Re: [Pyparsing] jsonParser.py patch

[thomas_h <th...@go...>]

Paul,

>  Hrm, well, I can see that this works, but I'm not sold.  A jsonObject really
>  *isn't* a choice between a jsonDict and a jsonArray.  It really is a '{'
>  members '}', as per the BNF.  I think the real problem is elsewhere.

Yes, I wasn't very happy with my naming but I'm not very good at
naming anyway and I thought it would do. 'jsonObject' should probably
be reserved for what I named 'jsonDict', and what I named 'jsonObject'
should be called something else.

>  The RFC link you gave was very helpful.  It states that the content of the
>  JSON string can be any JSON value, which you tripped over because you
>  encountered a JSON array when we were parsing for a JSON object.

I read this too, but wasn't sure I could trust my understanding.
Especially after I ran my current "reference implementation" (demjson)
on a file containing just "this is a value", which was dismissed as
illegal.

The RFC defines (if I get it right) 'JSON-text' as the start symbol
and says "A JSON text is a serialized object or array.

      JSON-text = object / array".

That might speak in favor of my approach. I think the BNF (I presume
you are referring to the sidebar of json.org) supports your
understanding as well since there is no dedicated start symbol.
(Actually, I can't find that text again where it says JSON is just a
JSON value ...).

>
>  The solution is not to change the definition of jsonObject.  The solution is
>  that we should have been invoking this:
>
>  jsonValue.parseString(testdata)

Fair enough. It might be overgenerating, though.

>  jsonValue already encompasses checking for all the different JSON value
>  types - not just object and array, but also number, true, false, etc.  Try
>  reverting to the original jsonParser.py, but change this statement instead
>  and see if you get successful parsing.

I'm sure this will work, I'll give it a try.

Keep up,
Thomas