Why is it better to use "#!/usr/bin/env NAME" instead of "#!/path/to/NAME" as my shebang?

Question

I notice that some scripts which I have acquired from others have the shebang #!/path/to/NAME while others (using the same tool, NAME) have the shebang #!/usr/bin/env NAME.

Both seem to work properly. In tutorials (on Python, for example), there seems to be a suggestion that the latter shebang is better. But, I don't quite understand why this is so.

I realize that, in order to use the latter shebang, NAME must be in the PATH whereas the first shebang does not have this restriction.

Also, it appears (to me) that the first would be the better shebang, since it specifies precisely where NAME is located. So, in this case, if there are multiple versions of NAME (e.g., /usr/bin/NAME, /usr/local/bin/NAME), the first case specifies which to use.

My question is why is the first shebang preferred to the second one?

Answer 1

It isn't necessarily better.

The advantage of #!/usr/bin/env python is that it will use whatever python executable appears first in the user's $PATH.

The disadvantage of #!/usr/bin/env python is that it will use whatever python executable appears first in the user's $PATH.

That means that the script could behave differently depending on who runs it. For one user, it might use the /usr/bin/python that was installed with the OS. For another, it might use an experimental /home/phred/bin/python that doesn't quite work correctly.

And if python is only installed in /usr/local/bin, a user who doesn't have /usr/local/bin in $PATH won't even be able to run the script. (That's probably not too likely on modern systems, but it could easily happen for a more obscure interpreter.)

By specifying #!/usr/bin/python you specify exactly which interpreter will be used to run the script on a particular system.

Another potential problem is that on many systems, the #!/usr/bin/env trick doesn't let you pass arguments to the intrepreter (other than the name of the script, which is passed implicitly). This usually isn't an issue, but it can be. Many Perl scripts are written with #!/usr/bin/perl -w, but use warnings; is the recommended replacement these days. Csh scripts should use #!/bin/csh -f -- but csh scripts are not recommended in the first place. But there could be other examples.

UPDATE: the env of FreeBSD has a -S option (copied by GNU coreutils env in 8.30 with a --split-string alias) that allows passing multiple arguments, so for example #!/usr/bin/env csh -f will fail on systems whose shebang accepts only one argument after the interpreter, but #!/usr/bin/env -S csh -f will work on those such as Linux where what's after the interpreter is bundled into one argument. Many env implementations such as NetBSD's, busybox, toybox still don't support it so that won't help you write portable scripts.

I have a number of Perl scripts in a personal source control system that I install when I set up an account on a new system. I use an installer script that modifies the #! line of each script as it installs it in my $HOME/bin. (I haven't had to use anything other than #!/usr/bin/perl lately; it goes back to times when Perl often wasn't installed by default.)

A minor point: the #!/usr/bin/env trick is arguably an abuse of the env command, which was originally intended (as the name implies) to invoke a command with an altered environment. Furthermore, some older systems (including SunOS 4, if I recall correctly) didn't have the env command in /usr/bin. Neither of these is likely to be a significant concern. env does work this way, a lot of scripts do use the #!/usr/bin/env trick, and OS providers aren't likely to do anything to break it. It might be an issue if you want your script to run on a really old system, but then you're likely to need to modify it anyway.

Another possible issue, (thanks to Sopalajo de Arrierez for pointing it out in comments) is that cron jobs run with a restricted environment. In particular, $PATH is typically something like /usr/bin:/bin. So if the directory containing the interpreter doesn't happen to be in one of those directories, even if it's in your default $PATH in a user shell, then the /usr/bin/env trick isn't going to work. You can specify the exact path, or you can add a line to your crontab to set $PATH (man 5 crontab for details).

Kevin's comment points out that Python's virtualenv creates a special case, where the environment installs a Python interpreter in a special directory that's inserted at the front of $PATH. For that particular environment (and perhaps others like it), the #!/usr/bin/env python trick (or python3?) is likely to be the best solution. (I haven't used virtualenv myself.)

Answer 2

Because /usr/bin/env can interpret your $PATH, which makes scripts more portable.

#!/usr/local/bin/python

Will only run your script if python is installed in /usr/local/bin.

