Replace custom strings in files under Linux

Question

I posted this question here on purpose, although - for my case - it is laTex related. The problem itself is generic though.

I've got several *.tex -source files containing references to images, eg image1.jpg and image2.png . I want to search all source file for a specified set of extensions (in this case jpg and png ) and replace them with their pdf counterparts. Consequently I want to end up with the references image1.pdf and image2.pdf . To complicate the problem, I've got a list of 20 files (image3.jpg, image 4.png etc.) that I don't want changed.

Is there a simple solution out there (sed-based or any tool suggestions?) which might help? I'm no regular expressions guru, though. ;)

Answer 1

A easier way is:

• image3.jpg, image4.png --> image3.JPG, image4.PNG
• *.jpg, *.png --> *.pdf
• image3.JPG, image4.PNG --> image3.jpg, image4.png

Answer 2

I'd suggest using find to get the file list, filter it with grep -v to remove the files you don't want changed and then use xargs to run on those files sed -r -i 's/image([12])\\.(jpg|png)/image\\1.pdf/g' . Take care, sed -i performs replacements in place, so it's better to backup your files in case something goes wrong. You can also use sed -ibak to let sed make a backup with bak extension before it modifies a file.

Answer 3

Replace the filenames on your blacklist into some "magic" strings so that they will not look like filenames and will not be matched by anything.

cat your.tex | \
sed 's/image3.jpg/MMMAAAGGGIIICCC1/g' | \
sed 's/image4.png/MMMAAAGGGIIICCC2/g' | \
# ... all others on your blacklist
# then do regexp replacements
sed 's/\(image[0-9]\+\)\.\(png\|jpg\)/\1.pdf/g' | \
# ... convert all "magics" back
sed 's/MMMAAAGGGIIICCC1/image3.jpg/g' | \
sed 's/MMMAAAGGGIIICCC2/image4.png/g' | \
# ... and many others
# then output
cat > output.tex