Suggesting a new feature - "Inverse Generators"

Bengt Richter bokr at oz.net
Fri Mar 25 23:11:20 CET 2005


On Fri, 25 Mar 2005 12:07:23 -0800, Michael Spencer <mahs at telcopartners.com> wrote:

>Scott David Daniels wrote:
>> Michael Spencer wrote:
>> 
>>> itertools.groupby enables you to do this, you just need to define a 
>>> suitable grouping function, that stores its state:
>> 
>> 
>> Michael, this would make a great Python Cookbook Recipe.
>> 
>OK, will do.  What would you call it?  Something like: "Stateful grouping of 
>iterable items"
>
>[Bengt]:
>> Nice, but I think "record" is a bit opaque semantically.
>> How about group_id or generate_incrementing_unique_id_for_each_group_to_group_by or such?
>> 
>> Regards,
>> Bengt Richter
>
>Agreed, it's an issue.  I think the most natural name is groupby - but that 
>would cause more trouble.  What do you think about 'grouping' ?
>I would use 'generate_incrementing_unique_id_for_each_group_to_group_by', but 
>then people might think I'm trying to outdo Bob Ippolito :-)
>
>[Serge]:
>> I think your example would
>> be more clear for Jordan if you used function attributes:
>> 
>> def record(item):
>>      if len(item) > 20:
>>          record.seq +=1
>>      return record.seq
>> record.seq = 0
>
>That does read better than the mutable default argument hack.  Is this use of 
>function attributes generally encouraged? (I tend to think of func_dict for 
>meta-data, used only outside the function)  Thoughts?
>
Personally, I don't like depending on an externally bound (and rebindable) (usually global)
name as a substitute for "self." You could always use a class to carry state, e.g. (untested)

    class Grouper(object):
        def __init__(self): self.group_id = 0
        def __call__(self, item):
            if len(item) > 20: self.group_id += 1 # advance id to next group
            return self.group_id
    # ...
    grouper = Grouper()
    # ...
    for groupnum, lines in groupby(linesource, grouper):
        print "".join(lines)

Or (guess I better actually try this one ;-)

 >>> linesource = """\
 ... Here is a long line, long line, long line
 ... and this is short
 ... and this is short
 ... Here is a long line, long line, long line
 ... and this is short""".splitlines()
 >>>
 >>> for groupnum, lines in groupby(linesource, type('',(),{'n':0, '__call__':lambda s,i: setattr(s,'n', s.n+(i>20)) or s.n})()):
 ...     print "".join(lines)
 ...
 Here is a long line, long line, long line
 and this is short
 and this is short
 Here is a long line, long line, long line
 and this is short

Regards,
Bengt Richter



More information about the Python-list mailing list