Bash filename completion explanation

I did some more research for you and here is what I found - the key to autocompletion is the bash command complete. You can print the rules for vim using:

complete -p vim

Likewise you can remove these specific rules with:

complete -r vim

This command resets it to defaults - the usual paths and file names completion without any extra logic. That's probably what you want.

For more info check out help complete or man bash and look for the section about complete command (it's a bash built-in, hence it's documented in the bash manpage).

One last point - the above changes only affect the current bash session, if you want to remove the vim rules after every login put the complete -r vim to your ~/.bashrc.

Hope that helps :)

Thanks to Muru for pointing me to the solution.

Apparently my issue was that every time my function was called it always returned the full list of possible matches. I needed to only return those possible completions that match what the user has typed so far. Once I return only a single item, then bash will auto-complete the rest of the line.

This is the function that works. Before adding each possible completion, I do a quick check to see if the completion starts with the word that was typed originally. I do this using regular expressions. I also had to test to see if they haven't entered anything yet - in which case I always return everything.

I also changed it to make some more variables "local".

_use () {
    local word="${COMP_WORDS[$COMP_CWORD]}"
    local files=( `compgen -f -X "!*.use" /opt/use/` )
    local output=()
    for file in "${files[@]}"; do
        fileBaseName=`basename $file .use`
        if [[ $fileBaseName =~ ^$word.* ]] || [ "$word" == "" ]; then
            output+=("$fileBaseName")
        fi
    done
    COMPREPLY=( ${output[@]} )
}