Assign types to a Dict returned by PyCall

General Usage
1

I’m attempting to call a Python function from Julia which returns a dictionary with known types. I’d like to efficiently get a Julia dictionary with these correct types.

It seems like there should be a way to provide the expected return types, although I may be misinterpreting this section of the readme:

By default, a PyDict is an Any => Any dictionary (or actually PyAny => PyAny ) that performs runtime type inference, but if your Python dictionary has known, fixed types you can instead use PyDict{K,V} given the key and value types K and V respectively.

julia> using PyCall

julia> d = py"{'foo': 123, 'bar': 456}"
Dict{Any, Any} with 2 entries:
  "bar" => 456
  "foo" => 123

julia> d = py"{'foo': 123, 'bar': 456}" :: Dict{String, Int64}
ERROR: TypeError: in typeassert, expected Dict{String, Int64}, got a value of type Dict{Any, Any}
Stacktrace:
 [1] top-level scope
   @ ~/.julia/packages/PyCall/L0fLP/src/pyeval.jl:232

julia> d = py"{'foo': 123, 'bar': 456}" :: PyDict{String, Int64}
ERROR: TypeError: in typeassert, expected PyDict{String, Int64, isdict} where isdict, got a value of type Dict{Any, Any}
Stacktrace:
 [1] top-level scope
   @ ~/.julia/packages/PyCall/L0fLP/src/pyeval.jl:232

Versions:

PyCall v1.93.0
Julia 1.6.4
2

d = Dict{String, Int}(py"{‘foo’: 123, ‘bar’: 456}")

1 Like
3

Thanks, I was concerned that passing the pycall result to a constructor or convert would cause two copies:

  1. Python dict → Julia Dict{Any, Any}
  2. Julia Dict{Any, Any} → Julia Dict{String, Int64}

Any suggestions on how to confirm whether this is the case?

4

Does it matter? Like do you really care this little overhead on the grand schemes of things like you’re using python?

5

Probably doesn’t matter. Just trying to be efficient where possible and have a better understanding of what the code is doing.

6

For reference,

PyDict{String, Int}(py"{'foo': 123, 'bar': 456}"o)

make no copies at all — it’s just a thin wrapper around the underlying Python dictionary. (Note the py"..."o, where the o suppresses automatic conversion and instead gives a raw Python object.)

3 Likes