agreed conventions

[mlee03]

Discussion to keep track of how python modules will be imported

[mlee03]

import numpy as np
import xarray as xr
import ctypes as ct
from dataclasses import dataclass

[fmalatino]

Regarding numpy arrays:

from numpy.typing import NDArray

def func(array: NDArray):

Also possible (as I think some of us have used already):

import numpy.typing as npt

def func(array: npt.NDArray):

I am partial to the former as it allows for a more natural use when type hinting and requires less actual typing (not just typing in the hinting sense).

If the need arises to specify the data type of the elements contained within the array aliasing can be done:

import numpy as np
from numpy.typing import NDArray[np.float64] as NDArray64

def func(array: NDArray64):

Thoughts?

[mlee03]

I kinda like the explicitness of npt.NDArray, it makes it clearer that NDArray is not a built-in python typing and it's coming from numpy.. but then I don't have much python experience. What do other people do? 🤣

[fmalatino]

Hey I am good with that too. I think this is one of those PEP things where it is bespoke at this level so long as the type hinting happens at all.

[mlee03]

Ok, we will go with import numpy.typing as npt

[fmalatino]

Python function declarations and argument lists
For Python functions with a number of arguments sufficient to run over the 79 characters to a line maximum I propose the format:

def func(
    arg1: Type1,
    arg2: Type2=None,
    arg3: Optional[Type3]=None,
    ...
) -> ReturnType:

[fmalatino]

Uses of Relative vs Absolute path
This mainly pertains to importing and elevating methods and classes in __init__ files. For a package with a structure like:

package
├── external_to_package
└── package_src
    ├── __init__.py
    ├── module1.py
    ├── module2.py
    └── module3.py

All __init__ files and modules within a package should use the absolute path to an imported module:

# __init__.py
from module1.class1 import Class1Obj, method
from module2.class2 import CONSTANTVAR

# within module3.py which is part of the package using the above __init__.py
from module2.class2 import CONSTANTVAR

All files which will be external to those contained within the main package files, but still within the repository containing said files, and for any file importing the package:

from package import Class1Obj, CONSTANTVAR

[mlee03]

avoid from numpy import *

[rem1776]

Private variables and functions
Variable and function names that should be private to a class should include leading double scores to hide the name from other files and any classes that inherit it.