Python Concepts/Interfacing with Unix

Objective Edit

To describe how to invoke and pass arguments to a python script on the Unix command line.
To describe how to invoke a Unix command from within a Python script.

Lesson Edit

Invoking python from Unix Edit

To enter python in interactive mode:

$ /usr/local/bin/python3.6
Python 3.6.3 (v3.6.3:2c5fed86e0, Oct  3 2017, 00:32:08) 
[GCC 4.2.1 (Apple Inc. build 5666) (dot 3)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> 3+4
7
>>> quit()
$

If your $PATH global variable contains the directory /usr/local/bin, you can invoke python with a file name rather than a path name:

$ echo $PATH
/usr/bin:/bin:/usr/sbin:/sbin:/usr/local/bin
$ python3.6
Python 3.6.3 (v3.6.3:2c5fed86e0, Oct  3 2017, 00:32:08) 
[GCC 4.2.1 (Apple Inc. build 5666) (dot 3)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> 5 * 7
35
>>> quit()
$

Let test.py be the name of a python script.

$ cat test.py
# test.py

print ('3+4 =',3+4)

$ 
$ python3.6 test.py
3+4 = 7
$ 
$ cat  test.py | python3.6
3+4 = 7
$
$ python3.6 < test.py
3+4 = 7
$ 
$ echo print\(\'3\+4\ \=\'\,3\+4\) | python3.6
3+4 = 7
$ 
$ echo "print ('3+4 =',3+4)" | python3.6
3+4 = 7
$ 
$ echo python3.6 ./test.py | ksh
3+4 = 7
$ echo python3.6 ./test.py | csh
3+4 = 7
$

Let the first line of your script contain the path name of python with appropriate syntax:

$ cat test.py
#!/usr/local/bin/python3.6

# test.py

print ('11-5 =',11-5)

$
$ ls -la test.py
-rwxr--r--  1 name  staff  62 Nov  8 16:13 test.py # Ensure that the script is readable and executable.
$ 
$ ./test.py # The script can be executed as a stand-alone executable.
11-5 = 6
$

Passing arguments to python Edit

$ cat test.py
#!/usr/local/bin/python3.6

# test.py

import sys

print ('''
sys.argv[1] = "{}"
sys.argv[2] = "{}"
sys.argv[3] = "{}"
Number of arguments received = {}
'''.format(
sys.argv[1] ,
sys.argv[2] ,
sys.argv[3] ,
len( sys.argv[1:] )
))
$
$ ./test.py    arg_1    arg\ 2    'arg 3'    "arg 4"    arg\*5

sys.argv[1] = "arg_1"
sys.argv[2] = "arg 2"
sys.argv[3] = "arg 3"
Number of arguments received = 5

$

Invoking Unix from python Edit

This section uses function subprocess.run() to pass a command to the Unix command line. Then we review the result of executing (or trying to execute) the Unix command. This section introduces several new objects such as CompletedProcess and the various error conditions and subclasses of those conditions.

The Unix command date is simple, can be invoked without arguments and is not likely to fail:

>>> import subprocess
>>> 
>>> a = subprocess.run('date') ; a
Sat Nov  9 16:57:51 GMT 2019
CompletedProcess(args='date', returncode=0)
>>> a.args
'date'
>>> a.returncode
0
>>> 
>>> type(a)
<class 'subprocess.CompletedProcess'>
>>>

The date command above displayed the date and returned code 0. It was invoked and went to normal completion. To capture the output of the invocation:

>>> a = subprocess.run('date',stdout=subprocess.PIPE) ; a
CompletedProcess(args='date', returncode=0, stdout=b'Sat Nov  9 17:35:36 CST 2019\n')
>>> a.args
'date'
>>> a.returncode
0
>>> a.stdout
b'Sat Nov  9 17:35:36 UST 2019\n'
>>> a.stdout.decode()
'Sat Nov  9 17:35:36 UST 2019\n'
>>>

To execute a Unix command with arguments:

