I found this code here https://www.tldp.org/HOWTO/Bash-Prompt-HOWTO/x700.html which nicely gives me the number of files in my directory.

ls -1 | wc -l

but I only want to know how many of those files' names start with 2009 (for example 20091210_005037.nc).

I tried ls -1 | wc -l 2009* but that slowly lists all the files and does not seem to give me a number.

  • 2
    ls is doing the listing, so the filenames should be given to ls.
    – muru
    May 28, 2019 at 8:51
  • 2
    You want to list all 2009-files and then count the output lines: ls 2009* | wc -l. Jsut remember the names of the command ls = list , wc = word count (-l = lines). Beware of dangers with ls for odd file namings, though. Using find might be saver.
    – FelixJN
    May 28, 2019 at 8:52
  • 1
    And the option -1 is not neccessary. For a piped output, -1 is default.
    – Philippos
    May 28, 2019 at 8:53

4 Answers 4

set -- 2009*
echo "$#"

This sets the list of positional parameters ($1, $2, ..., etc.) to the names matching 2009*. The length of this list is $#.

The issue with ls -1 | wc -l 2009* is that you execute wc -l directly on the files matching 2009*, counting the number of lines in each. Meanwhile, ls -1 is trying to write to the standard input of wc, which wc is not reading from since it was given an explicit list of files to work on.

You may have wanted to use ls -d 2009* | wc -l. This would have listed all the names that match 2009* (using ls with -d to not list the contents of directories), and would count the number of lines in the output. Note that -1 is not needed if you pipe the result of ls somewhere (unless ls is an alias or shell function that forces column output).

Note also that this would give you the wrong count if any filename contains a newline:

$ touch '2009
> was
> a
> good
> year'
$ ls
$ ls -l
total 0
-rw-r--r--  1 kk  wheel  0 May 28 11:09 2009?was?a?good?year
$ ls -1
$ ls | wc -l
$ ls -1 | wc -l


$ set -- 2009*
$ echo "$#"

(using set and outputting $# additionally does not use any external commands in most shells)

Using find to count recursively:

find . -type f -name '2009*' -exec echo . \; | wc -l

Here, we output a dot for each found pathname in or under the current directory, and then we count the number of lines that this produces. We don't count the filename strings themselves, and instead do it this way to avoid counting too many lines if a filename contains newlines.

With find we're able to more closely control the type of file that we count. Above, we explicitly test for regular files with -type f (i.e. not directories and other types of files). The * pattern in the shell does not distinguish between directories and files, but the zsh shell can use *(.) to modify the behaviour of the pattern to only match regular files (the zsh user would probably use 2009*(.) instead of 2009* in the non-find variations above and below).

Using ** in (with shopt -s globstar in bash, or set -o extended-glob in yash, or in any other shell that may support it), to count recursively:

set -- **/2009*
echo "$#"

The pattern ** matches almost like *, but also matches across / in pathnames.

  • GNU find also supports the -printf action, which saves you a fork and exec, and may be more efficient than -exec echo ..
    – Kevin
    May 28, 2019 at 19:19
  • Or even without GNU, -exec printf '.\n' + will use few fork/exec. May 31, 2019 at 2:40

Tried with below command and it worked fine and got the result

   find . -maxdepth 1 -type f -iname "2009*" | awk '{print NR}'| sed -n '$p'

Note: If you want to under subdirectory also Kindly remove maxdepth option

  • Re: the recent edit, see Kusalananda's answer where this would miscount files with newlines in their name.
    – Jeff Schaller
    May 28, 2019 at 16:43
  • Or just find ... | awk 'END {print NR}' (modulo the newline issue) May 31, 2019 at 2:38
  • This returns a) N if you have 1 or more matches or b) NULL if there's no match. Aspect b) was important to my application. Thanks for the hint!
    – porg

Thanks to the people in the comments, this is the answer to my question:

ls 2009* | wc -l

or using find

find 2009* | wc -l
  • The find command is a bit nonsensical. If you wanted to do a recursive find, you would not use 2009* as the search path.
    – Kusalananda
    May 28, 2019 at 9:12
  • See Kusalananda's answer where this would miscount files with newlines in their name.
    – Jeff Schaller
    May 28, 2019 at 16:41

To count number of files starting with a particular filename using awk.

root@ubuntu$ find . -name "2009*" | awk 'BEGIN{total=0}; {total=total+1} END {print "total files starting with 2009 is " ,total}'
total files starting with 2009 is  4
  • The BEGIN block is not needed. An uninitialised variable in awk will have the value zero when used in an arithmetic context.
    – Kusalananda
    May 28, 2019 at 9:36
  • thanks for pointing out :-)
    – Gowtham
    May 28, 2019 at 10:45
  • And awk's builtin NR already gives the record number, as in Praveen's answer May 31, 2019 at 2:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.