Regression using iterators using PyPy about pyfunctional HOT 2 CLOSED

2012-06-07: Different in custom list casting
(the commit in the last reply (5c7fa43f9bb6) was reverted 6 months later by da784fe683c0)

2012-10-06: PEP0424

2014-06-25: __len__ magic method is invoked when using list comprehensions

As long as __len__ exists, I guess the implementation can call it whenever it chooses.

When this happens with big sequences,
long breaks for computations may occur "randomly"
(actually whenever the implementation chooses to call __len__)
which the iterators optimization a slowdown in some cases.

The rewrite branch also has this problem.
There the optimization will just not work (when the interpreter arbitrarily calls __len__)
but at least not causing a slowdown.

My suggestion is not implementing __len__ at all
and name another method in place of it.
It would help if __len__ wasn't called under the hood here and there
but it seems this is not the case.

Perhaps you can have an attribute set to the length of a sequence when known
and experiment raising some exception when not,
imitating a smarter __length_hint__.
All this keeping aside this other length method
which will have a notice that it may get as computationally expensive as realizing the iterator.

from pyfunctional.

In the end went with the solution of removing __len__, while keeping the len and size methods. Removed here: bd391ae

That should resolve this issue.

from pyfunctional.