Using "subprocess" without lists. . .?

[Michael Williams]

I'm not sure you replied entirely to the correct post.  Basically I'm  
interested in encoding video with FFMPEG, but there will be variable  
length commands so I'd rather be able to type a single string for the  
"command" as opposed to having to enter it in the form of a list.   
And there is really no way to properly parse because some options  
have values and others are simply flags.

Thanks,
Michael

On May 13, 2007, at 2:17 PM, Steven Howe wrote:

> Michael Williams wrote:
>> Hi All,
>>
>> I've recently seen the "subprocess" module and am rather confused by
>> it's requirements.  Is it not possible to execute an entire string
>> without having to break them up into a list of arguments?  For
>> instance, I'd much rather do the following:
>>
>>
>> subprocess.call("ls -al | grep -i test")
>>
>>
>> . . .than to have to:
>>
>>
>> list = ["ls", "-a", "-l" ]  . . . . . . and so on and so forth.
>> subprocess.call(list. . .)
>>
>>
>> What is the best way to go about executing a massively complex single
>> line command?
>>
>>
>> Thanks,
>> Michael
>>
> How about a hybrid solution?:
> from subprocess import Popen, PIPE
> from string import strip
> (stdout, stderr) = Popen(['ls','-al'], stdout=PIPE,  
> stderr=PIPE).communicate()
> # process spins off. Ideally, you should wait for it to complete,  
> or assign the resulting object of Popen
> # to a variable and inquire if it's completed via the 'poll()'  
> operator or wait on it via the 'wait()'
> # operator.
> # then query the stderr via communicate()[1]; if good, i.e. len 
> (stderr) == 0, inspect the resulting
> # stdout string (via communicate()[0]) for your 'test' result. But  
> I'm being lazy. I assume the call to
> # ls -al will be faster then my next statement.
> res = []
> for line in stdout.strip():
>     if 'test' in line:
>        res.append( line )
>
> Do the work you need done at the shell prompt, do the rest in  
> Python. Also I wonder if, you were looking for the word 'test' in  
> the filename,  glob.glob wouldn't have been a bit more efficient?  
> But if you were looking for a 'test' owner or group then using:
> glob.glob(): get the list of  files, with directory. Example:   
> glob.glob('/bin/*')
> os.stat(): get an object of stats on a file. Example: os.stat('/bin/ 
> ls')
> stat.ST_UID: get just the group id, as an integer from the object  
> returned by os.stat. Example: os.stat('/bin/ls')[stat.ST_UID]
> pwd.getpwgid(): get the string translation of a UID. Example:
> pwd.getpwgid( os.stat('/bin/ls')[stat.ST_UID] )
> stat.ST_GID: get the group id, as an integer from the object  
> returned os.stat. Example:
> os.stat('/bin/ls')[stat.ST_GID]
> grp.getgrgid(): get the string translation of a UID. Example:
> grp.getgrgid( os.stat('/bin/ls')[stat.ST_UID] )
> Now I have a list, which I iterate over, getting the UID and GID,  
> as strings, that I can compare to. I never had to use a subprocess  
> and having to build up a 'wait on done loop'.  Note, grp and pwd, I  
> think, are Python/Unix functions. What one does on Windows I don't  
> know, but I'll bet there are similar functions under Python/ 
> Windows. Don't get me wrong, Subprocess is a fast and damn useful  
> tool. I often use it to verify a program is in my path, before  
> creating a subprocess to run some task python isn't built for (like  
> mencoder functionality).
>
> You see, the issue is: shell, python & shell or python alone. When  
> transitioning from one language to another, say C/C++/Java to  
> Python, it is often easier to use your knowledge of the shell  
> command (bash/Bourne for example) to get things done. But it's only  
> a transitional issue. The need to use the shell is a good  
> demonstrator that either the Interpretor is weak, or that you  
> haven't fully explored it's abilities. With the Python Interpretor  
> at v2.4 going onto v2.5, it's very likely that the desired feature  
> exists; you have but to do some clever research to find it. As to  
> clever research ... I find Google a good place to start.
>
> I had some familiarity with os.stat and glob. But the grp and pwd  
> functions were things I discovered from http://docs.python.org/ 
> modindex.html and about two minutes of searching for 'python uid to  
> name' @ Google.
>
> Also, os.listdir and os.path.join would have worked just as well as  
> glob.glob('dirname/*'). But glob.glob was faster to write. Even a  
> better choice would have been to use the 'os.walk' to iterate over  
> the directory tree.
>
> sph.
>
> -- 
> HEX: 09 F9 11 02 9D 74 E3 5B D8 41 56 C5 63 56 88 C0

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/python-list/attachments/20070513/a8f2c0e8/attachment.html>