Re: [core] SenML JSON syntax (with multiple base objects)

[Alexander Pelov <[e-mail]>]

Dear all,

I’m not sure to whether this has been discussed, so please forgive my ignorance on the topic. At one point, we considered the use of Deterministic Multimaps in CoOL.

A deterministic multimap, is a map, which can have duplicated key entries, and where the order of the elements is preserved. It seems to me that this structure fits perfectly the need of SenML.

Unfortunately, there is no native JSON/CBOR support for deterministic multimaps (DMMs). However, there is an extremely efficient way of representing them with arrays. In fact, it is as efficient, as the native map representation.

The representation is the following: 
[element1, element2, …, elementN]

Where, for every element:
- if it is NOT an array, it is the key, and the following element is the corresponding value.
- if it IS an array, its first element provides the key, and the rest of the array is the value, which is a DMM (processed recursively).

This allows for an efficient encoding of DMM, where the keys can be of any type EXCEPT arrays. In addition, it is fully compatible with JSON and CBOR. In CBOR we could have the luxury of allocating a tag for DMMs.

-------------------------------------------------------------------------------------------------------------
Two simple examples:

1)
[key1, value1, key2, value2, …, keyN, valueN]
Is interpreted as a DMM with the following structure:
{
  key1 : value1,
  key2 : value2, 
  …
  keyN : valueN
} // note that key1 can be the same as key2, …, keyN, and the order is preserved

2)
[key1, value1, key2, value2, [key3, keyA1, valueA1, keyA2, valueA2, [keyA3, keyB1, valueB1, keyB2, valueB2]], …, keyN, valueN]
Is interpreted as a DMM with the following structure:
{
  key1 : value1,
  key2 : value2, 
  key3 : {
    keyA1 : valueA1,
    keyA2 : valueA2,
    keyA3 : {
      keyB1, valueB1,
      keyB2, valueB2
    }
  },
  …
  keyN : valueN
}
-------------------------------------------------------------------------------------------------------------

Best,
Alexander




> Le 27 oct. 2015 à 09:12, Carsten Bormann <[e-mail]> a écrit :
> 
> Maps (JSON objects) are not fully compatible with streaming processing — you don’t know in which order the map elements will arrive.
> So this is OK if the assumption is that each element of the top level array will be fully ingested before acting on it.
> (Yes, we already have this problem with the other metadata in the measurements objects.)
> 
> I still like the two-level structure*) that we have been discussing, as it allows processing the elements of the measurements arrays knowing there is no serial dependency between their entries.  Might *look* more complicated, but actually is simpler, because the code can focus positionally on either processing base information or on processing measurements.
> 
> Grüße, Carsten
> 
> *) as in [+(base, [+measurement])]
> On 27 October 2015 at 07:06:56, Cullen Jennings ([e-mail] <mailto:[e-mail]>) wrote:
> 
>> If a base name is found in the the object, it applies to all the values in that object 
> _______________________________________________
> core mailing list
> [e-mail] <mailto:[e-mail]>
> https://www.ietf.org/mailman/listinfo/core <https://www.ietf.org/mailman/listinfo/core>