Chapter 161 explained how you can get access to your computer’s “command prompt”. If you do not remember, please go refresh your memory. The chapter also explained that you can start Python programs directly from the command prompt with the command:
$ python <programname>.py
This works as long as your system knows how to find Python and the program resides in the current directory. If the program is not in the current directory, it will still work as long as you specify the complete path to the program in the command.
Suppose you have written a program that asks a user for a filename and maybe for a few extra parameters, then processes the file with the indicated name using the parameters. I then ask you to run that program on all the files in a specific directory. The directory contains over 10,000 files. What will you do?
You might adapt the program so that instead of asking for the name of file and processing that file, it processes all files in a directory, using parameters that are not asked of the user. If the parameters that you have to use differ per file, there has to be a way to know what they are (perhaps you can derive them from the file name), so it should be possible to calculate them. This solves your problem. But then I ask you to run your program on a bunch of different directories. What will you do?
You can adapt your program so that it contains a list of all the directories that you need to process, then work through them one by one. And every time that I ask you to add an extra directory, just change the program. Regardless what I ask you, you can always adapt your program to encompass my requests. Though you might get a bit annoyed that I am asking you to change the program again and again and again.
There is a different way of handling such requests, and that is through
batch processing. All command lines support the running of so-called
“batch files,” which contain a list of commands that you give to the
operating system. Under Windows such files have the extension .bat
,
while under Mac and Linux they can have any name. However, you can
install different command shells which use different conventions –
potentially you could even use a Python shell and use Python itself to
write batch files.
The batch file can contain commands that use a close derivative of your first program, that processes just one file, and call it for every file that you want to process. The problem is, of course, that the program requires user input, and you are not prepared to type inputs every time the batch file calls the program. The solution is to change the program in such a way that it processes command line arguments.
On the command line, you can start a Python program with a list of arguments:
$ python <programname>.py <argument_1> <argument_2> … <argument_n>
The arguments are separated from each other by spaces and can be anything, though if you have an argument that contains spaces, you should enclose it in double quotes. This, of course, immediately raises the question what you should do if an argument contains a double quote, and unfortunately the answer is “that depends on the command shell that you are using.” Most commonly, you either precede the double quote with a backslash, or you write a double double quote.
Of course, just writing the arguments will not let your program process them automatically. You have to extend your program with code that processes command line arguments, which means you have to be able to access those command line parameters directly.
sys.argv
You get access to the command line arguments that your program was
supplied with via a pre-defined list that is available when you import
the sys
module. The list is called sys.argv
. It is a list of
strings, each string being one of the command line arguments that was
supplied to the program.
sys.argv
always contains at least one element, namely the complete
name of the Python file that you are running, including the directories
and (under Windows) the drive letter of the place where the program file
resides. To know how many arguments were supplied (including the
filename), you can, of course, use the len()
function.
When you work from an editor, in general you cannot supply command-line arguments to your Python programs. So if you want to experiment with this, you actually have to test your programs on the command line. That is a bit of a hassle, especially during development time. However, I can tell you how to set up your programs in such a way that handling command line arguments is optional.
When I code Python programs, I prefer working from an editor. There are a few editors that support supplying command line arguments to a program when testing it, but most do not. So I want to develop my programs in such a way that I can build them as if they process command line arguments, but that I can test them directly from an editor, and only need to test whether or not they actually process command line parameters correctly once. I do that as follows:
For every parameter that can be controlled via the command line, I create a global variable. I fill these global variables with default values. In the rest of the program, I use these variables as if they are constants. Only at the very start of the main program I check for the existence of command line arguments, and if I find those, I overwrite the variables with values that are supplied on the command line.
The advantage of this approach is that I can develop my program without using command line arguments. If I want to use different values for the command line arguments for testing, I simply use different values for the variables in which I will store the command line arguments. I can even set up the program in such a way that it will either fill a variable via a command line argument, or will ask the user for the value if it was not supplied on the command line.
Typically, such code looks as follows:
import sys
# 3 variables for holding the command line parameters
inputfile = "input.txt"
outputfile = "output.txt"
shift = 3
# Processing the command line parameters
# (works with 0, 1, 2, or 3 parameters)
if len( sys.argv ) > 1:
inputfile = sys.argv[1]
if len( sys.argv ) > 2:
outputfile = sys.argv[2]
if len( sys.argv ) > 3:
try:
shift = int( sys.argv[3] )
except TypeError:
print( sys.argv[3], "is not an integer." )
sys.exit(1)
In this code, three command line arguments are supported: the first two are strings, and the third is an integer. The third one is immediately converted from a string (which a command line argument always is) to an integer, and the program is aborted if this conversion fails. I could have built in more checks for demonstration, but I assume that at this point in your programming career that does not pose any problems to you.
All three arguments have a default value: the first string has default
value "input.txt"
, the second string has default value "output.txt"
,
and the integer has as default value 3. You are not expected to supply
all three arguments on the command line: if you supply zero, all three
default values will be used; if you supply one, the first string will be
overwritten with the command line argument, while the other two
variables will retain their default value; etcetera.
sys.exit()
In the example code above the program is aborted using sys.exit()
if
an argument is not meeting the requirements set to it. sys.exit()
was
introduced in Chapter
72.
However, I did not explain at that time that sys.exit()
can get a
numerical argument, which you see above. The use of this argument is
that it will be returned to the batch file that is running the program,
and the batch file can respond to it if it was designed to. The
numerical argument is supposed to be an error code. Typically, zero is
given as argument if everything was processed correctly (a program that
ends normally will also return zero), and otherwise another number is
used. As some systems are limited to the values zero to 255 as program
return values, it is convention to limit the sys.exit()
argument to
that range of values too.
argparse
If you want module support for command line processing, Python supplies
a standard module argparse
for that. Frankly, I do not see much use
for such a module, as command line processing is too simple to spend
much time on. However, some Python programs, in particular those that
are meant to enhance the system, support a large and complex variety of
command line arguments, and may benefit from such a module. It is up to
you to study it if you think it may help you.