In bash, how to sort strings with numbers in them?

If I have these files in a directory

cwcch10.pdf cwcch11.pdf cwcch12.pdf cwcch13.pdf cwcch14.pdf cwcch15.pdf cwcch16.pdf cwcch17.pdf cwcch18.pdf cwcch1.pdf cwcch2.pdf cwcch3.pdf cwcch4.pdf cwcch5.pdf cwcch6.pdf cwcch7.pdf cwcch8.pdf cwcch9.pdf 

how can I list them in Bash so that they are in ascending numeric order based on the number part of the string. So the resulting order is cwcch1.pdf, cwcch2.pdf, ..., cwcch9.pdf, cwcch10.pdf, etc.

What I'm ultimately trying to do is concatenate the pdfs with pdftk with something like the following

pdftk `ls *.pdf | sort -n` cat output output.pdf 

but that doesn't work as my sorting is wrong.

7 Answers

Something like this might do what you want, though it takes a slightly different approach:

pdftk $(for n in {1..18}; do echo cwcch$n.pdf; done) cat output output.pdf 

Your sort may have the ability to do this for you:

sort --version-sort 

For this particular example you could also do this:

ls *.pdf | sort -k2 -th -n 

That is, sort numerically (-n) on the second field (-k2) using 'h' as the field separator (-th).

You can use the -v option in GNU ls: natural sort of (version) numbers within text.

ls -1v cwcch* 

This does not work with BSD ls (e.g. on OS X), where the -v option has a different meaning.

Use shell expansion directly in a commandline. The expansion should order them properly. If I understand pdftk's commandline syntax properly, this will do what you want:

# shell expansion with square brackets pdftk cwcch[1-9].pdf cwcch1[0-9].pdf cat output output.pdf # shell expansion with curly braces pdftk cwcch{{1..9},{10..18}}.pdf cat output output.pdf 

Or you can try a different approach. When I need to do something like this, I usually try to get my numbers formatted properly ahead of time. If I'm coming into it late and the PDFs are already numbered like your example, I'll use this to renumber:

# rename is rename.pl aka prename -- perl rename script # this adds a leading zero to single-digit numbers rename 's/(\d)/0$1/' cwcch[1-9].pdf 

Now the standard ls sorting will work properly.

Here's a method just using sort:

ls | sort -k1.6n 

Sort -g is used to sort numbers in ascending order.

anthony@mtt3:~$ sort --help | egrep "\-g" -g, --general-numeric-sort compare according to general numerical value 

The following one liner iterates over a file with the names of the PDF files and grabs the numbers only with egrep -o and uses sort -g to sort the numbers in ascending order. Then it feeds these numbers to sed and plugs them in. Then rids the output of duplicates with uniq.

In place of uniq, you can also use awk:

awk '!x[$0]++' 

The above is equivalent to uniq.

What you're looking for is this one liner:

for i in `cat tmp | egrep -o "[0-9]*" | sort -g`; do cat tmp | sed "s/\(^[a-z]*\)\([0-9]*\)\(\.pdf\)/\1$i\3/g" | uniq; done 

Contents of tmp:

anthony@mtt3:~$ cat tmp cwcch10.pdf cwcch11.pdf cwcch12.pdf cwcch13.pdf cwcch14.pdf cwcch15.pdf cwcch16.pdf cwcch17.pdf cwcch18.pdf cwcch1.pdf cwcch2.pdf cwcch3.pdf cwcch4.pdf cwcch5.pdf cwcch6.pdf cwcch7.pdf cwcch8.pdf cwcch9.pdf 

EDIT:

Output of command:

anthony@mtt3:~$ for i in `cat tmp | egrep -o "[0-9]*" | sort -g`; do cat tmp | sed "s/\(^[a-z]*\)\([0-9]*\)\(\.pdf\)/\1$i\3/g" | uniq; done cwcch1.pdf cwcch2.pdf cwcch3.pdf cwcch4.pdf cwcch5.pdf cwcch6.pdf cwcch7.pdf cwcch8.pdf cwcch9.pdf cwcch10.pdf cwcch11.pdf cwcch12.pdf cwcch13.pdf cwcch14.pdf cwcch15.pdf cwcch16.pdf cwcch17.pdf cwcch18.pdf