search

NCAS-CMS / cf-python

A CF-compliant Earth Science data analysis library
http://ncas-cms.github.io/cf-python
MIT License
119 stars 19 forks source link

Most construct data wiped when unary operator taken of field #797

Open sadielbartholomew opened 1 month ago

sadielbartholomew commented 1 month ago

Here's a funky one for the end of the working week! When taking the additive inverse i.e. negative of a field, the additive inverse of the (field) data is correctly taken but it results, as an unintended side effect, in all of the data for the metadata constructs being wiped, somehow (no time to investigate until next week, and I won't speculate quite how this even happens!). Note that an operation of -f shows this bug but -1*f does not, which implies it is likely an issue with the setup/logic of the - operator on a field.

Quick examples:

>>> f = cf.example_fields()
>>> print(f[0].data)  # all good
[[0.007, ..., 0.013]] 1
>>> print(-f[0].data)  # all good
[[-0.007, ..., -0.013]] 1
>>> print(-f[0])  # not all good - see the dim coords
Field: specific_humidity (ncvar%q)
----------------------------------
Data            : specific_humidity(latitude(5), longitude(8)) 1
Cell methods    : area: mean
Dimension coords: latitude(5) = 
                : longitude(8) = 
                : time(1) = 
>>> print(-1*f[0])  # also all good
Field: specific_humidity (ncvar%q)
----------------------------------
Data            : specific_humidity(latitude(5), longitude(8)) 1
Cell methods    : area: mean
Dimension coords: latitude(5) = [-75.0, ..., 75.0] degrees_north
                : longitude(8) = [22.5, ..., 337.5] degrees_east
                : time(1) = [2019-01-01 00:00:00]
>>>
>>> print(-f[1])  # note from this result that other constructs are affected too
Field: air_temperature (ncvar%ta)
---------------------------------
Data            : air_temperature(atmosphere_hybrid_height_coordinate(1), grid_latitude(10), grid_longitude(9)) K
Cell methods    : grid_latitude(10): grid_longitude(9): mean where land (interval: 0.1 degrees) time(1): maximum
Field ancils    : air_temperature standard_error
Dimension coords: atmosphere_hybrid_height_coordinate(1) = 
                : grid_latitude(10) = 
                : grid_longitude(9) = 
                : time(1) = 
Auxiliary coords: latitude
                : longitude
                : long_name=Grid latitude name
Cell measures   : measure:area
Coord references: grid_mapping_name:rotated_latitude_longitude
                : standard_name:atmosphere_hybrid_height_coordinate
Domain ancils   : atmosphere_hybrid_height_coordinate
                : ncvar%b
                : surface_altitude

I noticed this during working through issues in the cf-plot basic test suite, whereby a call to cfp.vect(u=c, v=-g) was failing with the error AttributeError: DimensionCoordinate has no data due to the underlying issue whereby the - operation is taken for the field input as the v argument. Which indicates it is more than a representational issue (i.e. the data is missing from the printed output), but I am yet to explore further.

Environment

Using the main branch:

>>> cf.environment()
Platform: Linux-6.5.13-7-MANJARO-x86_64-with-glibc2.39 
HDF5 library: 1.14.2 
netcdf library: 4.9.2 
udunits2 library: /home/slb93/miniconda3/envs/cf-env-312/lib/libudunits2.so.0 
esmpy/ESMF: 8.4.2 /home/slb93/miniconda3/envs/cf-env-312/lib/python3.12/site-packages/esmpy/__init__.py
Python: 3.12.0 /home/slb93/miniconda3/envs/cf-env-312/bin/python
dask: 2024.7.0 /home/slb93/miniconda3/envs/cf-env-312/lib/python3.12/site-packages/dask/__init__.py
netCDF4: 1.6.5 /home/slb93/miniconda3/envs/cf-env-312/lib/python3.12/site-packages/netCDF4/__init__.py
psutil: 5.9.5 /home/slb93/miniconda3/envs/cf-env-312/lib/python3.12/site-packages/psutil/__init__.py
packaging: 23.2 /home/slb93/miniconda3/envs/cf-env-312/lib/python3.12/site-packages/packaging/__init__.py
numpy: 1.26.4 /home/slb93/miniconda3/envs/cf-env-312/lib/python3.12/site-packages/numpy/__init__.py
scipy: 1.11.3 /home/slb93/miniconda3/envs/cf-env-312/lib/python3.12/site-packages/scipy/__init__.py
matplotlib: 3.8.0 /home/slb93/miniconda3/envs/cf-env-312/lib/python3.12/site-packages/matplotlib/__init__.py
cftime: 1.6.2 /home/slb93/miniconda3/envs/cf-env-312/lib/python3.12/site-packages/cftime/__init__.py
cfunits: 3.3.7 /home/slb93/miniconda3/envs/cf-env-312/lib/python3.12/site-packages/cfunits/__init__.py
cfplot: 3.3.0 /home/slb93/git-repos/cf-plot/cfplot/__init__.py
cfdm: 1.11.1.0 /home/slb93/git-repos/cfdm/cfdm/__init__.py
cf: 3.16.2 /home/slb93/git-repos/cf-python/cf/__init__.py
davidhassell commented 1 month ago

Ouch! It's the case for all unary operator, I see.

sadielbartholomew commented 1 month ago

Ah, good spot! I should have thought to check for the more general case. I will update the title...