UNIX for Dummies Questions & Answers

Hi All,
I have created a bourne script that basically wants to split a file up in to different parts. I have this working if the file has all the information on different lines but if it doesn't then it doesn't work.
i.e.
If this is the file
hello
12345
good bye
6789
I could grep all the letters and out and put them in a file as the same with the numbers.
But if they were all on the same line when i use grep it copies the whole line.
i.e.
hello 12345 good bye 6789

So when i do grep '[a-z*]' file > letters
it sends the whole line.
I thought i could do like a sed command that substitutes everything except letters to nothing leaving just the letters then save that to a file.
i.e
sed 's/[a-z]// file > numbers

I hope that makes sense. What i'm trying to say is which is the best way to pull out specific text from a file without getting the whole line??

Thanks
Jason

to pull out lines that contain numbers, you may use:

Code:

grep -v '[a-z*]' somefile > num_file

similarly, to pull out lines containing alphabets, use:

Code:

grep -v '[0-9*]' somefile > letters_file

but in these cases, if a line contains both numbers and alphabets, it is left out. if you want to extract numeric or alphabetic part from such lines, then this approach is not good. you need to use something else then (grep alone can't help)

As you have seen, "grep" is a tool for identifying whole lines. You, however, need to identify words.

A cheap way to do this is to put each word on its own line before using grep:

Code:

tr ' '  '\n' somefile | grep '^[[:alpha:]]'

But then all your data ends up on individual lines. If you needed each line intact (minus the unwanted data of course) then your sed solution would be better.

If this is what you really want to do (see the points Yogesh and gus have raised) you can do that. In fact you have almost already done it, there is just a minor detail missing:

s/something/other/

will change only the first occurrence of "something" in every line to "other". If you want to change any occurrence you have to use the "global"-operator:

s/something/other/g

Hence your script would work this way:

sed 's/[a-z]//g' file > numbers

This will replace every (smallcap) character with "nothing", effectively deleting it. If you want to delete *every* character use: "[a-zA-Z]" or, better, "[^0-9.+-]". The "^" will invert the character mask, meaning "everything NOT mentioned here". The given expression will mean "everything, save for numbers 0-9, literal points *) and pluses and minuses **)".

So, your script should read:

sed 's/[^0-9.+-]//g' file > numbers

*) if you could have decimal numbers
**) if you could have signs

bakunin

Thanks a lot for the answers they really helped and was easy to understand. What about if the file had dates, money and text? As in if you wanted to extract them three seperately, but all the information was on the same line? Would i have to take the approach of putting one a new line after every word?

I used this command and the result was blank. Here is the data i am using from a file.
13/11/07 $50 new shoes
Here is the command i have used
tr ' ' '\n' < a file | grep '^[[:alpha:]]' > temp
tr