>>> a = subprocess.run(('ls','-laid','.'),stdout=subprocess.PIPE) ; a
CompletedProcess(args=('ls', '-laid', '.'), returncode=0, stdout=b'27647146 drwxr-xr-x  22 name  staff  748 Nov  8 16:33 .\n')
>>> a.args
('ls', '-laid', '.')
>>> a.returncode
0
>>> a.stdout.decode()
'27647146 drwxr-xr-x  22 name  staff  748 Nov  8 16:33 .\n'
>>>

Notice the difference between the next two commands:

>>> a = subprocess.run(( 'echo', 'print()', '|', 'python3.6' ),stdout=subprocess.PIPE) ; a
CompletedProcess(args=('echo', 'print()', '|', 'python3.6'), returncode=0, stdout=b'print() | python3.6\n')
>>>
>>> a.args
('echo', 'print()', '|', 'python3.6') # 4 arguments.
>>> a.stdout
b'print() | python3.6\n'
>>> 
>>> a = subprocess.run( 'echo "print(2*7)" | python3.6',stdout=subprocess.PIPE,shell=True) ; a
CompletedProcess(args='echo "print(2*7)" | python3.6', returncode=0, stdout=b'14\n')
>>> 
>>> a.args
'echo "print(2*7)" | python3.6' # 1 argument.
>>> a.stdout
b'14\n' # This is probably what you wanted.
>>>

Sending data to a pipe Edit

This process depends on the Unix command echo. On my system echo produced inconsistent results:

$ echo echo -n | ksh
$ echo echo -n | csh
$ echo echo -n | bash
$ echo echo -n | sh
-n
$

Python executable pecho.py produced consistent, predictable results across all 4 shells.

#!/usr/local/bin/python3.6

# pecho.py
import sys

Length = len( sys.argv[1:] )

if Length not in (1,2) :
        print ('pecho.py: Length {} not in (1,2).'.format(Length), file=sys.stderr)
        exit (39)

if Length == 2:
    if sys.argv[1] not in ('-n','-N') :
        print ('pecho.py: expecting "-n" for sys.argv[1]', file=sys.stderr)
        exit (38)
    print (sys.argv[2], end='') ; exit(0)

if sys.argv[1] in ('-n','-N') :
        print ('pecho.py: received only "-n"', file=sys.stderr)
        exit (37)

print (sys.argv[1])

The perl script below will be formatted and sent to a pipe for execution by perl.

import re
import subprocess

perlScript = r'''
# This is a 'short' perl script.

$a = 5 ;
$b = 8 ;

$newLine = "
";

(length($newLine) == 1) or die '$newLine'." must not contain extraneous white space.";

print ('a+b=',$a+$b,$newLine) ;
print ("b-a=",$b-$a,"\n") ;

'''

repl = """'"'"'"""
pS1 = re.sub( "'", repl, perlScript )
pS1 = "'" + pS1 + "'"

a = subprocess.run( "./pecho.py -n " + pS1 + " | perl" ,stdout=subprocess.PIPE ,stderr=subprocess.PIPE, shell=True)

print (''' 
a.args =   
{}
a.returncode = {} 
a.stdout = {}
{}]]]] 
'''.format(
a.args,
a.returncode ,
a.stdout ,
a.stdout.decode(),
),end='')

a.args =
./pecho.py -n '
# This is a '"'"'short'"'"' perl script.

$a = 5 ;
$b = 8 ;

$newLine = "
";

(length($newLine) == 1) or die '"'"'$newLine'"'"'." must not contain extraneous white space.";

print ('"'"'a+b='"'"',$a+$b,$newLine) ;
print ("b-a=",$b-$a,"\n") ;

' | perl
a.returncode = 0
a.stdout = b'a+b=13\nb-a=3\n'
a+b=13
b-a=3
]]]]

Processing errors Edit

If you execute command a = subprocess.run(.....) and there is an error, the value of a remains unchanged. This could be confusing later. The following code works well:

import subprocess
import sys
a = error = ''
try : a = subprocess.run(.........)
except : error = sys.exc_info()
# You expect exactly one of (a, error), preferably a, to be True.
set1 = {bool(v) for v in (a, error)}
if len(set1) != 2 : print ('Internal error.') ; exit (99)

