use numpy >=1.20 #70
use numpy >=1.20 #70
Conversation
| with: | ||
| python-version: 3.x | ||
| python-version: "3.10" |
There was a problem hiding this comment.
we cannot use latest Python until we resolve the netcdf4<1.6.1 pin.
| @@ -133,7 +133,7 @@ def from_dataframe(cls, df, output, **kwargs): | |||
| profile[j] = pname | |||
| row_size[j] = len(pfg) | |||
| if s_ind is not None: | |||
| s_ind[j] = np.asscalar(np.argwhere(station[:] == pfg[axes.station].dropna().iloc[0])) | |||
| @@ -72,6 +72,9 @@ def get_geographic_attributes(df, axes=None): | |||
|
|
|||
| if len(set(coords)) == 1: | |||
| geoclass = Point | |||
| # The set is to workaround the fact tht pocean | |||
| # relied in a shapely<2 bug to pass a vector here instead of a point. | |||
| coords = set(coords) | |||
There was a problem hiding this comment.
Not sure if you want to leave this as-is or to re-write it but it was relying on a bad behavior before. One cannot pass 4 coords to a Point class.
There was a problem hiding this comment.
From the Shapely 2.0 release notes:
The Point(..) constructor no longer accepts a sequence of coordinates consisting of more than one coordinate pair (previously, subsequent coordinates were ignored) (#1600).
| @@ -274,7 +274,7 @@ def create_ncvar_from_series(ncd, var_name, dimensions, series, **kwargs): | |||
| elif series.dtype.kind in ['U', 'S'] or series.dtype in [str]: | |||
| # AttributeError: cannot set _FillValue attribute for VLEN or compound variable | |||
| v = ncd.createVariable(var_name, get_dtype(series), dimensions, **kwargs) | |||
| elif series.dtype == np.object: | |||
| elif series.dtype == object: | |||
| @@ -29,12 +29,12 @@ classifiers = | |||
| packages = find: | |||
| install_requires = | |||
| cftime>=1.2.1 | |||
| netcdf4 | |||
| numpy>=1.15 | |||
| netcdf4<1.6.1 | |||
There was a problem hiding this comment.
This pin will hurt us when trying to install pocean in modern envs.
I was unable to fix the failing tests without it. Iris wrote a whole new class to solve that. I was hoping for an easier solution here :-/
|
@kwilcox with the I could not see a clear path ahead to easily fix this in Unidata/netcdf4-python#1192 For now I'm OK with this solution and it kind of helps us already. It would be nice to get someone more familiar with pocean-core to tackle the netcdf4 pin though. |
|
Thanks @ocefpaf! I agree, solving the |
I can look into that now. If I manage to fix this today maybe we can even issue a new release in the afternoon. |
The
np.object, and many other aliases, as deprecated and removed in numpy 1.24. This updates the code and set a new min numpy version.@kwilcox I'd appreciate if we can merge this quick and issue a new release. I can take care of the latter.