I was working on a shell script that needed to look at some lines in a bunch of files and perform some data mining. I started it in bash, so I am writing it in bash even though dave0 notes that I should have started in in perl. Point taken… I suffer for my mistake.

After a quick google I learned that a nice way to do what the topic of this post describes can be done using

    IFS='
    '
    declare -a foo=( $(cat $file) )

Which is great! Right?

Well, it turns out that it really sucks. For a few dozen lines this takes 8 seconds on my system, and it doesn’t account for shell startup or spinning up the disk.

    s=$(date +%s)
    declare -a foo=( $(cat $file) )
    e=$(date +%s)
    echo $(( $e - $s ))

8 seconds, really!?

So I tried a variant…

    s=$(date +%s)
    declare -a foo=( $(< $file) )
    e=$(date +%s)
    echo $(( $e - $s ))

Still 8 seconds.

But I know that bash isn’t always slow… maybe if I read one line at a time:

    s=$(date +%s)
    declare -a foo=( )
    while read txt ; do
            foo[${#foo[@]}]=$txt
    done < $file
    e=$(date +%s)
    echo $(( $e - $s ))

Now, it’s 0 seconds.

Well what about zsh?⌗

Since you asked… zsh runs all these in 0s.

I hate bash.

I am rewriting this in perl.