Thomas Robitaille

Have a registry of regions?

5

Would it be helpful to keep track of all region classes? We can do that with a simple meta-class (I could add it)

PR #71 adds to_mask for some of the single regions, but in future we'll need to implement it on Compound regions. This will have varying degrees of difficulty depending on...

visual should validate keys/values

3

Currently, the ``.visual`` property is just a free-form dictionary which allows things such as: ``` region.visual['banana'] = 'apple' ``` We should make sure we validate the keys/values for this dictionary...

Chess-like pattern when using rountrip_coords with GWCS, when using dask

2

I am seeing a very strange issue. The following example tries to reproject a dataset with a GWCS to a slightly different GWCS: ```python import os import asdf from copy...

Use multi-threading instead of multi-processing

40

The main functions in reproject, such as ``reproject_interp``, accept ``parallel=`` and ``block_size=`` arguments which, if used, will leverage dask behind the scenes to split up the data into chunks and...

At the moment, we force cdelt[0] to be negative but this isn't true for e.g. solar data. We should avoid this!

Implement option to use dask to do image co-addition

8

This is an experiment for now but it allows: * N-dimensional reprojection/co-addition * ``combine_function='median'`` (not implemented here but trivial to do) * ``reproject_and_coadd`` to optionally return a dask array (``return_type='dask'``)...

Implement parallelization across broadcasted dimensions

1

This builds on https://github.com/astropy/reproject/pull/374 to add a possible mode where if we are dealing with arrays with extra/broadcasted dimensions, we can split up the input array into chunks along the...

Experiment with Python limited API

2

Unfortunately enabling the limited Python API might not be easy yet because we use Cython and support for the limited API there is still experimental.

Avoid writing out a temporary memmap if the input data can already be recognized as a memmap

1

Currently when in parallel mode, if a user specifies a filename for the input file we load the data with astropy.io.fits, which even if memory mapped is then written out...