[SOLVED] sed to replace first and second instance

mhouston100 · 07-17-2011, 11:26 PM

Guys,

I might just be crazy or particularly dumb today but what I'm trying to do is have sed replace the first and second instance of a character on a line but not after that.

Basically it is a csv file with a date formatted as dd/mm/yyyy and I need to completely remove the slashes. No problem using:

Code:

sed 's/\///g' < file > outputfile

The only problem is that one of the columns after this may also contain these slashes and shouldn't be removed.

I must have read every single sed page on the net today to find a way to do it to no avail! Is it as simple as :

Code:

sed 's/\///1' < file > outputfile

And then run the same on the output file again? these two slashes will always be the first and second slashes in the file. I guess this would accomplish it but I would rather find the right way to do it!!

As always any help is greatly appreciated!

syg00 · 07-18-2011, 12:03 AM

I prefer to use regex differently - look for what you know exists, and concentrate on that.
If the date is always of that format, and always occurs before any other slash, the following should work.

Code:

sed -r 's@([^/]*[[:digit:]]{2})/([[:digit:]]{2})/([[:digit:]]{4}.*)@\1\2\3@'

If the leading zeroes could be missing you can also accommodate that with a minor change.

mhouston100 · 07-18-2011, 12:33 AM

Thanks for the quick reply syg00, tried your command and not quite what I expected:

I ran the command on the file with:

Code:

sed -r 's@([^/]*[[:digit:]]{2})/([[:digit:]]{2})/([[:digit:]]{4}.*)@\1\2\3@' < inputfile.csv > outputfile.csv

And the resulting file had no '\' removed, it was basically a copy of the original. So to test I just ran:

Code:

echo test1\test2\test3\test4\test5\ | sed -r 's@([^/]*[[:digit:]]{2})/([[:digit:]]{2})/([[:digit:]]{4}.*)@\1\2\3@'

and the output is test1test2test3test4test5 - Removing all of the slashes! This seems to be the result of everything I have tried, it will either remove all or none.

I did test with multiple '\\'s around the string just randomly and it seemed to remove 1 or two in a series and leave the rest, for instance:

Code:

echo test1\\\\test2\\\\test3\\\\ | sed -r 's@([^/]*[[:digit:]]{2})/([[:digit:]]{2})/([[:digit:]]{4}.*)@\1\2\3@'

Will result in test1\\test2\\test3

mhouston100 · 07-18-2011, 12:39 AM

Sorry, here is a sample line of the file:

Code:

"14#####1","AUD","64#####09","10/07/20","501","AUTOMATIC DRAWINGS"," ","W1####97          ORIX/FLEET       1#####",9####.45

I've hashed out a few numbers but that is a good sample, the thrid '\' in ORIX\FLEET needs to stay, but the date slashes need to go!

MAybe I should be looking at AWK instead of sed?

grail · 07-18-2011, 12:51 AM

Could this not be simpler and just be:

Code:

sed -r 's@([^/]+)/([^/]+)/@\1\2@' file

mhouston100 · 07-18-2011, 01:03 AM

Wow, you are a legend, works like a charm!!

I might have to look into it further cause I still cant wrap my head around the context of sed, every time I think I have it straight it turns out I was completely wrong!

But thanks again.

grail · 07-18-2011, 02:00 AM

No Probs

It is as syg00 said, sometimes it pays to look at what you have exactly as being too generic can cause issues.