sed and character classes , strange results
 

file "file" contains a aa aaa and sed -e 's/[[:lower]]/b/g' file gives :

b bb bbb

all good.

sed -e 's/[[:lower:]]/[[:upper:]]/g' file gives :

[[:upper:]] [[:upper:]] [[:upper:]] [[:upper:]] [[:upper:]] [[:upper:]] [[:upper:]] [[:upper:]] [[:upper:]]

Character classes seems not recognized. I have tried with LC_LANG=C / us / no.

Any ideas ?

You could do it with 'tr'

A sed command would be:
U - lower to upper
L - for upper to lower cases

Sed has also got a 'y' command that could do it, but the solution is quite weird:

It doesn't seem to accept ranges:(

Character classes are valid in regular expressions,
which the replacement string isn't. See sycamorex
post for "how to" ;}


Cheers,
Tink

Also, hheck out this sed tutorial. Someone recommended it on this forum before. It's got a lot of useful information.

THIS solved it.. I've used it in ranges/regexp and didn't think about it being nor valid in a replacement string.

Big thanks to all, checking out the howto:)

This one was new to me, looking forward to test it

It works! How comes it is not in the sed man page? How are we supposed to guess such things? Thank you sycamorex, this will save me from tedious and ineffective y//…

Yves.

It is in the info manual. Keep in mind that it is a GNU extension so it isn't portable.