How to replace one char with another in all filenames of the current directories?

In any shell, you can loop over the files whose name contains a space. Replacing the spaces with underscores is easy in bash, ksh and zsh with the ${VARIABLE//PATTERN/REPLACEMENT} construct.

for x in *" "*; do
  mv -- "$x" "${x// /_}"
done

On Debian, Ubuntu and derivatives, you can use the Perl rename (other distributions ship a different program as rename, and that program isn't helpful here).

rename 's/ /_/g' *

An obligatory zsh solution:

autoload zmv
zmv '(*)' '${1// /_}'

An obligatory POSIX solution:

for x in *" "*; do
  y=$(printf %sa "$x" | tr " " "_")
  mv -- "$x" "${y%a}"
done

If you need to rename files in subdirectories as well, and your find supports the -execdir predicate, then you can do

find /search/path -depth -name '* *' \
    -execdir bash -c 'mv -- "$1" "${1// /_}"' bash {} \;

Thank to @glenn jackman for suggesting -depth option for find and to make me think.

Note that on some systems (including GNU/Linux ones), find may fail to find files whose name contains spaces and also sequences of bytes that don't form valid characters (typical with media files with names with non-ASCII characters encoded in a charset different from the locale's). Setting the locale to C (as in LC_ALL=C find...) would address the problem.

You can use rename for this (here assuming the one from util-linux, not the perl one):

cd /path/to/dir
rename ' ' _ *\ *

This will find all files and directories space in the name and replace the space with an underscore. Since it uses glob file matching you need to be in the right directory to start with.

If you want to do recursive matches you can, but you might have to execute the rename a couple times to catch any items in directories that themselves got renamed:

cd /path/to/dir
shopt -s globstar
rename ' ' _ **/*\ *
!!; !!