Hi.

I have a data file containing the following columns:

$1 $2 $3 $4
record no. date time data

The date format at present is: yyyy/mm/dd
I need to put the date in the format yymmdd, ie drop the leading two digits and then remove the backslashes.

The time format at present is: hh:mm:ss.ms
I need to put this in the format hhmmss, ie drop the milisec part and remove the colons.

I only know basic sed commands and can't figure this out...

Can the individual characters of a string be referenced in a shell script?

Hi,

This should work:

sed 's%[0-9][0-9]\([0-9][0-9]\)/\([0-9][0-9]\)/\([0-9][0-9]\)%\1\2\3%' infile

- You can change sed's separator (/) to anything you like, I used %. This to avoid escaping the / in the date string.
- Sed can use backreferencing. I.e. something in the searchstring surrounded by () (need to be escaped) can be represented by \1 (\2 \3 etc) in the replace string.

Take a look here (it's about halfway down the article) for another sed backreference explanation/example.

The same can be done for the time string, but that's for you to figure out

If anything is unclear or you cannot get it to work: Just ask.

Thanks. Appreciate that.

Do you mean put an individual character into a variable? Awk will do this, but I don't have the syntax in front of me.