[SOLVED] xargs: placement of arguments

Problem statement: Given several files tmp1.pdf tmp2.pdf etc, build a single pdf:
Solution:
pdfunite tmp1.pdf tmp2.pdf ... final.pdf

But I don't want to type the n tmp files :-)

I am learning first with three files. After getting some feeling with the commands, I go further with several ones.

The simplest I could figure out:

$ls -l  |xargs -i pdfunite {} final.pdf

It fails. The reason I discover when I substitute echo for pdfunite:

$ls -l  |xargs -i echo {} final.pdf 
tmp1.pdf final.pdf
tmp2.pdf final.pdf
tmp3.pdf final.pdf
tmp4.pdf final.pdf

This is obviously wrong. Let try find:

$find . -maxdepth 1 -name tmp\*.pdf -print0
./tmp4.pdf./tmp1.pdf./tmp3.pdf./tmp2.pdf

I cannot figure out why it is out of order. This is not acceptable, the pages must be ordered. Try sort:

$find . -maxdepth 1 -name tmp\*.pdf -print0 | sort -z
./tmp1.pdf./tmp2.pdf./tmp3.pdf./tmp4.pdf

Fine. Let go on:

$find . -maxdepth 1 -name tmp\*.pdf -print0 | sort -z | xargs -0 -i echo {} final.pdf
./tmp1.pdf final.pdf
./tmp2.pdf final.pdf
./tmp3.pdf final.pdf
./tmp4.pdf final.pdf

No progress. I'm stuck. What do I need to know?

Operating sysyem: OpenSuse 12.3

At a bash command prompt

for i in tmp*.pdf; do list+=" $i"; done; pdfunite $list final.pdf

How about the obvious:

    pdfunite tmp*.pdf final.pdf

If you only have the files on stdin then you can do:

    ls | parallel -j1 -X pdfunite {} final.pdf

You can install GNU Parallel in 10 seconds:

    wget -O - pi.dk/3 | sh

Thank you alland and tange. Both techniques, loop and bash expansion work as long as tmp files are numbered with a single digit (1 to 9). Therefore, the proposed problem was solved.

However, if tmp files range to 10 or more, the first page after bash expansions is tmp10.pdf, then tmp1.pdf ... etc. In this case another approach must be chosen. Parallel I have not tried, just because, right now, I am trying to master xargs. BTW, thanks also for the nice script for installing parallel. I had never thought in running a script direct from the web via wget and sending it to sh. :-) Parallel is a tool I wish to master.

Posts with further suggestions using xargs would be appreciated, for learning purposes.

Originally Posted by otaviolb However, if tmp files range to 10 or more, the first page after bash expansions is tmp10.pdf, then tmp1.pdf ... etc. In this case another approach must be chosen.

Assuming you have tmp1.pdf ... tmp100.pdf:

    pdfunite tmp{1..100}.pdf final.pdf

PS. To master GNU Parallel: Watch pi.dk/1 and read the examples https://www.gnu.org/software/paralle...ment_appending

It works. Thanks again, including for the tips.

What you really should be using is standard shell globbing. That's the preferred way to build a command line with multiple files.

Beware though, as you've discovered, that the files will expand in ascii sorting order, which means that "10" will probably come just after "1", instead of "9". You have to get a bit fancy if you want to match non-zero-padded names with globs.

pdfunite tmp[1-9].pdf tmp1[0-9].pdf tmp2[0-9].pdf ... final.pdf

So you should really get into the habit of always zero-padding numbers in your filenames. "10" will always follow "09". With zero padding, the command would look like this (assuming 2 digits):

pdfunite tmp[0-9][0-9].pdf final.pdf

brace expansion, as tange posted above, can be used as well, and is a good way to deal with files that aren't zero-padded. It does have a drawback, however. Since the whole list of possible filenames is built first, any entries that don't match an actual filename will probably give you "file not found" errors.

globbing doesn't have this problem, as it only expands into files that actually exist. Do note however that if a glob matches nothing, the shell will try to use the raw string as is, unless the nullglob or failglob shell options are set.