`UnboundLocalError` of `reshape_order` in `FGmaxGrid.read_output`

[peanutfun]

Commit 437db74 introduced a bug in FGmaxGrid.read_output: If point_style is not one of 2, 3, or 4, the variable reshape_order is not set.

geoclaw/src/python/geoclaw/fgmax_tools.py

Lines 384 to 393 in 11479f6

	if point_style in [2,3,4]:
	if indexing == 'xy':
	reshape_order = 'C'
	elif indexing == 'ij':
	reshape_order = 'F'
	else:
	raise InputError("*** indexing must by 'xy' or 'ij'")
	self.indexing = indexing # make available to user

This leads to an error later, where reshape_order is always used, irrespective of the point_style:

geoclaw/src/python/geoclaw/fgmax_tools.py

Lines 479 to 486 in 11479f6

	X = numpy.reshape(d[:,0],fg_shape,order=reshape_order)
	Y = numpy.reshape(d[:,1],fg_shape,order=reshape_order)
	y0 = 0.5*(Y.min() + Y.max()) # mid-latitude for scaling plots
	h = numpy.reshape(d[:,ind_h],fg_shape,order=reshape_order)
	# AMR level used for each fgmax value:
	level = numpy.reshape(d[:,ind_level].astype('int'),fg_shape,
	order=reshape_order)

>       X = numpy.reshape(d[:,0],fg_shape,order=reshape_order)
                                                ^^^^^^^^^^^^^
E       UnboundLocalError: cannot access local variable 'reshape_order' where it is not associated with a value

There should be a default value set for reshape_order to avoid this issue.

[mandli]

@peanutfun good catch, reshape_order is not needed in the case of point_style in [0, 1], but since we always use it, we should just set it. Better would probably to code the numpy.reshape calls so they do not need it so that they have the expected default behavior, but maybe it's enough to take care of this.

Do you want to do a PR for a fix?

[peanutfun]

I think I can do that. Just to clarify: for point styles 0 and 1, the order parameter of numpy.reshape should be the default 'C'?

[mandli]

That is my guess as the ordering in principle should not matter for 1D arrays although there have been weird instances when this was not true in the past.