[erlang-questions] *current* value?

Paulo Sérgio Almeida <>
Thu Oct 18 00:42:08 CEST 2007


YC wrote:

> Hopefully this follow-up doesn't make me sound like a troll - but I 
> can't help but notice that ets and even Mnesia is recommended before 
> process dictionary (which is very heavily frowned upon) in this thread, 
> and I am not sure if I understand the reason, unless process dictionary 
> is technically inferior or broken.  From FP perspective, neither is 
> pure, so why one better than another?
> There must be a reason for the level of discouragement about a built-in 
> facility.

I think people are exagerating a bit on the discouragement. There are 
bad examples, like using it in general purpose libraries or doing 
intermodule sets and gets. But using it in a small module that contains 
the main loop of the process can result in something readable. This is a 
way to simulate objects, with "instance variables" being in the process 
dictionary. But philosophy apart, the sensible thing to do is:

- for small state, pass it around in a parameter in the process loop, 
e.g. in a record;

- for a large number of *SMALL* entries, or when advanced lookup (like 
select) is needed, use ets;

- for a large state with lots of substructure sharing (e.g. a forest 
with lots of common large sub-trees), use the process dictionary. An 
important difference from ets is that no copying is done: we are only 
referencing process memory, while with ets terms are copied to a 
separate memory area and no subterm sharing is preserved, resulting in 
possibly much larger memory requirements.


More information about the erlang-questions mailing list