Repository navigation
Converting numpy float64 of python float to Decimal is not the same (get rounded in numpy case) #1833
Description
Environment
- Pythonnet version: '2.5.1'
- Python version: 3.8.12
- Operating System: Win10
- .NET Runtime: 4.8.4470.0
- numpy version: 1.21.4
Details
- Describe what you were trying to get done.
I had to convert floats to Decimal but somehow the current Double I got out of Decimal when looking at it using ToDouble would give me only the round part of the double. I figured that my floats were not native python but numpy float64 . This issue happens only with numpy floats..
- What commands did you run to trigger this issue? If you can provide a
Minimal, Complete, and Verifiable example
this will help us understand the issue.
import clr
from System import Decimal
import numpy as np
d= 45.5
assert d == np.array([d])[0] # should be fine
assert Decimal(d) == Decimal(np.array([d])[0]) # assertion error
Activity
lostmsu commented on Jun 24, 2022
Python.NET 2.x permits all sorts of conversions in attempts to match the function arguments, some of which permit data loss. Please, try 3.0 previews of Python.NET. I believe 3.0 will tell you your Decimal(np.float64) call has no matching constructor overloads forcing you to explicitly convert np.float64 to either Python float or to System.Double first.
Also, to save time please add info about investigation from your side. E.g. in the example the equality of two decimals fails. What are their values at the time of comparison? Surely you checked before opening the bug.
make sure positions sent to actuators are python float not numpy float
filmor commented on Sep 20, 2022
This is fixed in rc6.