problem with isce3.cal.point_target_info.oversample

[kanglcn]

isce3 version 0.13.0, installed from conda-forge

I find when apply oversample to an array, the original array will be changed.

A minimum test case:

from isce3.cal.point_target_info import oversample
import numpy as np

a = np.random.rand(10,10)+1j*np.random.rand(10,10)
a_copy = a.copy()

np.testing.assert_array_equal(a,a_copy) # it passed with no error

a_oversampled = oversample(a,nov=5)

np.testing.assert_array_equal(a,a_copy) # it failed

---------------------------------------------------------------------------
AssertionError                            Traceback (most recent call last)
Cell In[38], line 1
----> 1 np.testing.assert_array_equal(a,a_copy)

    [... skipping hidden 1 frame]

File /lustre/work/client/users/kangl/.conda/envs/gslc_13/lib/python3.11/contextlib.py:81, in ContextDecorator.__call__.<locals>.inner(*args, **kwds)
     78 @wraps(func)
     79 def inner(*args, **kwds):
     80     with self._recreate_cm():
---> 81         return func(*args, **kwds)

File /lustre/work/client/users/kangl/.conda/envs/gslc_13/lib/python3.11/site-packages/numpy/testing/_private/utils.py:862, in assert_array_compare(comparison, x, y, err_msg, verbose, header, precision, equal_nan, equal_inf, strict)
    858         err_msg += '\n' + '\n'.join(remarks)
    859         msg = build_err_msg([ox, oy], err_msg,
    860                             verbose=verbose, header=header,
    861                             names=('x', 'y'), precision=precision)
--> 862         raise AssertionError(msg)
    863 except ValueError:
    864     import traceback

AssertionError: 
Arrays are not equal

Mismatched elements: 99 / 100 (99%)
Max absolute difference: 0.60713429
Max relative difference: 0.64424912
 x: array([[0.662784+0.404032j, 0.666327+0.449343j, 0.673974+0.373827j,
        0.501126+0.900433j, 0.676094+0.160175j, 0.771293+0.304292j,
        0.875867+0.731421j, 0.461425-0.130037j, 0.856009-0.075911j,...
 y: array([[0.662784+0.404032j, 0.642459+0.482853j, 0.632233+0.440759j,
        0.357471+0.9665j  , 0.629393+0.294318j, 0.669114+0.489666j,
        0.613705+0.962022j, 0.47772 +0.040072j, 0.815438+0.271247j,...

I suspect the problem is in

isce3/python/packages/isce3/cal/point_target_info.py

Lines 42 to 47 in cc62efc

	def shift_frequency(z, fx, fy):
	x = np.arange(z.shape[1])
	y = np.arange(z.shape[0])
	z *= np.exp(1j * fx * x)[None,:]
	z *= np.exp(1j * fy * y)[:,None]
	return z

The *= is in-place operator. It shouldn't be used here. But I haven't do any test on it.

[bhawkins]

Thanks for the report! I don't really mind shift_frequency operating in-place, but it should probably return None in order to prevent this sort of confusion. I agree it's surprising that oversample would modify the input array and will try to include a patch in the next release.