[erlang-questions] code_change for gen_fsm: how to actually handle a state change

[Daniel Dormont]

Thanks, that makes sense.

Here's another crazy idea, that I don't imagine is possible: can I
give code_change some kind of return value that says to the system "go
ahead and terminate this process, but call terminate/3 with the _old_
version of the code"? I know it sounds weird, but that would also be a
solution here.

dan

On Tue, Nov 22, 2011 at 5:35 AM, Richard Carlsson
<carlsson.richard@REDACTED> wrote:
> On 11/22/2011 08:40 AM, Bengt Kleberg wrote:
>>
>> Is it correct to say that the old definition of #state{} is
>> -record(state, {item1, item2}).
>> and that the new is
>> -record(state, {item1, item2, newfield}).
>>
>> I think you will find it difficult to get both these records to work in
>> the same file at the same time (during code change).
>
>
> Possible solution: let code_change call a separate module that transforms
> the old definition to an intermediate format that the gen_fsm module can use
> to initialize the new state. That way, the different record declarations are
> never present in the same module.
>
> code_change(#state{}=State) ->
>    State;  % already the new state
> code_change(OldState) ->
>    init_state(state_converter:old_to_proplist(OldState)).
>
> init_state([{foo, Foo} | Ps], State) ->
>    init_state(Ps, State#state{foo=Foo});
> ...
> init_state([], State) ->
>    State.
>
> A more quick-and-dirty solution is to re-tag the old state so you can use it
> with a different state declaration:
>
> -record(old_state, {item1, item2}).
> -record(state, {item1, item2, newfield}).
>
> code_change(#state{}=State) ->
>    State;  % already the new state
> code_change(OldState) ->
>    update_state(setelement(1, State, 'old_state'), #state{}).
>
> update_state(OldState, State) ->
>    State#state{item1 = OldState#old_state.item1,
>                item2 = OldState#old_state.item2,
>                newfield = 42}.
>
>
>    /Richard
> _______________________________________________
> erlang-questions mailing list
> erlang-questions@REDACTED
> http://erlang.org/mailman/listinfo/erlang-questions
>