--- NickC <[EMAIL PROTECTED]> wrote:
> I'd be very interested to hear your opinion on the 'namespace' module,
> which looks at addressing some of these issues (the Record object, in
> particular). The URL is http://namespace.python-hosting.com, and any
> comments should be directed to the [EMAIL PROTECTED]
> discussion list.
Hi Nick,
The namespace module looks interesting, thanks for the pointer! (I saw your
other message but didn't have a chance to reply immediately.)
I tried out the namespace.Record class. The resulting user code looks nice, but
I have two concerns:
- It requires a different coding style; until it is well established it will
surprise people.
- The runtime penalty is severe.
Attached is a simple adopt_timings.py script. If I run it with Python 2.4.1
under RH WS3, 2.8GHz Xeon, I get:
overhead: 0.01
plain_grouping: 0.27
update_grouping: 0.44
plain_adopt_grouping: 0.68
record_grouping: 10.85
I.e. record_grouping (using namespace.Record) is about 40 times slower than the
manual self.x=x etc. implementation.
My conclusion is that namespace.Record may have merits for specific purposes,
but is impractical as a general-purpose utility like I have in mind.
Note that the attached code includes a new, highly simplified "plain_adopt()"
function, based on the information I got through other messages in this thread.
Thanks to everybody for suggestions!
Cheers,
Ralf
____________________________________________________
Sell on Yahoo! Auctions no fees. Bid on great items.
http://auctions.yahoo.com/import sys, os
class plain_grouping:
def __init__(self, x, y, z):
self.x = x
self.y = y
self.z = z
class update_grouping:
def __init__(self, x, y, z):
self.__dict__.update(locals())
del self.self
def plain_adopt():
frame = sys._getframe(1)
init_locals = frame.f_locals
self = init_locals[frame.f_code.co_varnames[0]]
self.__dict__.update(init_locals)
del self.self
class plain_adopt_grouping:
def __init__(self, x, y, z):
plain_adopt()
try:
from namespace import Record
except ImportError:
Record = None
else:
class record_grouping(Record):
x = None
y = None
z = None
class timer:
def __init__(self):
self.t0 = os.times()
def get(self):
tn = os.times()
return (tn[0]+tn[1]-self.t0[0]-self.t0[1])
def time_overhead(n_repeats):
t = timer()
for i in xrange(n_repeats):
pass
return t.get()
def time(method, n_repeats):
g = method(x=1,y=2,z=3)
assert g.x == 1
assert g.y == 2
assert g.z == 3
t = timer()
for i in xrange(n_repeats):
method(x=1,y=2,z=3)
return t.get()
def time_all(n_repeats=100000):
print "overhead: %.2f" % time_overhead(n_repeats)
print "plain_grouping: %.2f" % time(plain_grouping, n_repeats)
print "update_grouping: %.2f" % time(update_grouping, n_repeats)
print "plain_adopt_grouping: %.2f" % time(plain_adopt_grouping, n_repeats)
if (Record is not None):
print "record_grouping: %.2f" % time(record_grouping, n_repeats)
if (__name__ == "__main__"):
time_all()
--
http://mail.python.org/mailman/listinfo/python-list
