Question

I have to delete every word containing at least one number from each file given through the command line as parameter. This is my code:

while [ "$*" != "" ]; do
    if [ ! -f $1 ]
        then echo "$1 not file"
    else
        sed -ie  "s/[^ ]*[0-9][^ ]*//g" $1
    fi
    shift
done

It works perfectly if I have only one file, but if I have more it gives the same result for each. After I run the script in each file there will be the result of the first one.

Can someone tell me what I’m missing?

EDIT2:

This is what I’m running now:

while [ "$#" -ne 0 ]; do
for file in "$@"; do
    if [ ! -e "$file" ]; then
        printf "file doesn't exist: %s\n" "$file"
        continue;
    fi
    if [ ! -f "$file" ]; then
        printf "not a file: %s\n" "$file"
        continue;
    fi
done
    sed -i "s/[^ ]*[0-9][^ ]*//g" "$file"
done

I was talking about the done for the while loop and the done for the for loop; but even without that my script keeps running.

EDIT:

Basically the same thing just a bit different. I have to delete the second and fourth word from each line from each file (word only contain alphanumeric characters). Its’ not working properly and I cant find the error. This is my code:

while [ "$*" != "" ]; do
    if [ ! -f $1 ]
        then echo "$1 not file"
    else
        sed -ie  's/^\( *[^ ]+\) +[^ ]+\(.*\)/\1\2/
                  s/^\( *[^ ]+\)\( +[^ ]+\) +[^ ]+\(.*\)/\1\2\3/g' $1
    fi
    shift
done

Answer 1

the while loop condition should check if there are no arguments, and if there are it should continue. So the right form would be

while [ "$#" -ne 0 ]; do

Now, what you really want is to get each argument and do something with it. That automatically implies a for loop. So what you really should be doing is

for file in $@; do

Now, a file can have spaces in it, so getting that filename and checking if it is really a file should be quoted, and you should also check for existance first

for file in "$@"; do
    if [ ! -e "$file" ]; then
        printf "file doesn't exist: %s\n" "$file"
        continue;
    fi
    if [ ! -f "$file" ]; then
        printf "not a file: %s\n" "$file"
        continue;
    fi
    sed -i "s/[^ ]*[0-9][^ ]*//g" "$file"
done

I could expand more on sed where the -i switch is restricted only to GNU sed. Apart from that you may also want to keep a backup of that file in case something goes wrong.
But this is another topic I guess.