An updated to the JSON distribution schema

[BobbyRBruce]

This distribution reformatting has the following benefits:

• The min and max of a given bin are clearly stated.
• Permits statistics with bins of different sizes or sparce bins.
• The 'numBins' and 'binSize', 'min', and 'max' stats are no explicitly
  (and redundantly) stated. They may be derived easily if needed.

[powerjg]

I like this schema! I think it gives the best flexibility.

However, I think it's a good idea to also allow a more compact format with just numbers as values. Basically a distribution could look like two things:

"hist" : {
  "value": [
    {"min": 0, "max": 1, "count": 7},
    {"min": 1, "max": 2, "count": 2},
    {"min": 2, "max": 3, "count": 4}
  ]
}

OR

"hist" : {
  "value": [7,2,3],
  "binSize": 1
}

I see how my previous implementation of the schema didn't quite work the way I expected. Do you think it's possible to have a secondary more compact representation? I think I could figure out the right json-schema syntax to get it to work if we can agree on a compact representation.

I have one comment below... I think that "value" should be "count" in the bins.

Thoughts?

[BobbyRBruce]

Hey @powerjg , thanks for the comments. I've change value to count, and altered the schema so the "value" of a distribution may be an array of numbers. However, I don't know how to specify via the schema that when an array of numbers is given, the binSize must be specified. Likewise, I don't know how to state binSize value should not be specified if using the more detailed "value" (where the min and max for each bin is specified).

I've also added more fields to the Distribution (overflow, underflow, sum, etc.).