GlobalToJSON-embeddedPython-pure#2

This package was triggered by the extra bonus points in the contest for writing the example
in embedded Python only.  Therefore PURE

Based on the previous example it looked like being a simple task. A mistake!
I met a bunch of issues that I had to workaround.
But I have to admit that there might be ready solutions and I just didn't find them.

1.  The first step was to reduce the class to 2 methods [language = python] only
   • generating the JSON object file
   • loading the JSON object file
2. Within the import iris.gref class I was unable to find an equivalent to $DATA()
   • checked with order() method for subscripts  to add 10 to my _d
   • tried with get() to access a value and add to my _d 1 for success
   • so my values were 0, 1, 10, 11   as I'm used to  
3. To scan the global nodes I used method query()  that didn't just return the next node but ALL nodes.
   • Even theoretical nodes. Referring to my model 
   • 
   • I expected to see ^X,^X(1),^X(1,2),^X(1,2,0) ,^X(1,2,1),^X(1,2,2),^X(2), ... and NEVER ever ^X(1,2),
   • Value returned None for this unexpected Zombie. But None could also signal Nullstring.
4. This worked for rather large Globals, but I had my doubts about size and available memory.
   • A test with ^oddDEF (1410829 nodes) ran with no result and no error. So be careful.
   • In addition, access to PPG, and extended references ( ^|"USER"^|global) ar not supported  
5. The next challenge was handling of $LISTBUILD() 
   • I found no trace of $LISTVALID() equivalent or any access to elements with $LIST()
   • so $LIST() constructs are just seen and handled as binary strings.
   • 
   • You see this in the rather unreadable val elements of the generated JSON object.
   • Though it has the advantage that it can be stored as a simple string during the load operation. 

Summary:

• The attempt to use and implement everything  just using embedded Python is a mistake
• Keep access to the core structures of IRIS as a subject for ObjectScript which is grown with it
• But instead, use  the range of Python libraries for those areas that are not covered by IRIS yet 
• The success will lay in a well-balanced mix of both worlds
• It's no surprise that version 1 of embedded Python has limits that were not mentioned yet anywhere
• But it is the power of this community to  learn from each other about such limits  

Video

GitHub