RSLC dataset and attribute issues

[nemo794]

It was noticed by ISRO that some attribute dtypes in GCOV sample products did not correspond to their respective dataset dtypes for GCOV; #213 addresses those reported issues.

As an additional follow-up, I implemented new checks into nisarqa via isce-framework/nisarqa#155 ; this nisarqa PR checks for such issues, and it also checks for empty datasets and invalid string datasets.

Here are the outputs from that QA PR, for a recent RSLC HDF5:

"Attribute has dtype float64, which does not match its dataset's dtype of float32. Path: science/LSAR/RSLC/metadata/geolocationGrid/alongTrackUnitVectorX -> _FillValue"
"Attribute has dtype float64, which does not match its dataset's dtype of float32. Path: science/LSAR/RSLC/metadata/geolocationGrid/alongTrackUnitVectorX -> valid_max"
"Attribute has dtype float64, which does not match its dataset's dtype of float32. Path: science/LSAR/RSLC/metadata/geolocationGrid/alongTrackUnitVectorX -> valid_min"
"Attribute has dtype float64, which does not match its dataset's dtype of float32. Path: science/LSAR/RSLC/metadata/geolocationGrid/alongTrackUnitVectorY -> _FillValue"
"Attribute has dtype float64, which does not match its dataset's dtype of float32. Path: science/LSAR/RSLC/metadata/geolocationGrid/alongTrackUnitVectorY -> valid_max"
"Attribute has dtype float64, which does not match its dataset's dtype of float32. Path: science/LSAR/RSLC/metadata/geolocationGrid/alongTrackUnitVectorY -> valid_min"
"Attribute has dtype float64, which does not match its dataset's dtype of float32. Path: science/LSAR/RSLC/metadata/geolocationGrid/elevationAngle -> _FillValue"
"Attribute has dtype float64, which does not match its dataset's dtype of float32. Path: science/LSAR/RSLC/metadata/geolocationGrid/elevationAngle -> valid_max"
"Attribute has dtype float64, which does not match its dataset's dtype of float32. Path: science/LSAR/RSLC/metadata/geolocationGrid/elevationAngle -> valid_min"
"Attribute has dtype float64, which does not match its dataset's dtype of float32. Path: science/LSAR/RSLC/metadata/geolocationGrid/incidenceAngle -> _FillValue"
"Attribute has dtype float64, which does not match its dataset's dtype of float32. Path: science/LSAR/RSLC/metadata/geolocationGrid/incidenceAngle -> valid_max"
"Attribute has dtype float64, which does not match its dataset's dtype of float32. Path: science/LSAR/RSLC/metadata/geolocationGrid/incidenceAngle -> valid_min"
"Attribute has dtype float64, which does not match its dataset's dtype of float32. Path: science/LSAR/RSLC/metadata/geolocationGrid/losUnitVectorX -> _FillValue"
"Attribute has dtype float64, which does not match its dataset's dtype of float32. Path: science/LSAR/RSLC/metadata/geolocationGrid/losUnitVectorX -> valid_max"
"Attribute has dtype float64, which does not match its dataset's dtype of float32. Path: science/LSAR/RSLC/metadata/geolocationGrid/losUnitVectorX -> valid_min"
"Attribute has dtype float64, which does not match its dataset's dtype of float32. Path: science/LSAR/RSLC/metadata/geolocationGrid/losUnitVectorY -> _FillValue"
"Attribute has dtype float64, which does not match its dataset's dtype of float32. Path: science/LSAR/RSLC/metadata/geolocationGrid/losUnitVectorY -> valid_max"
"Attribute has dtype float64, which does not match its dataset's dtype of float32. Path: science/LSAR/RSLC/metadata/geolocationGrid/losUnitVectorY -> valid_min"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyA/HH -> max_imag_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyA/HH -> max_real_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyA/HH -> mean_imag_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyA/HH -> mean_real_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyA/HH -> min_imag_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyA/HH -> min_real_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyA/HH -> sample_stddev_imag"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyA/HH -> sample_stddev_real"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyA/HV -> max_imag_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyA/HV -> max_real_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyA/HV -> mean_imag_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyA/HV -> mean_real_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyA/HV -> min_imag_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyA/HV -> min_real_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyA/HV -> sample_stddev_imag"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyA/HV -> sample_stddev_real"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyB/HH -> max_imag_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyB/HH -> max_real_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyB/HH -> mean_imag_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyB/HH -> mean_real_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyB/HH -> min_imag_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyB/HH -> min_real_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyB/HH -> sample_stddev_imag"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyB/HH -> sample_stddev_real"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyB/HV -> max_imag_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyB/HV -> max_real_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyB/HV -> mean_imag_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyB/HV -> mean_real_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyB/HV -> min_imag_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyB/HV -> min_real_value"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyB/HV -> sample_stddev_imag"
"Attribute has dtype float64, which does not match the half-precision of its dataset's dtype of complex64. Path: science/LSAR/RSLC/swaths/frequencyB/HV -> sample_stddev_real"

[hfattahi]

@nemo794 Thanks for looking at more details of the dataset attributes. However, many of the issues raised seem like false positives. Specifically, the flags regarding the min, max, and standard deviation of the real and imaginary components of the complex datasets are naturally real-valued.
For other complains about using float64 for _FillValue and min/max of float32 data, we can probably discuss!

[nemo794]

Thanks @hfattahi . What these messages are commenting on is the precision of the min/max/mean/std. (Not float vs. complex.)

If we follow significant figures, then if the main dataset is e.g. float32, then the metric computed on that should similarly be float32. For complex64 datasets, since the real component is float32 and the imag component is float32, then the metrics for each of those components would also be float32.

That said, perhaps there are subtleties in how the metrics are computed by the various ISCE3 SASes. For example, if the image arrays are actually e.g. float64 (and then simply stored in the HDF5 as float32), but the SAS computes the metrics on the float64 version of the array, then we could rationalize why the metrics are allowed to be the higher-precision of float64 in the HDF5.