fix for Basemap.pcolormesh 'ortho' projection #476
Conversation
| @@ -3426,6 +3426,27 @@ def pcolormesh(self,x,y,data,**kwargs): | |||
| if the dimensions are the same, then the last row and column of data will be ignored. | |||
| """ | |||
| ax, plt = self._ax_plt_from_kw(kwargs) | |||
| # fix for invalid grid points | |||
| if ((np.any(x > 1e20) or np.any (y > 1e20)) and | |||
| len(x.shape) == 2 and len(y.shape) == 2): | |||
There was a problem hiding this comment.
instead of len(x.shape) == 2, do x.ndim == 2.
| @@ -3426,6 +3426,27 @@ def pcolormesh(self,x,y,data,**kwargs): | |||
| if the dimensions are the same, then the last row and column of data will be ignored. | |||
| """ | |||
| ax, plt = self._ax_plt_from_kw(kwargs) | |||
| # fix for invalid grid points | |||
| if ((np.any(x > 1e20) or np.any (y > 1e20)) and | |||
| # fix for invalid grid points | ||
| if ((np.any(x > 1e20) or np.any (y > 1e20)) and | ||
| len(x.shape) == 2 and len(y.shape) == 2): | ||
| if not x.shape == y.shape: |
There was a problem hiding this comment.
or clearer would be if x.shape != y.shape:
| if ((np.any(x > 1e20) or np.any (y > 1e20)) and | ||
| len(x.shape) == 2 and len(y.shape) == 2): | ||
| if not x.shape == y.shape: | ||
| raise Exception('pcolormesh: x and y need same dimension') |
There was a problem hiding this comment.
This should be a ValueError, which is typically used when there are problems with input arguments.
| ) | ||
| # we do not want to overwrite original array | ||
| data = data[:nx-1,:ny-1].copy() | ||
| data[mask] = np.nan |
There was a problem hiding this comment.
I need to do some testing with this. I am pretty sure some numpy rules changed recently regarding boolean indexing using masks that don't exactly match the shape of the array being indexed.
There was a problem hiding this comment.
This is why I make a copy of slice of data with the correct dimensions: data = data[:nx-1,:ny-1].copy()
|
*It seems to be working with current numpy; irrespective, the aim of the
code as written is that the dimensions match.*
*Thanks for all the suggestions. Do I need to do a new pull request you
you will integrate a version with your modifications?*
…On Tue, 27 Aug 2019 at 03:52, Benjamin Root ***@***.***> wrote:
***@***.**** commented on this pull request.
------------------------------
In lib/mpl_toolkits/basemap/__init__.py
<#476 (comment)>:
> + nx,ny = x.shape
+ if nx < data.shape[0] or ny < data.shape[1]:
+ raise Exception('pcolormesh: data dimension needs to be at least that of x and y.')
+ mask = (
+ (x[:-1,:-1] > 1e20) |
+ (x[1:,:-1] > 1e20) |
+ (x[:-1,1:] > 1e20) |
+ (x[1:,1:] > 1e20) |
+ (y[:-1,:-1] > 1e20) |
+ (y[1:,:-1] > 1e20) |
+ (y[:-1,1:] > 1e20) |
+ (y[1:,1:] > 1e20)
+ )
+ # we do not want to overwrite original array
+ data = data[:nx-1,:ny-1].copy()
+ data[mask] = np.nan
I need to do some testing with this. I am pretty sure some numpy rules
changed recently regarding boolean indexing using masks that don't exactly
match the shape of the array being indexed.
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<#476?email_source=notifications&email_token=AAJW2FUM5YLKB6AEVGJ77XLQGQJ7NA5CNFSM4IPGEWB2YY3PNVWWK3TUL52HS4DFWFIHK3DMKJSXC5LFON2FEZLWNFSXPKTDN5WW2ZLOORPWSZGOCCWHNTQ#pullrequestreview-279738062>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AAJW2FWIFGNTKURXGRCY5WTQGQJ7NANCNFSM4IPGEWBQ>
.
|
|
Is it still possible to merge this fix? |
|
Hi @2sn, you do not need to create a new pull request, you can just create a new commit in your |
|
@molinav thanks for your follow-up My apologies, I have since transitioned my code to |
|
No worries @2sn, then I will just merge the pull request as it is and I will apply the review changes afterwards in |
fix for #470 as discussed in comment there