EXI Grammar Format in JSON

[danielpeintner]

Hi Youenn,

Thank you very much for the EXI decoder. The code looks very promising!

In the schema-informed case, the EXI processor needs to be informed about the expected content which in your case is done with JSON structures.
e.g., tests\streams\schemas\Weblog_grammars.json

May I ask you to explain a bit the object notation (i.e. the meaning) of "grammar"?
For example what each construct means... e.g., in

{
    "name" : "Weblog",
    "uri" : "",
    "grammar" : {
      "initialChoice" : 0,
     "content" : {
        "choices" : [ 0, 1, 1 ],
        "children" : [ 1, 1, -1 ],
        "next" : [ 1, 1, -1 ]
      },
      "xsiTypeNil" : 0
    }
  }

or in

{
    "name" : "entry",
    "uri" : "",
    "grammar" : {
      "initialChoice" : 0,
      "content" : {
        "choices" : [ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ],
        "children" : [ 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, -1 ],
        "next" : [ 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, -1 ]
      },
      "xsiTypeNil" : 0
    }
  }

Thanks again!

-- Daniel

[youennf]

"grammar" represents a state machine more or less.
A "grammar" has a number of state (the size of children array basically).
IIRC:
"initialChoice" represents the possible transitions when entering the particular element grammar.
"start" represents the start tag grammar.
"content" represents the element content grammar.

"content.choices" represents the number of possible transitions at a given state.
The "content.children" identifies the event. In case of content, we can have either SE(QName) or generic events. SE(QName) are positive integers. The integer allows getting the qname and its corresponding element grammar.
When it is negative, it is a generic event, like EE, CH or wildcard.

The same applies to StartTag parsing with attributes. This is defined in "grammar.start" and "grammar.start.children", the integers refering to attribute qnames this time.

One difference between "grammar.start" and "grammar.content" is that attributes are ordered so the parsing is one-way. On the contrary, for content grammars, there may be repetition.
The handling of this is done through the "content.next" grammar.

You can see an example for instance in Deviations1_grammars.json, the "book" content element grammar may be interesting to study.
I can dig a little bit more in the way "next" is used.