Design thought for callbacks

[Marko Rauhamaa]

Chris Angelico <rosuav at gmail.com>:

> On Sun, Feb 22, 2015 at 7:34 PM, Marko Rauhamaa <marko at pacujo.net> wrote:
>> Refloops are not to be worried about, let alone removed.
>
> Why?

Because the whole point of GC-languages is that you should stop worrying
about memory. Trying to mastermind and micromanage GC in the application
is, pardon my French, an antipattern.

> They force the use of the much slower cycle-detecting GC, rather than
> the quick and efficient CPython refcounter.

Java's Hotspot doesn't bother with refcounters but is much faster than
Python. CPython's refcounters are a historical accident that a Python
application developer shouldn't even be aware of.

> I don't know how other Pythons work, but mark-and-sweep has its own
> costs, and I don't know of any system that's both prompt and able to
> detect refloops.

It's exceedingly difficult (and pointless) to detect cycles in your
object structures. Python is going to have to do a GC occasionally
anyway. Yes, your worst-case response times are going to suffer, but
that's the cost of doing business.

> Helping it along means your program doesn't waste memory. Why such a
> blanket statement?

Because worrying Python programmers with evil spirits (reference loops)
leads to awkward coding practices and takes away one of the main
advantages of Python as a high-level programming language.


Marko