Sidney Mau

Note that the implementation proposed in the above PR ends up being fairly inefficient because it can't fully leverage nodes for, e.g., projections and filtering. This functionality could be included—basically...

Just noting that I am also running into (after discovering I had to install `python3-devel` to make `Python.h` available on fedora). Not sure if this is due to changes in...

There may need to be some additional handling in case the requested types include `noshear` and psf types but not normal shear types (in which case `len(types) > 1` would...

My bad, didn't catch the existing test. Updated accordingly

There are a few motivations: - It would be convenient for testing the detection & fitting pipeline from metadetect in a simplified case. For example, I want to compare the...

The alternative I was coming up with was to reconstruct the `MEDSifier` to get the detection stage run on the `mbobs` but then I may be missing some of the...

In this case I am running code from `metadetect` (i.e., `metadetect.do_metadetect` and/or using the `Metadetect` class) -- that code runs `get_all_metacal` which runs `get_all`, which is how I arrived here....

I do appreciate that it is bad to break standing behavior -- but if the decision is to keep this as-is I do think it would be good to at...

I think the only change that would be necessary is to extract out the "get_unsheared" bit from https://github.com/esheldon/ngmix/blob/d08b471f4c4d5887df9f9f2551efaf8f2e226150/ngmix/metacal/metacal.py#L181-L185 and rather have the code run each requested type in sequence (rather...

A few more changes: - update `ParquetReader` to use pyarrow as well - implement `ParquetWriter` - add support for reading/writing pyarrow tensors (e.g., for covariance matrices) -- this is a...