Newbie question: how to replace a bunch of text in a file?

Question

Good afternoon,
I have a very stupid problem here: I want to replace text in a file.
At first glance, this is very simple: just do:
sed -e 's/letter_a/alpha/g' test.txt > test2.txt
and the word "letter_a" will be replaced by the word "alpha" everywhere in the file "test.txt" and the results will be placed in file "test2.txt".
However my problem is a bit more complex:
I want to do different replacements in one file.
If I do:
sed -e 's/letter_a/alpha/g' test.txt
then the result will be shown on my screen, but I want the results to be stored in my file "test.txt".
I have already tried to do:
sed -e 's/letter_a/alpha/g' test.txt > test.txt
but this simply empties the file "test.txt".
As I have more than 50 replacements to do, I cannot do something like:
sed -e 's/letter_a/alpha/g' test.txt > test1.txt
sed -e 's/letter_b/beta/g' test1.txt > test2.txt
sed -e 's/letter_c/gamma/g' test2.txt > test3.txt
+...+
mv test50.txt test.txt
rm test1.txt
+...+
rm test50.txt
(It's possible in theory, but it's really too much work)
Do you know an easy command that does the following:
Replace "letter_a" by "alpha" in <filename> (without needing to specify another file where the output will be written)?
Thanks
Dominique
Edited by: scampsd on Oct 5, 2011 5:17 PM

Answer 1

Why don't you write a wrapper script that renames the input file to a temp file and run the sed to the original name. Something like:
<pre>
#!/bin/sh
# $1 = old string
# $2 = new string
# $3 = input/output file
cp -p $3 $3.zz$$ # preserves i-node of file
sed -e s/"$1"/"$2"/g < $3.zz$$ > $3
rm $3.zz$$
</pre>