#!/usr/bin/env python

Will interpret your $PATH, and find python in any directory in your $PATH.

So your script is more portable, and will work without modification on systems where python is installed as /usr/bin/python, or /usr/local/bin/python, or even custom directories (that have been added to $PATH), like /opt/local/bin/python.

Portability is the only reason using env is preferred to hard coded paths.

Answer 3

Objective Criteria/Requirements:

In determining whether to use an absolute or logical (/usr/bin/env) path to an interpreter in a shebang, there are two key considerations:

a) The interpreter can be found on the target system

b) The correct version of the interpreter can be found on the target system

If we AGREE that "b)" is desirable, we also agree that:

c) It's preferable our scripts fail rather than execute using an incorrect interpreter version and potentially achieve inconsistent results.

If we DON'T AGREE that "b)" matters, then any interpreter found will suffice.

Testing:

Since using a logical path — /usr/bin/env to the interpreter in the shebang — is the most extensible solution allowing the same script to execute successfully on target hosts with different paths to the same interpreter, we'll test it — using Python, due to its popularity — to determine whether it meets our criteria.

1. Does /usr/bin/env live in a predictable, consistent location on POPULAR (not "every") operating systems? Yes:

   • RHEL 7.5
   • Ubuntu 18.04
   • Raspbian 10 ("Buster")
   • OSX 10.15.02

2. Below Python script executed both inside and outside of virtual envelopes (Pipenv used) during tests:

   #!/usr/bin/env pythonX.x
   import sys
   print(sys.version)
   print('Hello, world!')
   

3. The shebang in the script was varied by the Python version number desired (all installed on the same host):

   • #!/usr/bin/env python2
   • #!/usr/bin/env python2.7
   • #!/usr/bin/env python3
   • #!/usr/bin/env python3.5
   • #!/usr/bin/env python3.6
   • #!/usr/bin/env python3.7

4. Expected results: that print(sys.version) = env pythonX.x.  Each time ./test1.py was executed using a different installed Python version, the correct version specified in the shebang was printed.

5. Testing Notes:

   • Tests were exclusively limited to Python
   • Perl, like Python, MUST live in /usr/bin according to the FHS
   • I've not tested every possible combination on every possible number of Linuxy/Unixy Operating System and version of each Operating System.

Conclusion:

Although it's TRUE that #!/usr/bin/env python will use the first version of Python it matches in the user's Path, we can enforce an express preference by specifying a version number such as #!/usr/bin/env pythonX.x.  Indeed, developers don't care which interpreter is found "first"; all they care about is that their code is executed using the specified interpreter they know to be compatible with their code to ensure consistent results — wherever that may live in the filesystem...

In terms of portability/flexibility, using a logical — /usr/bin/env — rather than absolute path not only meets requirements a), b) & c) from my testing with different versions of Python, but also has the benefit of fuzzy-logic finding the same version interpreter even if they live at different paths on different Operating Systems.  And although MOST distros respect the FHS, not all do.

So, where a script will FAIL if the binary lives in a different absolute path than specified in the shebang, the same script using a logical path SUCCEEDS as it keeps going until it finds a match, thereby offering greater reliability & extensibility across platforms.

Answer 4

Specifying the absolute path is more precise on a given system. The downside is that it's too precise. Suppose you realize that the system installation of Perl is too old for your scripts and you want to use your own instead: then you have to edit the scripts and change #!/usr/bin/perl to #!/home/myname/bin/perl. Worse, if you have Perl in /usr/bin on some machines, /usr/local/bin on others, and /home/myname/bin/perl on yet other machines, then you'd have to maintain three separate copies of the scripts and execute the appropriate one on each machine.

#!/usr/bin/env breaks if PATH is bad, but so does almost anything. Attempting to operate with a bad PATH is very rarely useful, and indicates that you know very little about the system the script is running on, so you can't rely on any absolute path anyway.

There are two programs whose location you can rely on on almost every unix variant: /bin/sh and /usr/bin/env. Some obscure and mostly retired Unix variants had /bin/env without having /usr/bin/env, but you're unlikely to encounter them. Modern systems have /usr/bin/env precisely because of its widespread use in shebangs. /usr/bin/env is something you can count on.

