Shell Programming and Scripting

That's a rather pointless construct. It's inefficient (requires fork-exec to create a subshell and run ls), prone to breakage if any of the file names contain an IFS character (by default, space, tab, and newline), and is bound by the system's exec()'s limit (ARG_MAX).

A simpler, safer, more efficient alternative: for txtfile in "$system"/*.txt

Regards,
Alister

Thank you all for ur efforts, unfortunately none of the suggested solutions has worked for me.
I managed to get the file name, the question is how to inser the file name before the concatenation happens.


generated txtfile: no such file or directory


has generated missing argument to -exec


xargs: ksh: No such file or directory

what OS are you on?
Try:

Code:

find . -type f -name '*.txt' | xargs -I {} -i /bin/ksh 'echo //$(dirname {}); cat {}'

Linux Ubuntu 10.04.2

Hi,

If i understand what you want, and if there is no IFS characters on filenames and dirnames, this could do the trick.

In my example, i suppose that all dirname with txt files to cat are named Class*.
But you can remove the -name 'Class*' to include all dirs.

Code:

#!/bin/bash
echo "Let's do the job . . ."
for txtdir in $(find . -type d -name 'Class*'); do
    # Remove old output file.
    rm $txtdir.out &>/dev/null
    for txtfile in $(find $txtdir -name '*.txt'); do
        [ -e $txtdir.out ] && echo "Processing $txtdir.out . . ."
        echo "  Adding $(basename $txtfile)."
        echo "// $(basename $txtfile)" >>$txtdir.out
        cat $txtfile >>$txtdir.out
    done
done
echo "Done."

Hello, miss_dodi:

Unless I missed it, you never made it clear if the output filename varies with directory and if so how to choose its name. To collect the desired concatenation of all .txt files in a directory, with each file's contents preceded by its filename, the solution that follows will create a file named ALL-TEXT-FILES.txt in each directory

You mention that you have 145 directories, but neglected to explain how the code is expected to visit them. Do you have a list to feed the script, either via a pipe or command line arguments? Or are they all in a hierarchy which can be simply traveresed with find from a single root location? I will assume the later and the following script can take a single argument, the location of this starting directory. If absent, the current working directory is assumed.

Code:

#!/bin/sh

find "${1:-.}" -type d -exec sh -c '
        for d; do
                out=$d/ALL-TEXT-FILES.txt
                for f in "$d"/*.txt; do
                        { [ -f "$f" ] && [ -r "$f" ]; } || continue
                        printf "//%s\n" "${f##*/}" >> "$out"
                        cat "$f" >> "$out"
                done
        done 
' sh {} +

I tested it and it works as I intend.

However, there is a bug in this code (which is also present in some of the other suggestions). It's unlikely to be triggered, but it's lurking ... sleeping ... hoping.

In case anyone would prefer to find it themselves...

***** CAUTION: SPOILERS AHEAD *****





If a directory happens to contain a file whose name is identical to the output file, cat will enter an infinite loop of reading-writing to itself until the machine explodes. The non-lazy solution would be to use a unique tempfile (or at least a filename that is guaranteed to be outside the traversal).

Regards,
Alister