I need to convert a python script into ruby. I use for that the gems Pandas and Numpy which make the work quite simple.
For example I have these kind of lines:
# python
# DF is a dataframe from Pandas
DF['VAL'].ewm(span = vDAY).mean()
DF['VOLAT'].rolling(vDAY).std()
so no question asked, I convert like this:
# ruby
df['VAL'].ewm(span: vDAY).mean
df['VOLAT'].rolling(vDAY).std
easy.
But I have a function apply from Pandas which takes a function as first argument and I really don't know how to convert it in ruby.
It's something like that :
# python
import numpy as np
DF['VAL'].rolling(vDAY).apply(lambda x: np.polyfit(range(len(x)), x, 1)[0])
# output=> NaN or Float
I tried to decomposed the lambda like this:
# ruby
polyfit = ->(x) { t = Numpy.polyfit((0...x.size).to_a, x, 1); t[0] }
puts polyfit.call(<insert Array argument>)
#=> I have a satisfying output for my lambda
# but...
df['VAL'].rolling(vDAY).apply(&polyfit)
# output=> `apply': <class 'TypeError'>: must be real number, not NoneType (PyCall::PyError)
# or
df['VAL'].rolling(vDAY).apply{ |x| polyfit.call(x) }
# output=> `apply': <class 'TypeError'>: apply() missing 1 required positional argument: 'func' (PyCall::PyError)
# or
df['VAL'].rolling(vDAY).apply(polyfit)
#output=> `apply': <class 'TypeError'>: must be real number, not NoneType (PyCall::PyError)
# or
df['VAL'].rolling(vDAY).apply(:polyfit)
# output=> `apply': <class 'TypeError'>: 'str' object is not callable (PyCall::PyError)
It's obviously not working. The problem is this "x" argument in the python inline syntax that I really don't know how to get it "the ruby way"
If someone can "translate" this apply function from python syntax to ruby, it would be really nice :)
I just want to point out that I'm a ruby/rails developer and I don't know python professionally speaking.
UPDATE:
Ok, it's a complete misunderstanding of python code for my part: apply needs a function argument as a callable object. So in ruby it's not a lambda but a Proc I need.
So the solution for those who encounter the same problem:
# ruby
polyfit = Proc.new { t = Numpy.polyfit((0...x.size).to_a, x, 1); t[0] }
df['VAL'].rolling(vDAY).apply(polyfit)
The solution is to use an Proc (see "UPDATE" section in the original question)
# ruby
polyfit = Proc.new { t = Numpy.polyfit((0...x.size).to_a, x, 1); t[0] }
df['VAL'].rolling(vDAY).apply(polyfit)