UNIX for Dummies Questions & Answers

I have text with upper and lower case words. I want to find something and replace it with something new. But it should match the case - Meaning - it should replace old upper cased word with NEW upper case word and lower with lower.

example:

Code:

this text is very simple TEXT.

now I want to replace both cases of "text" with word to match with case.
output should be

Code:

this word is very simple WORD.

text is replace with word and TEXT with WORD.

Is there a way to do this in sed or some other command - without the use of a loop.
Thanks in Advance

Whenever you replace something in a text file you either have to memory map the whole in C, make changes, then close the file. Or. Loop through the original file processing line by line and write the new file. In Linux

Code:

sed -i

will do this - update in place. (It does the loop behind closed curtains). Other UNIXES, no.

We need:
Your shell and your unix OS to go further.

And it would be nice to know why loops are out of the question. Sounds misguided to me.

Your requirements aren't very clear.

What should the output be if your input is:

Code:

this text is very simple TEXT in the context of a TEXTILE mill.
This Text is in a mixed case ConTeXt.

when you're replacing text with word?

Will the original text to be replaced always contain the same number of characters as the replacement text? If not, what should happen with the text:

Code:

ThIs iS.

when you want to change is to eater?

Hi Jim - Here is my OS:

Code:

$>uname -o
GNU/Linux

And I am trying to reduce my lines in my code, hoping that awk or sed has something to take care of case even for the replace code.

Don - it should change all occurrences of "text/TEXT" and change it to the appropriate "word/WORD". Mixed case is not in my file and would never be there, so it doesn't matter in case of mixed case.
its like global replace, but it has to take care of cases too:
text should be replaced with word
TEXT should be replace with WORD
TeXt should be replaced with WoRd (not critical)
find string and replace string will always have same number of characters.

Hope I am clear now.

So maybe you want something like:

Code:

#!/bin/ksh
awk -v from="${1:-text}" -v to="${2:-word}" '
BEGIN {	lf = tolower(from)
	uf = toupper(from)
	lt = tolower(to)
	ut = toupper(to)
}
{	gsub(lf, lt)
	gsub(uf, ut)
}
1'

If you save this in a file (for example tester) and make it executable with:

Code:

chmod +x tester

then you can invoke it with:

Code:

./tester from_string to_string < input_file > output_file

If you invoke it without operands it will default to text and word, respectively. This will not work reliably if any characters in the 1st operand have a special meaning in an extended regular expression (but for alphanumeric strings this is not an issue). This will not modify mixed case input, but you said that is not an issue.

If you want to try this on a Solaris/SunOS system, change awk to /usr/xpg4/bin/awk, /usr/xpg6/bin/awk, or nawk.

Thanks for your time Don. Really appreciate your help.