We actually use records and we include a version number in the record
name 'foo-bar:0' that way we can easily match on them, and update them
lazily. Concerning the usage of dicts, we are not sure if that's
actually a good idea given that their representation doesn't seem to
be stable across erlang runtime versions, please correct me if I am
wrong.

Ali

On Mon, Apr 26, 2010 at 9:39 PM, David Weldon <[email protected]> wrote:
> Usually I'm all for records because of pattern matching and
> compile-time checking. IMHO, broken upgrades are reason enough not to
> use them directly for storage. I have yet to design a system where the
> data structures never changed after the initial release. I hate to say
> it but, for me, the answer may be to have functions which map between
> records and orddicts/proplists. I guess this really isn't a
> riak-specific issue, but its interesting to hear with other people
> think about it.
>
> Dave
>
> On Mon, Apr 26, 2010 at 10:18 AM, Eric Cestari <[email protected]> wrote:
>>
>> Le 26 avr. 2010 à 15:10, Sean Cribbs a écrit :
>>
>>> Yes, the primary weakness of records would be upgrading them.  You could do 
>>> it with a map-reduce job, but it would be better in general to use a more 
>>> fluid data structure.
>>>
>> In the defense of records :
>> Records can be pattern matched whereas dicts and proplists can't.
>> Using a record with dict attributes and single attributes for pattern 
>> matching ("state" attribute, or "type" ...) could be interesting.
>>
>> Eric
>>
>> http://twitter.com/cstar
>>
>>
>
> _______________________________________________
> riak-users mailing list
> [email protected]
> http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
>

_______________________________________________
riak-users mailing list
[email protected]
http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com

Reply via email to