Interface weirdness in get_value/1

[filmor]

The following works as expected

% some process
gproc:reg({n, l, some_proc}, [], [{value, value}]).

% some other process
gproc:get_attribute({n, l, some_proc}, value) =:= value.

The following however doesn't:

% some process
gproc:reg({n, l, some_proc}, value).

% some other process
gproc:get_value({n, l, some_proc}) =:= value.

The problem is, that get_value/1 will always look the value up in the current process while get_attribute/2 will do the smarter thing and look up the attribute on a process depending on the type of the registration, in particular for names it will always give you the attribute value for the current name-holder.

Is this intended behaviour or just an oversight? Would a PR changing get_value/1 to similar semantics be accepted?

[uwiger]

It might be weird behavior, but it is in line with the documentation. I will confess that I can't account for how I was thinking at the time, but here's a try:

• get_value/1 is only intended for fetching values corresponding to my own reg objects
• get_value/2 expects me to know which process holds the value, assuming it's an error if this is no longer true. As with get_value/1, it expects the requested value to be there.
• lookup_value/1 does what you are asking for. It exits if you pass it a property, since these are not unique, or if given a name, no such key exists.
• lookup_values/1 will work for both names and properties, returning all matching [{Pid,Value}] entries.

[filmor]

Oh, I guess I forgot about lookup_value/1, sorry about that. Still, I'm not using lookup_attribute and still get the "lookup" behaviour there.

The explanation is fine by me though, from my side this can be closed.