TimeoutExpired Edit

try :  a = subprocess.run(('sleep', '100'), stdout=subprocess.PIPE, timeout=2 )
except subprocess.TimeoutExpired:
    print ('Detected timeout.')
    t1 = sys.exc_info()
    print ('sys.exc_info() =',t1)
    isinstance(t1,tuple) or exit(99)
    len(t1)==3 or exit(98)
    isinstance(t1[1], subprocess.TimeoutExpired) or exit(97)
    print('''                                                                                                                                
cmd = {}                                                                                                                                     
timeout = {}                                                                                                                                 
stdout = {}                                                                                                                                  
stderr = {}                                                                                                                                  
'''.format(
t1[1].cmd,
t1[1].timeout,
t1[1].stdout,
t1[1].stderr,
))

Detected timeout.
sys.exc_info() = (<class 'subprocess.TimeoutExpired'>, TimeoutExpired(('sleep', '100'), 2), <traceback object at 0x10186c988>)

cmd = ('sleep', '100')
timeout = 2
stdout = b''
stderr = None

CalledProcessError Edit

If check is true, and the process exits with a non-zero exit code, a CalledProcessError exception will be raised.

try :  a = subprocess.run('cat crazy_file_name', shell=True, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE )
except subprocess.CalledProcessError :
    error = sys.exc_info()
    print ('''
CalledProcessError detected.
cmd = {}
returncode = {}
stdout = {}
stderr = {}
'''.format(
error[1].cmd,
error[1].returncode,
error[1].stdout,
error[1].stderr
))

CalledProcessError detected.
cmd = cat crazy_file_name
returncode = 1
stdout = b''
stderr = b'cat: crazy_file_name: No such file or directory\n'

Normal processing of errors Edit

If you're not using the check or timeout arguments, most errors, perhaps all, can be processed by referring to the contents of the CompletedProcess object.

a=error=''
try :  a = subprocess.run(('cat', 'crazy_file_name'), stdout=subprocess.PIPE, stderr=subprocess.PIPE )
except : error = sys.exc_info()

set1 = {bool(v) for v in (a, error)}
if len(set1) != 2 : print ('Internal error.') ; exit (99)

if a :
    print ('a =',a)
    print ('''
args = {}
returncode = {}
stdout = {}
stderr = {}
'''.format(
a.args ,
a.returncode ,
a.stdout ,
a.stderr ,
))

a = CompletedProcess(args=('cat', 'crazy_file_name'), returncode=1, stdout=b'', stderr=b'cat: crazy_file_name: No such file or directory\n')

args = ('cat', 'crazy_file_name')
returncode = 1
stdout = b''
stderr = b'cat: crazy_file_name: No such file or directory\n'

In the above invocation of subprocess.run(...), the exception was not taken and all information about the error is available in a.returncode and a.stderr.

With a few well chosen Unix commands, subprocess.run gives you easy access to much good information about the Unix environment.

# Does the file ../ObjOrPr/test.py exist?
>>> try : subprocess.run('head -1 test.py', shell=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE,cwd='../ObjOrPr' )
... except : sys.exc_info()
... 
CompletedProcess(args='head -1 test.py', returncode=0, stdout=b"print ('4 + 9 =',4+9)\n", stderr=b'')
# Yes, and it's readable.
>>> try : subprocess.run('cat /dev/null >> test.py', shell=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE,cwd='../ObjOrPr' )
... except : sys.exc_info()
... 
CompletedProcess(args='cat /dev/null >> test.py', returncode=1, stdout=b'', stderr=b'/bin/sh: test.py: Permission denied\n')
>>> 
# But it's not writeable.

try and except statements provide good clean info about errors:

>>> try : subprocess.run('pwd', stdout=subprocess.PIPE, stderr=subprocess.PIPE,cwd='../OjOrPr' )
... except : sys.exc_info()
... 
(<class 'FileNotFoundError'>, FileNotFoundError(2, "No such file or directory: '../OjOrPr'"), <traceback object at 0x10199e2c8>)
>>>

Without try and except:

>>> subprocess.run('pwd', stdout=subprocess.PIPE, stderr=subprocess.PIPE,cwd='../OjOrPr' )
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Library/Frameworks/Python.framework/Versions/3.6/lib/python3.6/subprocess.py", line 403, in run
    with Popen(*popenargs, **kwargs) as process:
  File "/Library/Frameworks/Python.framework/Versions/3.6/lib/python3.6/subprocess.py", line 709, in __init__
    restore_signals, start_new_session)
  File "/Library/Frameworks/Python.framework/Versions/3.6/lib/python3.6/subprocess.py", line 1344, in _execute_child
    raise child_exception_type(errno_num, err_msg, err_filename)
FileNotFoundError: [Errno 2] No such file or directory: '../OjOrPr': '../OjOrPr'
>>>

Addendum Edit

Default value shell=False Edit

The reference recommends default value shell=False wherever possible because the method manages the conversion of python string to Unix string so that the Unix executable receives an accurate copy of each argument supplied.

