Can't "connect" to events, PyCall can't create weak references to Julia function

[marcusps]

I am trying to "connect" to PyPlot events so I can visually inspect some plots and extract peaks from 1D spectra.

I tried the following

julia> using PyPlot
'import sitecustomize' failed; use -v for traceback

julia> x = Float64[]; y = Float64[]
0-element Array{Float64,1}

julia> function mycall(event)
         append!(x,[event.xdata]); append!(y,[event.ydata])
       end
mycall (generic function with 1 method)

julia> connect("button_press_event", mycall)
ERROR: PyError (PyObject_Call) <type 'exceptions.TypeError'>
TypeError("cannot create weak reference to 'PyCall.jl_Function' object",)
  File "/usr/local/opt/python/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/matplotlib/pyplot.py", line 395, in connect
    return get_current_fig_manager().canvas.mpl_connect(s, func)
  File "/usr/local/opt/python/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/matplotlib/backend_bases.py", line 2195, in mpl_connect
    return self.callbacks.connect(s, func)
  File "/usr/local/opt/python/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/matplotlib/cbook.py", line 308, in connect
    self._func_cid_map[s][func] = cid
  File "/usr/local/opt/python/Frameworks/Python.framework/Versions/2.7/lib/python2.7/weakref.py", line 262, in __setitem__
    self.data[ref(key, self._remove)] = value

 in pyerr_check at /Users/msilva/.julia/PyCall/src/exception.jl:58
 in pycall at /Users/msilva/.julia/PyCall/src/PyCall.jl:85
 in connect at /Users/msilva/.julia/PyPlot/src/PyPlot.jl:235

Looking through the PyPlot code it seems like connect should work in this instance, but I have not been able to find any examples of people using this function, so I am not sure if it is functional.

[ronisbr]

I do really need to connect a signal in fig[:canvas] using fig[:canvas][:mpl_connect]. However, I am getting this same error message. I am trying to get the position of a mouse click over a plot. Is there any way to solve the problem now?

Thanks,
Ronan

[stevengj]

This is due to JuliaPy/PyCall.jl#158, and looks like it requires some implementation work in PyCall to support Python weak references to Julia objects.

[ronisbr]

Ok, thanks @stevengj.

[ronisbr]

@stevengj

Hi! Did something change? I was able to connect an event to a Julia function using the fig[:canvas][:mpl_connect] without problems. The weak reference error does not show anymore.

Here is an example:

using PyCall
pygui(:qt5)
using PyPlot

function test(event)
    println("Click: $(event[:xdata]), $(event[:ydata])")
end

fig = figure()
plot(0:1:10, 0:1:10)
fig[:canvas][:mpl_connect]("button_press_event", test)

Then, after clicking on the figure, I see:

julia> Click: 4.423387096774194, 4.476190476190478
Click: 5.820564516129033, 4.11904761904762
Click: 1.806451612903226, 7.2738095238095255
Click: 3.691532258064516, 7.690476190476192
Click: 6.286290322580645, 7.660714285714288
Click: 7.572580645161291, 5.30952380952381
Click: 5.997983870967743, 2.7202380952380962
Click: 1.5625, 3.7916666666666674
Click: 0.941532258064516, 6.708333333333336

[ronisbr]

Here is a gif of my Control Toolbox package that I was developing. I was really wanting this feature to have an interactive plot:

EDIT: I tested this with both Python 2 and 3 using Qt4, Qt5, and Tk as backend. Everything worked as expected!

[lukasfischer83]

What Julia/PyCall version were you using for that? Did you also test on linux?
I still get the same "weak reference" error in 0.6.2 and after a Pkg.update() on ubuntu.

[ronisbr]

I was using Julia 0.6.2 and latest released version of PyCall in openSUSE linux. I tested it on macOS and Linux. In linux, I performed a test using different versions of Python and QT5. Everything worked. You can see this package here:

https://github.com/ronisbr/ControlToolbox.jl

[lukasfischer83]

Unfortunately in my setup (ubuntu 14.04, tried Python 3.4 and 2.7) it also doesn't work with your example.

rlocus(G)
ERROR: PyError (ccall(@pysym(:PyObject_Call), PyPtr, (PyPtr, PyPtr, PyPtr), o, arg, C_NULL)) <class 'TypeError'>
TypeError("cannot create weak reference to 'PyCall.jlwrap' object",)
  File "/usr/lib/python3/dist-packages/matplotlib/backend_bases.py", line 2324, in mpl_connect
    return self.callbacks.connect(s, func)
  File "/usr/lib/python3/dist-packages/matplotlib/cbook.py", line 495, in connect
    self._func_cid_map[s][func] = cid
  File "/usr/lib/python3.4/weakref.py", line 365, in __setitem__
    self.data[ref(key, self._remove)] = value

Stacktrace:
 [1] pyerr_check at /home/lukas/.julia/v0.6/PyCall/src/exception.jl:58 [inlined]
 [2] pyerr_check at /home/lukas/.julia/v0.6/PyCall/src/exception.jl:62 [inlined]
 [3] macro expansion at /home/lukas/.julia/v0.6/PyCall/src/exception.jl:82 [inlined]
 [4] #_pycall#88(::Array{Any,1}, ::Function, ::PyCall.PyObject, ::String, ::Vararg{Any,N} where N) at /home/lukas/.julia/v0.6/PyCall/src/PyCall.jl:710
 [5] _pycall(::PyCall.PyObject, ::String, ::Vararg{Any,N} where N) at /home/lukas/.julia/v0.6/PyCall/src/PyCall.jl:698
 [6] #pycall#92(::Array{Any,1}, ::Function, ::PyCall.PyObject, ::Type{PyCall.PyAny}, ::String, ::Vararg{Any,N} where N) at /home/lukas/.julia/v0.6/PyCall/src/PyCall.jl:732
 [7] pycall(::PyCall.PyObject, ::Type{PyCall.PyAny}, ::String, ::Vararg{Any,N} where N) at /home/lukas/.julia/v0.6/PyCall/src/PyCall.jl:732
 [8] #call#93(::Array{Any,1}, ::PyCall.PyObject, ::String, ::Vararg{Any,N} where N) at /home/lukas/.julia/v0.6/PyCall/src/PyCall.jl:735
 [9] (::PyCall.PyObject)(::String, ::Vararg{Any,N} where N) at /home/lukas/.julia/v0.6/PyCall/src/PyCall.jl:735
 [10] plot_rlocus(::Array{Array{Complex{Float64},1},1}, ::Array{Float64,1}, ::Array{Complex{Float64},1}, ::Array{Complex{Float64},1}, ::Array{Complex{Float64},1}, ::Array{Complex{Float64},1}) at /home/lukas/.julia/v0.6/ControlToolbox/src/plots/plot_rlocus.jl:45
 [11] #rlocus#22(::Bool, ::Function, ::ControlToolbox.TransferFunction, ::Array{Any,1}) at /home/lukas/.julia/v0.6/ControlToolbox/src/analysis/rlocus.jl:222
 [12] rlocus(::ControlToolbox.TransferFunction) at /home/lukas/.julia/v0.6/ControlToolbox/src/analysis/rlocus.jl:27

[ronisbr]

One question, did you install matplotlib from pip?