Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.python > #40832

Re: Interesting list() un-optimization

Show key headers only | View raw

On 3/7/2013 3:41 PM, Wolfgang Maier wrote:
>
>>>>> Iterators do not generally have __len__ methods.
>>>>>
>>>>>>>> len(iter(range(10)))
>>>>> Traceback (most recent call last):
>>>>>    File "<stdin>", line 1, in <module>
>>>>> TypeError: object of type 'range_iterator' has no len()
>>>>
>>>> But iterators have a length hint method that are used for some
>>>> optimizations and preallocations, too.
>>>>
>>>>>>> i = iter(range(10))
>>>>>>> i.__length_hint__()
>>>> 10
>>>>
>>>> See http://www.python.org/dev/peps/pep-0424/
>>>
>
> very interesting (hadn't heard of it)! Just checked the PEP,
> then tested list()'s behavior, and it is just as described:
>
> class stupid(list):
> 	def __len__(self):
> 		print ('len() called')
> 		return NotImplemented
> 	
> 	def __length_hint__(self):
> 		print ('hint requested')
> 		l=iter(self).__length_hint__()
> 		print (l)
> 		return l
>
> a=stupid((1,2,3))
> len(d)
> ======>
>    len() called
>
>    Traceback (most recent call last):
>      File "<pyshell#79>", line 1, in <module>
>        len(d)
>    TypeError: an integer is required
>
> list(d)
> ======>
>    len() called
>    hint requested
>    3
>    [1, 2, 3]
>
> so list() first tries to call the iterable's __len__ method. If that raises a
> TypeError it falls back to __length_hint__ .
> What I still don't know is how the listiterator object's __length_hint__ works.
> Why, in this case, does it know that it has a length of 3 ? The PEP does not
> provide any hint how a reasonable hint could be calculated.

'How' depends on the iterator, but when it has an underlying concrete 
iterable of known length, it should be rather trivial as .__next__ will 
explicitly or implicitly use the count remaining in its operation. Part 
of the justification of adding __length_hint__ is that in many cases it 
is so easy. Iterators based on an iterator with a length_hint can just 
pass it along.

The list iterator might work with a C pointer to the next item and a 
countdown count initialized to the list length. The __next__ method 
might be something like this mixed C and Python pseudocode:

   if (countdown--) return *(current--);
   else raise StopIteration

(For non-C coders, postfix -- decrements value *after* retrieving it.)
Then __length_hint__ would just return countdown. Tuples would work the 
same way. Sets and dicts would need current-- elaborated to skip over 
empty hash table slots. Range iterators would add the step to current, 
instead of 1.

-- 
Terry Jan Reedy

Back to comp.lang.python | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

Interesting list() un-optimization Roy Smith <roy@panix.com> - 2013-03-06 22:20 -0500
  Re: Interesting list() un-optimization Dave Angel <davea@davea.name> - 2013-03-06 22:38 -0500
  Re: Interesting list() un-optimization Tim Chase <python.list@tim.thechases.com> - 2013-03-06 21:57 -0600
  Re: Interesting list() un-optimization Kev Dwyer <kevin.p.dwyer@gmail.com> - 2013-03-07 07:31 +0000
  Re: Interesting list() un-optimization Wolfgang Maier <wolfgang.maier@biologie.uni-freiburg.de> - 2013-03-07 11:22 +0000
  Re: Interesting list() un-optimization Ian Kelly <ian.g.kelly@gmail.com> - 2013-03-07 09:00 -0700
  Re: Interesting list() un-optimization Christian Heimes <christian@python.org> - 2013-03-07 17:20 +0100
  Re: Interesting list() un-optimization Ian Kelly <ian.g.kelly@gmail.com> - 2013-03-07 10:31 -0700
  Re: Interesting list() un-optimization Stefan Behnel <stefan_ml@behnel.de> - 2013-03-07 20:19 +0100
  Re: Interesting list() un-optimization Ian Kelly <ian.g.kelly@gmail.com> - 2013-03-07 13:26 -0700
  Re: Interesting list() un-optimization Terry Reedy <tjreedy@udel.edu> - 2013-03-07 15:29 -0500
  Re: Interesting list() un-optimization Terry Reedy <tjreedy@udel.edu> - 2013-03-07 15:34 -0500
  Re: Interesting list() un-optimization Wolfgang Maier <wolfgang.maier@biologie.uni-freiburg.de> - 2013-03-07 20:41 +0000
  Re: Interesting list() un-optimization Terry Reedy <tjreedy@udel.edu> - 2013-03-07 17:53 -0500
  Re: Interesting list() un-optimization Steven D'Aprano <steve+comp.lang.python@pearwood.info> - 2013-03-08 17:59 +0000
    Re: Interesting list() un-optimization Roy Smith <roy@panix.com> - 2013-03-08 13:45 -0500
  Re: Interesting list() un-optimization Roy Smith <roy@panix.com> - 2013-03-10 09:05 -0400
    Re: Interesting list() un-optimization Terry Reedy <tjreedy@udel.edu> - 2013-03-10 17:39 -0400
      Re: Interesting list() un-optimization Roy Smith <roy@panix.com> - 2013-03-10 18:34 -0400
        Re: Interesting list() un-optimization Steven D'Aprano <steve+comp.lang.python@pearwood.info> - 2013-03-10 23:35 +0000
          Re: Interesting list() un-optimization Roy Smith <roy@panix.com> - 2013-03-10 19:50 -0400

csiph-web