>>> a = subprocess.run( ('ls','-laid','.','..',) ,stdout=subprocess.PIPE ,stderr=subprocess.PIPE)
>>> print(a.stdout.decode())
27647146 drwxr-xr-x   28 name  staff   952 Nov 16 14:45 .
26949099 drwxrwxrwx  170 name  staff  5780 Nov  8 14:36 ..

>>>
>>> a = subprocess.run( ('echo','Te$t p9tte8n:~!@:$%^&*()_+') ,stdout=subprocess.PIPE ,stderr=subprocess.PIPE)
>>> print(a.stdout.decode())
Te$t p9tte8n:~!@:$%^&*()_+

>>>

However:

>>> a = subprocess.run( ('echo','Te$t p9tte8n:~!@:$%^&*()_+', '|', 'od', '-h') ,stdout=subprocess.PIPE ,stderr=subprocess.PIPE)
>>> print (a.returncode)
0
>>> print(a.stdout.decode())
Te$t p9tte8n:~!@:$%^&*()_+ | od -h # This probably is not what you wanted.

>>>

Try again with shell=True.

>>> a = subprocess.run( ('echo "Te$t p9tte8n:~!@:$abcde%^&*()_+" | cat') ,stdout=subprocess.PIPE ,stderr=subprocess.PIPE,shell=True)
>>> print(a.stdout.decode())
Te p9tte8n:~!@:%^&*()_+ # '$t' and '$abcde' are missing from output.

>>>

'$t' and '$abcde' are missing from output above because a Unix string beginning with '"' provides variable substitution.

>>> a = subprocess.run( ("echo 'Te$t p9tte8n:~!@:$abcde%^&*()_+' | cat") ,stdout=subprocess.PIPE ,stderr=subprocess.PIPE,shell=True)
>>> print(a.stdout.decode())
Te$t p9tte8n:~!@:$abcde%^&*()_+ # No variable substitution in string beginning with "'".

>>> a = subprocess.run( ("echo 'Te$t p9tte8n:~!@:$abcde%^&*()_+' | wc") ,stdout=subprocess.PIPE ,stderr=subprocess.PIPE,shell=True)
>>> print(a.stdout.decode())
       1       2       32 # 31 characters is correct length of pattern. "wc" adds new line.

>>>

Unix strings and Python strings Edit

Provided that the string to be passed to Unix does not contain a single quote "'", the simplest way to prepare the string is to enclose it in single quotes:

>>> st1 = '\n  line 1  \n line 2\n'
>>> print (st1, end='')

  line 1  
 line 2
>>> st1a = "'" + st1 + "'"
>>>
>>> a = subprocess.run( ("echo " + st1a) ,stdout=subprocess.PIPE ,stderr=subprocess.PIPE, shell=True)
>>> print (a.args)
echo '
  line 1  
 line 2
'
>>> a.stdout.decode()[:-1] == st1
True
>>>

If the string contains a single quote, convert the single quote to a pattern which Unix will recognize.

