Modifications to datasets.py #141
Conversation
Started type hinting Dartboard. The conditional statements can be simplified by rewriting them as guard clauses. Copied attributes from coords can be rewritten as `@property` to save memory
The numpy type hints are a little long, but Python 3.8 doesn't support type aliasing.
"Beautiful is better than ugly."
The init logic can be simplified by treating the if statement as a guard clause
init function has been simplified as well
| @@ -131,7 +131,7 @@ def __init__(self, cell_size: float, npix: int) -> None: | |||
| self.packed_q_centers_2D = np_fft.fftshift(self.sky_q_centers_2D) | |||
| self.packed_phi_centers_2D = np_fft.fftshift(self.sky_phi_centers_2D) | |||
|
|
|||
| self.q_max = ( | |||
| self.q_max = float( | |||
There was a problem hiding this comment.
Just curious, what's the reason for the float call? Shouldn't this already be as a float type?
There was a problem hiding this comment.
Some of the numpy fft calls end up returning values that the linter can't infer its type from. The float call just makes it explicit for use with mypy later.
| @@ -126,7 +155,7 @@ def to(self, device): | |||
|
|
|||
|
|
|||
| # custom dataset loader | |||
| class UVDataset(Dataset): | |||
| class UVDataset(torch_ud.Dataset): | |||
There was a problem hiding this comment.
I think this class is stale and we'll want to eventually rewrite it, using some of the ideas put forth in #126. But for now it can come along for the ride. If I recall correctly, is it used anywhere else in the codebase?
There was a problem hiding this comment.
You are right -- it is stale. I searched for references in the code base but there were none. I thought I'd keep it there in case I missed it.
__init__functions' logic have been simplified by implementing guard clausesq_edgesandphi_edgesparameterGridder.from_image_propertiesinitializer