Failure when using pmap with pycall

[garrison]

The following three-line program fails using PyCall 0.7.3 with Julia release-0.3 or master

@everywhere using PyCall
@everywhere @pyimport math
pmap(math.sin, linspace(0, pi, 100))

with a rather long error message, the first few lines being

fatal error on 4: fatal error on 2: fatal error on fatal error on 5: 3: ERROR: type: setfield!: expected Ptr{PyObject_struct}, got Symbol
 in deserialize at serialize.jl:565
 in deserialize at serialize.jl:520
 in handle_deserialize at serialize.jl:351
 in deserialize at serialize.jl:334
 in anonymous at serialize.jl:354
 in ntuple at tuple.jl:30
 in deserialize_tuple at serialize.jl:354

[garrison]

A simple variation of the above code seems to work, however, so perhaps sending a python function directly to pmap is just unsupported?

@everywhere using PyCall
@everywhere @pyimport math
@everywhere f(x) = math.sin(x)
pmap(f, linspace(0, pi, 100))

[stevengj]

If each of the processes imports its own Python math module, it should work; I'm not sure why you are getting the weird error message here — it seems like the second example should be equivalent to the first.

I have to admit that I don't really know how pmap works. @StefanKarpinski, any ideas?

[StefanKarpinski]

No idea. @JeffBezanson?

[garrison]

Closing, as this works nowadays without error. The equivalent code that works today is:

using Distributed
@everywhere using PyCall
@everywhere @pyimport math
pmap(math.sin, range(0, stop=pi, length=100))