just playing around - doesn't work with 3.0 due to lack of pattern binding
(which i think is coming back in 3.1?)
>>> from collections import defaultdict
>>> from itertools import count
>>> def mkdict(): return defaultdict(count().next)
...
>>> converters = defaultdict(mkdict)
>>> def to_float((i, s)):
... try: return float(s)
... except: return converters[i][s]
...
>>> def convert(line):
... return map(to_float, zip(count(), line.split(',')))
...
>>> convert('1,2,red')
[1.0, 2.0, 0]
>>> convert('1,2,red')
[1.0, 2.0, 0]
>>> convert('1,2,blue')
[1.0, 2.0, 1]
>>> convert('1,2,blue,blue')
[1.0, 2.0, 1, 0]
andrew cooke wrote:
> Carl Banks wrote:
>> import collections
>> import itertools
>>
>> def createInitialCluster(fileName):
>> fixedPoints = []
>> # quantization is a dict that assigns sequentially-increasing
>> numbers
>> # to values when reading keys that don't yet exit
>> quantization = defaultdict.collections(itertools.count().next)
>> with open(fileName, 'r') as f:
>> for line in f:
>> dimensions = []
>> for s in line.rstrip('\n').split(","):
>> if isNumeric(s):
>> dimensions.append(float(s))
>> else:
>> dimensions.append(float(quantization[s]))
>> fixedPoints.append(Point(dimensions))
>> return Cluster(fixedPoints)
>
> nice reply (i didn't know defaultdict worked like that - very neat).
>
> two small things i noticed:
>
> 1 - do you need a separate quantization for each column? the code above
> might give, for example, non-contiguous ranges of integers for a
> particular column if a string occurs ("by accident" perhaps) in more than
> one.
>
> 2 - don't bother with isNumeric. just return the cast value or catch the
> exception:
>
> [...]
> try:
> dimensions.append(float(s))
> except:
> dimensions.append(float(quantization[s]))
>
> (not sure float() is needed there either if you're using a recent version
> of python - only reason i can think of is to avoid integer division in
> older versions).
>
> andrew
>
>
> --
> http://mail.python.org/mailman/listinfo/python-list
>
>
--
http://mail.python.org/mailman/listinfo/python-list
