even)

[rehoot]

thx.

The "locale" program works on 10.4 but not 10.3. I could change 10.4 to have the date format that I don't want but that matches the format from 10.3. The ls program in 10.3 does not seem to respond to the international settings in System Preference, and I don't see a locale program. I realize that my duct-tape solution is fragile.

[rehoot]

Originally Posted by Hal Itosis 

Goodgawd man! What are you trying to do? (Surely "finding spaces" isn't the ultimate goal).

If that's the "simple" version, I guarantee there's probably an even simpler version that will work.

(try awk maybe... or stat -f even)

Yes, it is an ugly and fragile script. The ugly sed thing is as much of a UNIX oddity as it is a solution to anything. I use awk for similarly crazy scripts. The space delimited output from ls is reliable (within a given version of the OS) becuse I stop parsing before I get to the file name, which is on the far right when you use the -T option. My main home-made archiving system feeds data into MySQL, and that has been working well.