>>> st1 = '''ls -laid "Bill's file"''' # The string is: ls -laid "Bill's file"
>>> st1a = re.sub("'", """'"'"'""", st1)
>>> st1b = "'" + st1a + "'"  # 'ls -laid "Bill'"'"'s file"' = 'ls -laid "Bill' + "'" + 's file"'
>>> a = subprocess.run( ("echo " + st1b) ,stdout=subprocess.PIPE ,stderr=subprocess.PIPE, shell=True)
>>> print (a.args)
echo 'ls -laid "Bill'"'"'s file"'
>>> a.stdout.decode()[:-1] == st1
True
>>> a.stdout.decode()[:-1] 
'ls -laid "Bill\'s file"'
>>> 
>>> 
>>> st1 = '''ls -laid "../ObjOrPr/Bill's file"'''
>>> st1a = re.sub("'", """'"'"'""", st1)
>>> st1b = "'" + st1a + "'" 
>>> a = subprocess.run( ("echo " + st1b) ,stdout=subprocess.PIPE ,stderr=subprocess.PIPE, shell=True)
>>> a.stdout.decode()[:-1] 
'ls -laid "../ObjOrPr/Bill\'s file"'
>>> a.stdout.decode()[:-1] == st1
True
>>> a = subprocess.run( ("echo " + st1b + ' |ksh') ,stdout=subprocess.PIPE ,stderr=subprocess.PIPE, shell=True)
>>> print (a.args)
echo 'ls -laid "../ObjOrPr/Bill'"'"'s file"' |ksh
>>> print(a.stdout.decode(),end='')
27706761 -rw-r--r--  1 name  staff  0 Nov 18 16:22 ../ObjOrPr/Bill's file
>>>

To execute a shell script Edit

import re
import subprocess

shellScript = '''
cat ../ObjOrPr/"Bill's file" | wc
echo
ls -l ../ObjOrPr/Bill"'s f"ile
echo
ls -laid "../ObjOrPr/Bill's file" | od -c
'''
# Substiute 'Æ' for "'". This keeps the length the same and
# makes a.args slightly more readable.
sS1 = re.sub("'", 'Æ', shellScript)
sS2 = "'" + sS1 + "'"
pattern = "/Æ/s//'/g" ; p1 = '"' + pattern + '"'
a = subprocess.run( ("echo " + sS2 + ' | sed ' + p1 + ' | ksh') , stdout=subprocess.PIPE, stderr=subprocess.PIPE, shell=True)

print ('''
a.args =
{}
a.returncode = {}
a.stdout =
{}]]]]
a.stderr =
{}]]]]
'''.format(
a.args,
a.returncode ,
a.stdout.decode(),
a.stderr.decode(),
),end='')

a.args =
echo '
cat ../ObjOrPr/"BillÆs file" | wc
echo
ls -l ../ObjOrPr/Bill"Æs f"ile
echo
ls -laid "../ObjOrPr/BillÆs file" | od -c
' | sed "/Æ/s//'/g" | ksh
a.returncode = 0
a.stdout =
       0       0       0

-rw-r--r--  1 _____Bill______  staff  0 Nov 18 16:22 ../ObjOrPr/Bill's file

0000000    2   7   7   0   6   7   6   1       -   r   w   -   r   -   -
0000020    r   -   -           1       _   _   _   _   _   B   i   l   l
0000040    _   _   _   _   _   _           s   t   a   f   f           0
0000060        N   o   v       1   8       1   6   :   2   2       .   .
0000100    /   O   b   j   O   r   P   r   /   B   i   l   l   '   s
0000120    f   i   l   e  \n
0000125
]]]]
a.stderr =
]]]]

Assignments Edit

Under "Sending data to a pipe" above:

Modify the python script so that data sent to the pipe does not contain blank lines or comments.

Build the Unix command in different ways such as the following and check the results.

a = subprocess.run( "perl -e " + pS1 ,stdout=subprocess.PIPE ,stderr=subprocess.PIPE, shell=True)
a = subprocess.run( ("perl", '-e', perlScript) ,stdout=subprocess.PIPE ,stderr=subprocess.PIPE, )
a = subprocess.run( ('./pecho.py', '-n', perlScript, '|', 'perl') ,stdout=subprocess.PIPE ,stderr=subprocess.PIPE, )

Try the following Unix commands and verify that a.stdout.decode() matches perlScript exactly:

a = subprocess.run( "./pecho.py -n " + pS1 ,stdout=subprocess.PIPE ,stderr=subprocess.PIPE, shell=True)
a = subprocess.run( ("./pecho.py", "-n", perlScript ) ,stdout=subprocess.PIPE ,stderr=subprocess.PIPE, )

References Edit

1. Python's documentation:

Using the subprocess Module

2. Python's methods:

sys.argv

3. Python's built-in functions:

print()