Apart from /bin/sh, the only time you should use an absolute path in a shebang is when your script isn't meant to be portable, so you can count on a known location for the interpreter. For example, a bash script that only works on Linux can safely use #!/bin/bash. A script that is only meant to be used in-house can rely on house interpreter location conventions.

#!/usr/bin/env does have downsides. It's more flexible than specifying an absolute path but still requires knowing the interpreter name. Occasionally you might want to run an interpreter that isn't in the $PATH, for example in a location relative to the script. In such cases, you can often make a polyglot script that can be interpreted both by the standard shell and by your desired interpreter. For example, to make a Python 2 script portable both to systems where python is Python 3 and python2 is Python 2, and to systems where python is Python 2 and python2 doesn't exist:

#!/bin/sh
''':'
if type python2 >/dev/null 2>/dev/null; then
  exec python2 "$0" "$@"
else
  exec python "$0" "$@"
fi
'''
# real Python script starts here
def …

Answer 5

Specifically for perl, using #!/usr/bin/env is a bad idea for two reasons.

First, it's not portable. On some obscure platforms env isn't in /usr/bin. Second, as Keith Thompson has noted, it can cause trouble with passing arguments on the shebang line. The maximally portable solution is this:

#!/bin/sh
exec perl -x "$0" "$@"
#!perl

For details on how it works, see 'perldoc perlrun' and what it says about the -x argument.

Answer 6

The reason there is a distinction between the two is because of how scripts are executed.

Using /usr/bin/env (which, as mentioned in other answers, is not in /usr/bin on some OSes) is required because you can't just put an executable name after the #! - it must be an absolute path. This is because the #! mechanism works on a lower level than the shell. It's part of the kernel's binary loader. This can be tested. Put this in a file and mark it executable:

#!bash

echo 'foo'

You will find it prints an error like this when you attempt to run it:

Failed to execute process './test.sh'. Reason:
The file './test.sh' does not exist or could not be executed.

If a file is marked executable and begins with a #!, the kernel (which does not know about $PATH or the current directory: these are user-land concepts) will look for a file using an absolute path. Because using an absolute path is problematic (as mentioned in other answers), someone came up with a trick: You can run /usr/bin/env (which is almost always in that location) to run something using $PATH.

Answer 7

There are two more problems with using #!/usr/bin/env

1. It doesn't solve the problem of specifying the full path to the interpreter, it just moves it to env.

   env is no more guaranteed to be in /usr/bin/env than bash is guaranteed to be in /bin/bash or python in /usr/bin/python.

2. env overwrites ARGV[0] with the name of the interpreter (e..g bash or python).

   This prevents your script's name from appearing in, e.g., ps output (or changes how/where it appears) and makes it impossible to find it with, e.g, ps -C scriptname.sh

[update 2016-06-04]

And a third problem:

3. Changing your PATH is more work than just editing the first line of a script, especially when scripting such edits is trivial. e.g:

   printf "%s\n" 1 i '#!'$(type -P python2) . w | ed foo.py

   Appending or pre-pending a directory to $PATH is fairly easy (although you still have to edit a file to make it permanent - your ~/.profile or whatever - and it's far from easy to script THAT edit because the PATH could be set anywhere in the script, not in the first line).

   Changing the order of PATH directories is significantly more difficult....and far harder than just editing the #! line.

   And you still have all the other problems that using #!/usr/bin/env gives you.

   @jlliagre suggests in a comment that #!/usr/bin/env is useful for testing your script with multiple interpreter versions, "by only changing their PATH / PATH order"

   If you need to do that, it's much easier to just have several #! lines at the top of your script (they're just comments anywhere but the very first line) and cut/copy-and-paste the one you want to use now to the first line.

   In vi, that would be as simple as moving the cursor to the #! line you want, then typing dd1GP or Y1GP. Even with an editor as trivial as nano, it would take seconds to use the mouse to copy and paste.

---

Overall, the advantages of using #!/usr/bin/env are minimal at best, and certainly don't even come close to outweighing the disadvantages. Even the "convenience" advantage is largely illusory.

