recursive mkdir

Question

Is there a linux command that I'm overlooking that makes it possible to do something along the lines of: (pseudo)

$ mkdir -R foo/bar/zoo/andsoforth

Or is there no alternative but to make the directories one at a time?

Answer 1

$ mkdir -p foo/bar/zoo/andsoforth

Parameter p stands for 'parents'.

Answer 2

Using mkdir -p is a simple way for most modern OSes:

mkdir -p foo/bar/zoo/andsoforth

However, mkdir -p is not recommended in many manuals. Read documentation for of GNU make and autoconf about problems with using mkdir -p:

• http://www.gnu.org/savannah-checkouts/gnu/autoconf/manual/autoconf-2.69/html_node/Limitations-of-Usual-Tools.html#Limitations-of-Usual-Tools
• http://www.gnu.org/savannah-checkouts/gnu/make/manual/html_node/Utilities-in-Makefiles.html

The cross platform installation and configuration systems have their own safe alternatives for mkdir -p.

CMake to use in shell command line:

 cmake -E make_directory foo/bar/zoo/andsoforth

Autoconf to use in script with preprocessing:

AS_MKDIR_P(foo/bar/zoo/andsoforth)

or:

AC_PROG_MKDIR_P(foo/bar/zoo/andsoforth)

But these solutions require cmake or autoconf (M4) tools to be installed (and possible preprocessing)

You can use also install-sh script with -d option:

install-sh -d foo/bar/zoo/andsoforth

This script is used by autoconf and automake project. I think it must be the safest solution.

At the time I was searching for a cross platform solution for standard /bin/sh without dependences, but haven't found one. Therefore I wrote the next script that may be not ideal, but I think it is compliant to most cross platform requirements:

#! /bin/sh

cdirname() # cross platform alternative for 'dirname'
{
  # $1 - path
  test $# -eq 1 || { echo "Procedure 'cdirname' must have only one parameter. Scripting error."; exit 1; }
  echo "$1" | sed -n -e '1p' | sed  -e 's#//*#/#g' -e 's#\(.\)/$#\1#' -e 's#^[^/]*$#.#' -e 's#\(.\)/[^/]*$#\1#' -
}

mkd() # cross platform alternative for 'mkdir -p'
{
  # $1 - directory to create
  test $# -eq 1 || { echo "Function 'mkd' can create only one directory (with it's parent directories)."; exit 1; }
  test -d "$1"  && return 0
  test -d "$(cdirname "$1")" || { mkd "$(cdirname "$1")" || return 1; }
  test -d "$1" || { mkdir "$1" || return 1; }
  return 0
}

This script can be used for old systems, where option -p for mkdir is absent.

sed-based cross platform version of dirname was added to the code. It works with a way similar to dirname (correct with path /, paths with base name only, paths with trailing /, paths with and without trailing \ns). This function can't work correct if the path has newlines or some invalid characters for current locale. It also replaces any combination of / (//, ///) with single /

Changed line mkdir "$1" || return 1 to test -d "$1" || { mkdir "$1" || return 1; } because mkdir terminates with error if path exists and this check is needed for paths containing constructions like aaa\. (If aaa doesn't exist previous version creates aaa and then tries to create it again).

This version of mkd doesn't generate an error if path already exist (but it still has the possibility to generate such an error in parallel execution) and can't get several directories in command line.

Answer 3

Option -p in command mkdir makes parent directories as needed (no error if existing):

mkdir -p foo/bar/zoo/andsoforth

Another way is, for example using && (error if the specified folder exists):

mkdir foo && mkdir foo/bar && mkdir foo/bar/zoo && mkdir foo/bar/zoo/andsoforth

The chaining operator && is used to chain commands together, such that the next command is run if and only if the preceding command exited without errors.

Of course, the first way is better.

Answer 4

mkdir foo foo/bar foo/bar/zoo/ foo/bar/zoo/andsofort.

This should do it.