Amend the changes #80 to return only flat, 1-D columns of longitudes and latitudes from the glass.points sampling functions, even for inputs with extra dimensions. This makes it easier to pass the results to other functions, which might not accept ragged lists of arrays.
If extra dimensions are present in the inputs, the output count is an array of counts with the shape of the extra dimensions. This is convenient, because it can be passed directly e.g. as the size= argument in random sampling functions.
A count array also makes it easy to create a column of labels for the sampled populations of points:
label_col = np.repeat(labels, count.flat)
The change also introduces a broadcast_leading_axes() utility function that broadcasts a varying number leading axes for given arrays.
BREAKING CHANGE: Point sampling functions return flat arrays.
Amend the changes #80 to return only flat, 1-D columns of longitudes and latitudes from the
glass.pointssampling functions, even for inputs with extra dimensions. This makes it easier to pass the results to other functions, which might not accept ragged lists of arrays.If extra dimensions are present in the inputs, the output
countis an array of counts with the shape of the extra dimensions. This is convenient, because it can be passed directly e.g. as thesize=argument in random sampling functions.A
countarray also makes it easy to create a column of labels for the sampled populations of points:The change also introduces a
broadcast_leading_axes()utility function that broadcasts a varying number leading axes for given arrays.BREAKING CHANGE: Point sampling functions return flat arrays.
Fixes: #85