Groups


191 of 99+  



julia-users ›
It is getting harder to match function prototypes
3 posts by 2 authors  



David van Leeuwen 	

May 2


Hello, 

I am trying to get GaussianMixtuers.jl to pass the tests in julia-0.5.0-dev+3476 .  I am more and more often running into problems related to not being able to find any function to dispatch to, it seems.  This time, I am stuck with the error

MethodError: no method matching xμTΛxμ!(::Array{Float64,2}, ::Array{Float64,2}, ::Array{Float64,1}, ::UpperTriangular{Float64,Array{Float64,2}})
Closest candidates are: 
  xμTΛxμ!(::Array{T,2}, ::Array{T,2}, ::Array{T,1}, ::UpperTriangular{T,S<:AbstractArray{T,2}})



The badly named `xμTΛxμ!()` is a function in GaussianMixtures.   In this case, I can't really see which of the arguments does't match to the closest candidate.  The type `T` in the closest candidate is not specified in the function prototype, so I guess that implicitly the four `T`s in the closest candidate are, in fact, implicitly not the same type.  But even if they are implicitly the same type, the call clearly is with `T==Float64` for the five `T`s in the closest candidate.  The prototype is





function xμTΛxμ!(Δ::Matrix, x::Matrix, μ::Vector, ciΣ::UpperTriangular)



In julia-0.3 and julia-0.4, this problem does not occur.   Does anyone know what might be going on?



Thanks, 



---david

 


Stefan Karpinski 	

May 2


That's strange – that method appears to match what is being looked for. Have you defined your own Matrix, Vector or UpperTriangular types that might be disjoint from the ones provided by Julia??
- show quoted text -
 


David van Leeuwen 	

May 2


Not that I know.  I do have some logic to sort out UpperTriangular for versions < v"0.4.0-dev", but this does not run for julia-0.5. 

It seems though, that this particular error only occurs in testing with 2 subprocesses for parallel computation.  I've made sure that the version of the code in ~/.julia/v0.5/GaussianMixtures is the same as the codebase I am developing in.  

Perhaps the processes spawned by `addprocs()` run a different version of Julia than what I am using for testing?  Could any of these types have changed identity in such version mismatch?  I have /usr/local/bin/julia pointing to julia-0.4 . 
- show quoted text -