BUG : fix _reshape_2D bug with [(n, 1), ..] input #3284
Conversation
cbook._reshape_2D the input can be _at most_ 2D, however if it is passed a list of (n, 1) shaped ndarray this condition will be violated and it fails to work (which is less than great, but fine from a documentation stand point). However, boxplot _used_ to work with a list of such ndarrays (and users might expect to call a (n, 1) shaped array a 1D array). The added `ravel` makes sure that even if we get a list of any dimensional ndarrays in, the output will be a list of (n, ) shaped arrays. Closes matplotlib#3220
|
I will try to pull tests together for @phobson Comments? |
|
Note to self... this function needs serious commenting. Furthermore this function (as well as others elsewhere) will need to be updated to properly account for len()-able data types that don't implement As far as I can tell, this makes sense. |
|
This is set to merge into master, not 1.4.x. Was that intentional? |
|
nope, product of doing this too late at night. I can make a new PR targeted at 1.4, one of use can do the merged to 1.4 locally and then merge to master (which will auto-close this), or we can cherry-pick. Up to you @WeatherGod |
|
It is just a single commit. I'll cherry-pick it to 1.4.x after merging to master. |
BUG : fix _reshape_2D bug with [(n, 1), ..] input
BUG : fix _reshape_2D bug with [(n, 1), ..] input
|
Done |
BUG : fix _reshape_2D bug with [(n, 1), ..] input
cbook._reshape_2Dthe input can be at most 2D, however if itis passed a list of (n, 1) shaped ndarray this condition will
be violated and it fails to work (which is less than great, but
fine from a documentation stand point). However, boxplot used
to work with a list of such ndarrays (and users might expect to
call a (n, 1) shaped array a 1D array). The added
ravelmakessure that even if we get a list of any dimensional ndarrays in, the
output will be a list of (n, ) shaped arrays.
Closes #3220