Re: [Jython-dev] PyType publication with better concurrency

[Jeff A. <ja...@fa...>]

PyType really is quite hairy code. It has certainly tested my 
understanding of Java initialisation. I've not changed anything 
materially yet, but produced a version plastered with print statements 
that others might find amusing:

https://bitbucket.org/tournesol/jython-fgtype/commits/094f87e1ddaa5d30b4d1be9fcf3209e0ff6432bd

It produces output like this:

addBuilder(class o.p.c.PyShadowString, class o.p.c.PyShadowString$PyExposer)
--- --- --- ---
fromClass(class o.p.c.PyShadowString, true) -> null
|   addFromClass(class o.p.c.PyShadowString)
|   createType(class o.p.c.PyShadowString)
|   getBuilder(class o.p.c.PyShadowString)
|   Class.forName(class o.p.c.PyShadowString) with static init.
|   fromClass(class o.p.c.PyType, false) -> <type 'type'>
|   PyType.init(class o.p.c.PyShadowString)
|   fromClass(class o.p.c.PyString, false) -> <type 'str'>
|   fromClass(class o.p.c.PyMethodDescr, false) -> <type 'method_descriptor'>
|   fromClass(class o.p.c.PyMethodDescr, false) -> <type 'method_descriptor'>
|   fromClass(class o.p.c.PyMethodDescr, false) -> <type 'method_descriptor'>
|   fromClass(class o.p.c.PyMethodDescr, false) -> <type 'method_descriptor'>
|   fromClass(class o.p.c.PyMethodDescr, false) -> <type 'method_descriptor'>
|   fromClass(class o.p.c.PyMethodDescr, false) -> <type 'method_descriptor'>
|   +-- init'd <type 'shadowstr'> (from class o.p.c.PyShadowString)
|   +-- created <type 'shadowstr'> from class o.p.c.PyShadowString
+-> <type 'shadowstr'> (from class o.p.c.PyShadowString)
--- --- --- ---

There is some odd-but-not-incorrect behaviour amongst the 15000 or so 
lines, including PyArray initialised twice and _codecs$EncodingMap left 
in the bootstrap types at the end (until you look up the right kind of 
codec). Now I understand the logic better, and have a baseline, I'll try 
the ClassValue idea, and exhibit it in the same place.

Jeff