Added weak references support to jlwrap types (for #21 and #158)

[amyascwk]

I've managed to get weak references to work (#158), and have mainly tested it directly with the python weakref module and on julia functions:

using PyCall
@pyimport weakref

po = PyObject(x -> x + 1)  #create jl_FunctionType
r = weakref.ref(po)
r()                        #returns (anonymous function)

po = 1                     #remove reference to the PyObject
gc()                       #run gc
r()                        #returns nothing

I've also tested it on connecting julia functions to matplotlib's figure events:

import PyPlot.plt

plt[:plot]()
fig = plt[:gcf]()
onpress(mev) = println("($(mev[:xdata]),$(mev[:ydata]))")
cid = fig[:canvas][:mpl_connect]("button_press_event", onpress)

Left clicking within the axes outputs pairs of data coordinates at the location of the click.

[amyascwk]

There appears to be an occasional segfault when using matplotlib's event connection to julia functions, especially when redrawing. Will try to identify the problem.

[ronisbr]

hi @amyascwk!
Do you have any news about weak references support?

[goretkin]

I second the request for any news, @amyascwk. Thanks

[amyascwk]

I'm sorry, I'm afraid that I'm unable to keep up with the development of Julia and PyCall at the moment. I will still keep the patched code (which worked partially for Julia 0.2) available, in case anyone wishes to continue working on it before I get the chance to revisit this again.