Creating symbols from Python

[ChrisRackauckas]

Is this possible? I would like a user to be able to pass a symbol from Python to a Julia API. Do I instead need to implement a conversion from a string for comparisons? That sounds dangerous.

[stevengj]

Call the Julia Symbol(string) function?

[ChrisRackauckas]

>>> import julia
>>> jul = julia.Julia()

>>> jul.Symbol("x")
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'Julia' object has no attribute 'Symbol'
>>> jul.symbol("x")
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'Julia' object has no attribute 'symbol'

[stevengj]

That's #144. For now you probably need jul.eval("Symbol")("x")

But a separate problem is that it might convert it back to a string when it returns the value to Python, since PyCall by default converts Julia Symbols to Python strings.

[stevengj]

In Julia, you would currently have to do PyCall.pyjlwrap_new(:x) to create a Python wrapper for a Julia Symbol. So maybe jul.eval("PyCall.pyjlwrap_new(:x)")?

(I'm guessing slightly here because I don't use pyjulia myself, so I'm working off of how I remember it working and how PyCall works in Julia.)

[ChrisRackauckas]

>>> jul.eval("a=:x")
'x'
>>> jul.eval("typeof(a)")
<PyCall.jlwrap Symbol>

I can use that to define a jul.symbol("x") function. Thanks.

But a separate problem is that it might convert it back to a string when it returns the value to Python, since PyCall by default converts Julia Symbols to Python strings.

This is only going to be used Python->Julia, so it's safe. I think I got all I need now, other than JuliaPy/PyCall.jl#497 and the tag. It worked out surprisingly smooth.

[iffanh]

In Julia, you would currently have to do PyCall.pyjlwrap_new(:x) to create a Python wrapper for a Julia Symbol. So maybe jul.eval("PyCall.pyjlwrap_new(:x)")?

(I'm guessing slightly here because I don't use pyjulia myself, so I'm working off of how I remember it working and how PyCall works in Julia.)

I know this is an old thread but this works for me, thanks!

new_var = jl.eval(f"PyCall.pyjlwrap_new(:{old_var})")