IMO, it's a silly idea promoted by a particular kind of programmer who thinks that operating systems are not something to be worked with, they are a problem to be worked around (or ignored at best).

PS: here's a simple script to change the interpreter of multiple files at once.

change-shebang.sh:

#!/bin/bash

interpreter="$1"
shift

if [ -z "$(type -P $interpreter)" ] ; then
  echo "Error: '$interpreter' is not executable." >&2
  exit 1
fi

if [ ! -d "$interpreter" ] && [ -x "$interpreter" ] ; then
  shebang='#!'"$(realpath -e $interpreter)" || exit 1
else
  shebang='#!'"$(type -P $interpreter)"
fi

for f in "$@" ; do
  printf "%s\n" 1 i "$shebang" . w | ed "$f"
done

Run it as, e.g., change-shebang.sh python2.7 *.py or change-shebang.sh $HOME/bin/my-experimental-ruby *.rb

Answer 8

Adding another example here:

Using env is also useful when you want to share scripts between multiple rvm environments for example.

Running this on the cmd line, shows which ruby version will be used when #!/usr/bin/env ruby is used inside a script:

env ruby --version

Therefore, when you use env, you can use different ruby versions through rvm, without changing your scripts.

Answer 9

If you're writing purely for yourself or for your job and the location of the interpreter you're calling is always in the same place, by all means use the direct path. In all other cases use #!/usr/bin/env.

Here's why: in your situation the python interpreter was in same place regardless of which syntax you used but for many people it could have installed in a different place. Though most major programming interpreters are located in /usr/bin/ a lot of newer software is defaulting to /usr/local/bin/.

I would even argue to always use #!/usr/bin/env because if you have multiple versions of the same interpreter installed and you don't know what your shell will default to, then you should probably fix that.

Answer 10

For portability and compatiblity reasons it is better to use

#!/usr/bin/env bash

instead of

#!/usr/bin/bash

There are multple possiblities, where a binary may be located on an Linux/Unix system. Check the hier(7) manpage for a detailed description of the file system hierarchy.

FreeBSD for example installs all software, which is not part of the base system, in /usr/local/. Since the bash is not part of the base system, the bash binary is installed at /usr/local/bin/bash.

When you want a portable bash/csh/perl/whatever script, which runs under most Linux Distributions and FreeBSD, you should use #!/usr/bin/env.

Also take note, that most Linux installations have also (hard)linked the env binary to /bin/env or softlinked /usr/bin into /bin which should not be used in the shebang. So don't use #!/bin/env.

Answer 11

Because this is a very old post, it understandably fails to acknowledge the modern Mac OS/X wherein one cannot update /bin/sh or /bin/bash due to Apple's unwillingness to accept the terms of the GPL3 -- even root permissions are insufficient to override this decision in my experience. What this often means is that such scripts may not work at all due the old version of sh and/or bash that are/is installed by Apple. </rant>

Such problems reinforce the value of using #!/usr/bin/env discussed at-length above; however, as previously-stated by other respondents, that may not always be a viable solution. That being said, I've been using the following "shebang" in my Perl scripts for many years on many different Unix-like systems:

eval '(exit $?0)' && eval 'exec perl -S $0 ${1+"$@"}' #-*-perl-*-
& eval 'exec perl -S $0 $argv:q'
        if 0;

• Where eval '(exit $?0)' determines if there is a script to execute.
• Where eval 'exec perl -S $0 ${1+"$@"}' is used if the user's current shell is sh, bash or some similar variant that has built-in eval and exec commands.
• Where #-*-perl-*- identifies the script as perl for my preferred Emacs editor.
• Where eval 'exec perl -S $0 $argv:q' is used if the user's current shell is [t]csh or some similar variant that has built-in eval and exec commands.

It seems to me that this is easily adapted to execute scripts written in other languages, and just the embedded interpreter needs to be changed accordingly:

eval '(exit $?0)' && eval 'exec PREFERRED-INTERPRETER $0 ${1+"$@"}'
& eval 'exec PREFERRED-INTERPRETER $0 $argv:q'
        if 0;