Date: Thu, 02 Jul 2009 00:07:53 +0300 From: Giorgos Keramidas <keramida@ceid.upatras.gr> To: Wojciech Puchar <wojtek@wojtek.tensor.gdynia.pl> Cc: Alexandre Vieira <nullpt@gmail.com>, freebsd-questions@freebsd.org Subject: Re: scripting tip needed Message-ID: <87k52saz86.fsf@kobe.laptop> In-Reply-To: <alpine.BSF.2.00.0907012202070.1817@wojtek.tensor.gdynia.pl> (Wojciech Puchar's message of "Wed, 1 Jul 2009 22:02:48 %2B0200 (CEST)") References: <755cb9fc0907011040o28b82cdbjd5760b139f797050@mail.gmail.com> <87tz1wqkmu.fsf@kobe.laptop> <alpine.BSF.2.00.0907012202070.1817@wojtek.tensor.gdynia.pl>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 1 Jul 2009 22:02:48 +0200 (CEST), Wojciech Puchar <wojtek@wojtek.tensor.gdynia.pl> wrote:
>> Using an interactive language like Python you can actually *test* the
>> code as you are writing it. This is a major win most of the time.
>
> could you explain what you mean? You can and you have to test a code on
> any language be it bash, ksh python or C
Yes. I mean that one can directly interact with the interpret in a REPL
prompt, doing stuff like:
>>> import re
>>> devre = re.compile(r'(/dev/\S+)\s+(\S+)\s.*$')
>>> devre
<_sre.SRE_Pattern object at 0x28462780>
>>> devre.match('/dev/ad0s1d 1012974 390512 541426 42% /var')
<_sre.SRE_Match object at 0x28432e78>
>>> devre.match('/dev/ad0s1d 1012974 390512 541426 42% /var').groups()
('/dev/ad0s1d', '1012974')
>>> devre =
>>> re.compile(r'(/dev/\S+)\s+(\S+)\s+(\S+)\s+(\S+)\s+(\S+)\s+(\S+).*$')
>>> devre.match('/dev/ad0s1d 1012974 390512 541426 42% /var').groups()
('/dev/ad0s1d', '1012974', '390512', '541426', '42%', '/var')
See how I am 'refining' the initial regular expression without ever
leaving the Python prompt? That sort of interactivity is entirely lost
when you have to edit a file, save it, switch screen(1) windows or type
^Z to background the editor, run a script, watch it fail and repeat.
Then I can keep testing bits and pieces of code:
>>> from subprocess import Popen, PIPE
>>> pipe = Popen(['df', '-k'], shell=False, stdout=PIPE).stdout
>>> for l in pipe:
... m = devre.match(l)
... if m:
... print "device %s, size %ld KB" % (m.group(1), long(m.group(2)))
...
device /dev/ad0s1a, size 1012974 KB
device /dev/ad0s1d, size 1012974 KB
device /dev/ad0s1e, size 2026030 KB
device /dev/ad0s1f, size 10154158 KB
device /dev/ad0s1g, size 284455590 KB
device /dev/md0, size 19566 KB
>>>
So piping df output to a Python bit of code works! That's nice. Then
once I have a 'rough idea' of how I want the script to work, I can
refactor a bit the repetitive bits:
>>> def devsize(line):
... m = devre.match(line)
... if m:
... return (m.group(1), m.group(2))
...
>>> devsize('/dev/ad0s1d 1012974 390512 541426 42% /var')
('/dev/ad0s1d', '1012974')
So here's a short function to return a nice 2-item tuple with two values
(device name, number of 1 KB blocks). Can we pipe df output through it?
>>> pipe = Popen(['df', '-k'], shell=False, stdout=PIPE).stdout
>>> pipe = Popen(['df', '-k'], shell=False, stdout=PIPE).stdout
>>> map(devsize, pipe.readlines())
[ None, ('/dev/ad0s1a', '1012974'), None, ('/dev/ad0s1d', '1012974'),
('/dev/ad0s1e', '2026030'), ('/dev/ad0s1f', '10154158'),
('/dev/ad0s1g', '284455590'), None, None, None, None, None, None,
None, None, None, None, None, None, None, None,
('/dev/md0', '19566'), None]
>>>
It looks we can do that too, but the tuple list may be more useful if we
trim the null items in the process:
>>> pipe = Popen(['df', '-k'], shell=False, stdout=PIPE).stdout
>>> [t for t in map(devsize, pipe.readlines()) if t]
[ ('/dev/ad0s1a', '1012974'), ('/dev/ad0s1d', '1012974'),
('/dev/ad0s1e', '2026030'), ('/dev/ad0s1f', '10154158'),
('/dev/ad0s1g', '284455590'), ('/dev/md0', '19566') ]
So there it is. A nice structure, supported by the core of the
language, using a readable, easy syntax, and listing all the /dev nodes
of my laptop along with their sizes in KBytes.
The entire thing was built 'piece by piece', in the same Python session,
and I now have not only a 'rough idea' of how the code should work, but
also a working copy of the code in my history.
Note the complete *lack* of care about how to append to a list, how to
create dynamic pairs of devicename-size tuples, how to map all elements
of a list through a function, and more importantly the complete and
utter lack of any sort of '"${[]}"' quoting for variable names, values,
nested expansions, and so on.
That's what I am talking about. Shell scripts are nice, but if we are
not constrained for some reason to use only /bin/sh or ksh, there's no
excuse for wasting hours upon hours to decipher cryptic quoting rules
and exceptional edge-cases of "black quoting magic", just to get a short
job done. Being able to _easily_ use higher level structures than a
plain 'stream of bytes' is nice :)
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?87k52saz86.fsf>