What is the difference between using `sh` and `source`?

What is the difference between sh and source?

source: source filename [arguments]
    Read and execute commands from FILENAME and return.  The pathnames
    in $PATH are used to find the directory containing FILENAME.  If any
    ARGUMENTS are supplied, they become the positional parameters when
    FILENAME is executed.

And for man sh:

NAME
       bash - GNU Bourne-Again SHell

SYNOPSIS
       bash [options] [file]

COPYRIGHT
       Bash is Copyright (C) 1989-2004 by the Free Software Foundation, Inc.

DESCRIPTION
       Bash  is  an sh-compatible command language interpreter that executes commands read from the standard input or from a file.  Bash also incorporates
       useful features from the Korn and C shells (ksh and csh).

       Bash is intended to be a conformant implementation of the IEEE POSIX Shell and Tools specification (IEEE Working Group 1003.2).

When you call source or . (the one is an alias to the other. source cmd not POSIX - kind of bashism), you load and execute a bash script into the current bash process. So you can

• read variables set in the sourced script,
• use functions defined within it.
• and even execute forks and/or subprocess if script do this.

When you call sh, you initiate a fork (sub-process or child) that runs a new session of /bin/sh, which is usually a symbolic link to bash. In this case, environment variables set by the sub-script would be dropped when the sub-script finishes.

Caution: sh could be a symlink to another shell.

Practical sample

For example, if you want to change current working directory by a specific manner, you could not do

$ cat <<eof >myCd2Doc.sh
#!/bin/sh
cd /usr/share/doc
eof

$ chmod +x myCd2Doc.sh

This won't do what you expect:

$ cd /tmp
$ pwd
/tmp
$ ~/myCd2Doc.sh
$ pwd
/tmp

because current working dir is part of environment and myCd2Doc.sh would run in a subshell.

But:

$ cat >myCd2Doc.source <<eof
# Shell source file
myCd2Doc() {
    cd /usr/share/doc
}
eof

$ . myCd2Doc.source
$ cd /tmp
$ pwd
/tmp
$ myCd2Doc
$ pwd
/usr/share/doc

Have a look at mycd function!! (With [tag:bash] completion based on Associative Array).

Execution level $SHLVL

$ cd /tmp
printf %b '\43\41/bin/bash\necho This is level \44SHLVL.\n' >qlvl.sh

$ bash qlvl.sh 
This is level 2.

$ source qlvl.sh 
This is level 1.

Recursion (when a script run from itself)

$ cat <<eoqlvl2 >qlvl2.sh 
#!/bin/bash

export startLevel recursionLimit=5
echo This is level $SHLVL started:${startLevel:=$SHLVL}.
(( SHLVL < recursionLimit )) && ./qlvl2.sh
eoqlvl2
$ chmod +x qlvl2.sh

$ ./qlvl2.sh 
This is level 2 started:2.
This is level 3 started:2.
This is level 4 started:2.
This is level 5 started:2.

$ source qlv2.sh 
This is level 1 started:1.
This is level 2 started:1.
This is level 3 started:1.
This is level 4 started:1.
This is level 5 started:1.

A little futher

$ sed '$a ps --sid $SID fw' qlvl.sh >qlvl3.sh
$ chmod +x qlvl3.sh 
$ export SID
$ read SID < <(ps ho sid $$)
$ echo $SID $$
8983 8983

( Current PID ($$ == process Id) are same identifier than SID (session ID). It's not alway true.)

$ ./qlvl3.sh 
This is level 2.
  PID TTY      STAT   TIME COMMAND
 8983 pts/10   Ss     0:00 /bin/bash
10266 pts/10   S+     0:00  \_ /bin/bash ./qlvl3.sh
10267 pts/10   R+     0:00      \_ ps --sid 8983 fw

$ . qlvl3.sh 
This is level 1.
  PID TTY      STAT   TIME COMMAND
 8983 pts/10   Ss     0:00 /bin/bash
10428 pts/10   R+     0:00  \_ ps --sid 8983 fw

Dot . is an alias of source. So the only difference between two command are slash replaced by space.

And a final test:

$ printf %b '\43\41/bin/bash\necho Ending this.\nsleep 1;exit 0\n' >finalTest.sh

$ bash finalTest.sh 
Ending this.

$ source finalTest.sh
Ending this.

... You may notice a different behaviour between the two syntaxes. ;-)

The main difference is that they are executed in a different process.

So if you source a file foo which does a cd, the sourcing shell (e.g. your interactive shell in the terminal) is affected (and its current directory will change)

If you execute sh foo the cd does not affect the sourcing shell, only the freshly created sh process running foo

Read the Advanced Bash Scripting Guide.

^{That difference is not specific to Linux; every Posix implementation would have it.}

As others have mentioned, when you run sh test.sh, any changes that test.sh makes to your shell environment won't persist after the process has ended.

However, also note that any element of your environment that isn't exported (e.g., variables, aliases, and shell functions) won't be available to the code in test.sh when it is executed as a subprocess (i.e. with sh test.sh).

For example:

$ cat > test.sh
echo $foo
$ foo=bar
$ sh test.sh
$ . test.sh
bar

Example 2:

lap@my-ThinkPad:~$ cat test.sh
#!/bin/sh
cd /etc
lap@my-ThinkPad:~$ sh test.sh 
lap@my-ThinkPad:~$ pwd
/home/savoury
lap@my-ThinkPad:~$ source test.sh 
lap@my-ThinkPad:/etc$ pwd
/etc
lap@my-ThinkPad:/etc$ 

source (or . ) - runs inside current shell and changes its attribute/environment.

sh do fork and runs in a subshell and hence can't change attributes/environment.

For example

My shell script is -

elite12!rg6655:~/sh_pr [33]$ cat changeDir.sh
#!/bin/bash
cd /home/elt/rg6655/sh_pr/justdir
pwd
echo $$

My Current Shell -

elite12!rg6655:~/sh_pr [32]$ echo $$
3272

Process id of my current shell is 3272

Running with the source -

elite12!rg6655:~/sh_pr [34]$ source changeDir.sh
/home/elt/rg6655/sh_pr/justdir
3272
elite12!rg6655:~/sh_pr/justdir

Observer two things -

1. The process id (3272) is same as my shell, which confirms source executes in the current shell.
2. cd command worked and directory got changed to justdir.

Running with sh -

elite12!rg6655:~/sh_pr [31]$ sh changeDir.sh
/home/elt/rg6655/sh_pr/justdir
13673
elite12!rg6655:~/sh_pr

In this case, process id (13673) is different and directory remains the same which means it is running in a different process or subshell.

When you execute a program with sh command:

• your terminal will use sh or Bourne Shell to execute the program .

• a new process is created because Bash makes an exact copy of itself . this child process has the same environment as its parent, only the process ID number is different. (this process called forking )

• you need to need to have execution permission to execute it ( since it is forking )

and when you use source command :

• you execute the program with your default interpreter
• you execute the process in your current terminal ( technically your *nix command interpreted )
• Since the program will be executed in current terminal you dont need to give it execution permission