[issue1943] improved allocation of PyUnicode objects

[Marc-Andre Lemburg]

Marc-Andre Lemburg added the comment:

Agreed, those optimizations do make the implementation more complicated.
It's also not clear whether they would really be worth it.

#1629305 only provided speedups for the case where you write s += 'abc'.
The usual idiom for this is to use a list and then concatenate in one
go. If you want a really fast approach, you'd use cStringIO or perhaps
the bufferarray. I don't think that optimizing for just one particular
use case warrants making the code more complicated or changing the C
interface radically.

In your case, I think that closing the door for being able to easily
extend the type implement at the C is the main argument against it. The
speedups are only marginal and can also be achieved (to some extent) by
tuning the existing implementation's parameters.

__________________________________
Tracker <report at bugs.python.org>
<http://bugs.python.org/issue1943>
__________________________________