homepage

2013年11月8日 Martin v. Löwis <report@bugs.python.org>:
> I'd be +0 on extracting common identifiers. I (...) expect a slight reduction of memory usage, and a tiny reduction in runtime.
Only duplicated Py_IDENTIFIER structures would be removed in memory,
but these structures are very small (3 pointers or something like
that). The identifier strings are already interned, so not duplicated
in memory.
> It's not really worth the effort, but I fail to see that it causes harm.
Initializing an identifier has to decode the literal byte string from
UTF-8, but Python UTF-8 decoder is really fast. I'm not sure that it's
possible to see a difference on the startup time.
> I see no point in reverting cases where this approach is already taken.
I only reverted shared identifiers added a few days ago in issue
#19512. I agree to leave the old code unchanged.
> I don't quite understand Victor's interest in this, either, as there are hundreds of open real bugs that could use his attention.
I tried to explain my motivation on using more identifiers in the issue #19512.