Pyuvdata

Added
- Added support for multiple spectral windows for UVH5, UVFITS, MIR, and MIRIAD files.
- Added the `flex_spw` attribute to the `UVData` class, which can be set to True by using the new `_set_flex_spw` method.
- Added the optional `flex_spw_id_array` attribute, of type=int and shape=(`Nfreqs`,), which indexes individual channels along the frequency axis to `spw_array`.
- Adjustment to digital gain removal from mwa_corr_fits files to account for a division by 64 due to a bit selection shift.
- Options to remove coarse band shape and digital gains from mwa_corr_fits files.
- Support for cotter flags in mwa_corr_fits files.

Changed
- When `flex_spw=True`, the `channel_width` attribute of UVData is expected to be an array of type=float and shape=(`Nfreqs`,). Individual channels are allowed to have different channel widths.
- Changed `freq_array` to be of shape (1, `Nfreqs`) (formerley (`Nspws`, `Nfreqs`)).
- Changed `data_array`, `flag_array`, `nsample_array` to be of shape (`Nblts`, 1, `Nfreqs`, `Npols`) (formerly (`Nblts`, `Nspws`, `Nfreqs`, `Npols`)).
- Changed UVH5 metadata byte conversion method from tobytes() to bytes()

Added
- Added read-only support for MIR files, adding the `Mir` class to `UVData`.
- Added the `MirParser` class, which allows for python access to MIR files.
- A new `check_warnings` method in our test module that behaves like `pytest.warns`
but adds the ability to check for multiple warnings.
- The 'ant_str' option when making selects on UVFlag objects
- A check that the uvws match the antenna positions, as part of the acceptability checking.

Changed
- Updated the numpy requirement to >= 1.18
- Moved `parse_ants` to `utils.py` to allow any UVBased object the potential to use it.

Deprecated
- The `checkWarnings` function in our test module, use the new `check_warnings`
function instead.

Added
- `UVFlag` can now take `PosixPath` objects as input.
- An option to average rather than select a single baseline in `UVData.compress_by_redundancy`,
controlled by the `method` keyword.
- `nsample_array_dtype` keyword for `UVData.read`, which sets the datatype for the nsample_array. Currently only used for mwa_corr_fits files.
- `utils` cython module to speed up some common utility functions.
- Added support for installing the package on Windows.
- 'background_lsts' keyword in `uvdata.read` to calculate lsts in the background while reading. Defaults to True.
- `background` keyword in `uvdata.set_lsts_from_time_array` to allow lst calculation in a background thread.
- `read_data` keyword to read_mwa_corr_fits, allows from metadata only reads of MWA correlator fits files.
- `propagate_flags` keyword for `UVData.frequency_average` which flags averaged samples if any contributing samples were flagged

Changed
- Nants data calculation changed to use numpy functions for small speed up.
- Changed `input` variable to `indata` in UVFlag to avoid shadowing python builtin.
- `select` now also accepts a list of baseline indices for the `bls` parameter.
- `data_array_dtype` keyword for `UVData.read` is now also respected by mwa_corr_fits files. Previously it was only used by uvh5 files.
- The `time_range` parameter on UVCal is no longer required and it is suggested to only set it if the Ntimes axis is length one.
- FHD now supports metadata only reads.
- Updated formatting and added more explicit typing in corr_fits.pyx
- Updated Nants calculations for speed up.
- updated HERA telescope location to match the HERA defined center of array.
- `utils.uvcalibrate` now incorporates many more consistency checks between the uvcal and uvdata object. The new keywords `time_check` and `ant_check` were added to control some of these checks.

Deprecated
- The UVBeam methods `set_cs_params`, `set_efield`, `set_power`, `set_simple`, and `set_phased_array` have been made private. The public methods will be removed in version 2.2.
- The UVCal methods `set_gain`, `set_delay`, `set_unknown_cal_type`, `set_sky`, and `set_redundant` have been made private. The public methods will be removed in version 2.2.
- `utils.uvcalibrate` will error rather than warn if some of the newly added checks do not pass, including if antenna names do not match between the objects, starting in version 2.2. In addition, the `flag_missing` keyword is deprecated and will be removed in version 2.2.

Fixed
- UVFlag.__add__ now properly concatenates all existing data-like parameters of the object, including optional ones.
- A bug in `UVData.downsample_in_time` where the data were not being properly weighted by the nsample array and the nsample_array was not being properly weighted by the integration times.
- A bug in `UVData.downsample_in_time` that lead to duplicated data on the final object if a baseline had varying integration times and some integration times were greater than or equal to the requested minimum integration time.

Added
- New `_corr_fits` C extension for performing the index and conjugation mapping calculation for read_mwa_corr_fits
- `copy` method to UVCal and UVBeam objects
- `chunks` to `UVH5.write_uvh5` and `UVData.write_uvh5` for HDF5 dataset chunking.
- `multidim_index` to `UVH5.read_uvh5` and `UVData.read` for multidimensional slicing into HDF5 datasets

Changed
- Various serialized calculations in uvdata.py, uvbeam.py, and utils.py to be vectorized with numpy
- Miriad interface was re-written from a hand written CPython interface to a Cython interface which dynamically creates the CPython during package setup/build.
- uvfits, calfits and beamfits files should now be read faster due to a simplified handling of fits header objects, especially when the history is very long.
- The UVData methods `set_drift`, `set_phased`, and `set_unknown_phase_type` have been made private. The public methods will be removed in version 2.2.

Changed
- UVData allows conjugation of metadata only objects.
- Handling of strings in UVFlag files has been made more widely compatible.

Added
- Option to provide parameters for RectBivariatespline through interp
- `weights_square_array` (optional) parameter on UVFlag - stores sum of squares of weights when converting to waterfall
- `frequency_average` method on UVData to average data along the frequency axis.

Fixed
- A bug in `select` where baseline-polarization tuples would not be conjugated correctly in `UVData` and `UVFlag`, and select on read for Miriad files.
- `metafits_ppds.fits` files can now be passed to `mwa_corr_fits.read` without throwing an error.
- UVParameters that are array_like and have NaNs are now properly identified as equal if the NaNs are in the same locations and all non-NaN entries are equal.
- A bug in `mwa_corr_fits.read` in filling flag and nsample arrays.
- A bug in `UVData.downsample_in_time` in calculating the number of new blts.

Changed
- All references to Python 2 removed from codebase.

Added
- Routine flagging of MWA coarse band edges and center channels, as well as beginning and end integrations. Only done during `read_mwa_corr_fits`.
- Added a return_weights_square option to `utils.collapse` function, which returns the sum of the squares of the weights.

Removed
- Previously deprecated code marked for removal in version > 1.5:
- reading multiple files with file format specific read functions (e.g. read_mirad). Multi-file reads can only be performed with `uvdata.read`
- `read_metadata` keyword in various specific read functions (e.g. `read_miriad`)
- `metadata_only` keyword in `select` and `get_redundancies`
- `uvdata.miriad.read_miriad_metadata`
- `phase_center` keyword in `uvdata.read_mwa_corr_fits`
- `phase_data` keyword in `uvdata.read_mwa_corr_fits`
- `uvdata.get_antenna_redundancies`
- `uvdata.get_baseline_redundancies`
- `uvdata.uvfits.read_uvfits_metadata`
- `uvdata.uvfits.read_uvfits_data`