Shell Programming and Scripting

Globbing (aka pathname expansion) never occurs within any quotes (single or double).

In that case, the string "*two*.txt" in the for loop is not expanded. The loop will always execute one time. That one time, the value of $i will be the literal string "*two*.txt". However, when the ls command is executed, the value of i is expanded to "*two*.txt" and then, at that time, since it is unquoted, that's where the globbing occurs (in short, file globbing will occur after variable expansion, if the variable expansion is unquoted).

Since globbing never occurs within quotes, that's why you notice that it never happens when you quote both the for's list and the variable in the loop.


That's correct, but let's be clear about where the field splitting is occuring. What is being split is the unquoted expansion of $i when the ls command is run. Pathname expansion, when it occurs, happens after field splitting (it actually happens after everything, except quote removal).


In this particular instance, those two loops give the same result, but they are not interchangeable. Two very different things are happening. In the first instance, ls is called once, with three arguments. In the second, it is called three times, with one argument each time.

In the first, "[[:digit:]]*" is quoted, so no globbing occurs when the for loop's list is evaluated. The for loop will only execute once. After $i is expanded, the resulting "[[:digit:]]*" will not be split into fields since it does not contain any IFS characters (by default, space, tab, and newline). And then, finally, pathname expansion will occur and correctly deal with the whitespace that those filenames contain, because the globbing occurs after the field splitting step. But, remember, since the for loop list is quoted, it can only ever result in one word and the loop will only execute once (not much point in having a loop, in that case ... just write: ls [[:digit:]]*).

In the second version, since the for loop's list is not quoted, [[:digit:]]* will expand to all the filenames that match the pattern and the loop will execute once per filename. So, obviously, these two forms are not equivalent. Depending on what you are doing, you may need access to each individual filename, which is not possible with the first version.

Example that makes it clear (1 iteration versus 3):

Code:

$ j=0; for i in "[[:digit:]]*"; do printf '%d: ' $((++j)); ls $i; done
1: 1 2 3.txt    3 two 1.txt     9 10 11.txt
$ j=0; for i in [[:digit:]]*; do printf '%d: ' $((++j)); ls "$i"; done
1: 1 2 3.txt
2: 3 two 1.txt
3: 9 10 11.txt

Globs inside quoted strings are never expanded.

Regards,
Alister

Thank you very much for all of the information!!!! That's awesome! I tried pseudocoder's first script that was deleted and it seems to have worked as far as I could tell. But the warning by Alister made me a bit paranoid. And everything after that is a bit over my head. I wasnt' sure just how to adapt this part:

Code:

for i in *; do
    case $i in
        file[12].txt) continue;;
    esac
    ... further processing ...
done

I would definitely like to know exactly how this works. I'm trying to understand the problem that arises with 'IFS characters' and how to control/manipulate text in files. How can this script be set up to perform what I need to happen? Thanks again.

Best regards.

I think drewk's solution should fit your needs. Assuming that the files you want to process all have the extension .c, and that your html header file (everything up to textarea) is h.html, and that your html footer file (everything after textarea) is f.html:

Code:

for f in *.c; do
    cat h.html "$f" f.html > "${f%.c}.html"
done

Alister you are the sh glob god.