erlangweb-users

[Erlangweb-users] implementation of edict

[Ulf W. <ul...@wi...>]

Just thought I'd think out loud about the implementation of edict.
Note that I've just been browsing the code on the web, so this
is just from a quick read-through.

The data representation seems a bit inefficient. For example:

   fset(Key, Value) ->
   	case ets:lookup(?MODULE, self()) of
   		[{_, Dict}] -> ets:insert(?MODULE, {self(), dict:store(Key, Value, Dict)});
   		[]          -> exit(no_dict_attached)
   	end.

This strategy obviously scales very poorly if a large number of keys
is inserted.

An alternative strategy would be:

fset(Key, Value) ->
   case ets:member(?MODULE, self()} of
      true ->
         ets:insert(?MODULE, {{self(),Key}, Value});
      false ->
         exit(no_dict_attached)
   end.

...and so on.

The penalty will be incurred on terminate_state(), which will
look something like this:

terminate_state() ->
   ets:delete(?MODULE, self()),
   ets:match_delete(?MODULE, {self(),'_'},'_'}).

(the ets table being an ordered_set, of course.)


Since it would appear as if this module will be called many times while
building a page, I would think that this ought to give better characteristics.


One more thing - I believe this function to be wrong:

fset(List, Key, Value) ->
	case ets:lookup(?MODULE, self()) of
		[{_, Dict}] -> ets:insert(?MODULE, {self(), dict:store(List, [{Key,
Value}|dict_fetch(List, Dict)], Dict)});
		[]          -> exit(no_dict_attached)
	end.


The dict_fetch/2 function returns 'undefined' if there is no occurence
of the key List.
That would mean that calling e.g. fset("list", "person", "john")
before storing an
initial value for "list", would result in the term {"list",
[{"person","john"}|undefined]}
in the dictionary. It would be better to return an error or initialize
it to [] first.

BR,
Ulf W

Re: [Erlangweb-users] implementation of edict

[Michal S. <mi...@er...>]

Hi,
Thank you for your email. Both patches should be included in the 1.1  
release, which we are working on right now.

Thanks,
Michal

-- 
Michal Slaski
http://www.erlang-consulting.com

On 30 Aug 2008, at 17:19, Ulf Wiger wrote:

> Just thought I'd think out loud about the implementation of edict.
> Note that I've just been browsing the code on the web, so this
> is just from a quick read-through.
>
> The data representation seems a bit inefficient. For example:
>
>    fset(Key, Value) ->
>    	case ets:lookup(?MODULE, self()) of
>    		[{_, Dict}] -> ets:insert(?MODULE, {self(), dict:store(Key,  
> Value, Dict)});
>    		[]          -> exit(no_dict_attached)
>    	end.
>
> This strategy obviously scales very poorly if a large number of keys
> is inserted.
>
> An alternative strategy would be:
>
> fset(Key, Value) ->
>    case ets:member(?MODULE, self()} of
>       true ->
>          ets:insert(?MODULE, {{self(),Key}, Value});
>       false ->
>          exit(no_dict_attached)
>    end.
>
> ...and so on.
>
> The penalty will be incurred on terminate_state(), which will
> look something like this:
>
> terminate_state() ->
>    ets:delete(?MODULE, self()),
>    ets:match_delete(?MODULE, {self(),'_'},'_'}).
>
> (the ets table being an ordered_set, of course.)
>
>
> Since it would appear as if this module will be called many times  
> while
> building a page, I would think that this ought to give better  
> characteristics.
>
>
> One more thing - I believe this function to be wrong:
>
> fset(List, Key, Value) ->
> 	case ets:lookup(?MODULE, self()) of
> 		[{_, Dict}] -> ets:insert(?MODULE, {self(), dict:store(List, [{Key,
> Value}|dict_fetch(List, Dict)], Dict)});
> 		[]          -> exit(no_dict_attached)
> 	end.
>
>
> The dict_fetch/2 function returns 'undefined' if there is no occurence
> of the key List.
> That would mean that calling e.g. fset("list", "person", "john")
> before storing an
> initial value for "list", would result in the term {"list",
> [{"person","john"}|undefined]}
> in the dictionary. It would be better to return an error or initialize
> it to [] first.
>
> BR,
> Ulf W
>
> ---------------------------------------------------------------------- 
> ---
> This SF.Net email is sponsored by the Moblin Your Move Developer's  
> challenge
> Build the coolest Linux based applications with Moblin SDK & win  
> great prizes
> Grand prize is a trip for two to an Open Source event anywhere in  
> the world
> http://moblin-contest.org/redirect.php?banner_id=100&url=/
> _______________________________________________
> Erlangweb-users mailing list
> Erl...@li...
> https://lists.sourceforge.net/lists/listinfo/erlangweb-users