def test_004_inverse_constructor():
try:
a = t.Inverse()
except TypeError:
a = t.Identity()
b = t.Inverse(a)
assert (f"{b}" == "Transform( Inverse Identity )")
else:
assert (False) # Should have thrown an error
a = t.Identity()
b = t.Inverse(a)
c = t.Inverse(b)
assert (f"{c}" == "Transform( Inverse Inverse Identity )")
def test_007_composition():
a = t.PlusOne_()
b = a.inverse()
try:
c = t.Composition(a, b)
assert False, "composition of a non-list should throw an error"
except:
pass
# Construct an inverse
c = t.Composition([a, b])
d = c.inverse()
assert ( f"{c}" == "Transform( ( (_PlusOne) o (Inverse _PlusOne) ) )"\
and f"{d}" == "Transform( Inverse ( (_PlusOne) o (Inverse _PlusOne) ) )")
# Check that Composition flattens compositions into one list
e = t.Composition([c, c])
assert (
f"{e}" ==
"Transform( ( (_PlusOne) o (Inverse _PlusOne) o (_PlusOne) o (Inverse _PlusOne) ) )"
)
# Check that Composition tracks dimensions correctly. In particular, idim/odim propagates through
# transforms with 0 dim to the first nonzero one
b = t.Identity()
c = t.Composition([b, b])
assert (c.idim == 0 and c.odim == 0)
c = t.Composition([a, b])
assert (c.idim == 1 and c.odim == 1)
c = t.Composition([b, a])
assert (c.idim == 1 and c.odim == 1)
def test_006_inverse_method():
a = t.Identity()
b = a.inverse() # Identity is idempotent -- should get same transform back
assert (f"{b}" == f"{a}" and f"{b}" == "Transform( Identity )")
a = t.PlusOne_()
assert (f"{a}" == "Transform( _PlusOne )")
# PlusOne_ is non-idempotent - should get Inverse
b = a.inverse()
assert (f"{b}" == "Transform( Inverse _PlusOne )")
# Inverting the inverse via method should unwrap the inversion
c = b.inverse()
assert (f"{a}" == f"{c}")
def test_012_resample():
# Since the main engine is in helpers (resmple is a wrapper), this is
# really just a basic functionality test that the wrapper works right.
# Makes use of Scale, which is in Basic -- but since this is a test, not
# the main code, so there's no circular dependence.
# Simple 7x5 array with a single nonzero spot
a = np.zeros([7, 5]) # 7 tall, 5 wide
a[2, 2] = 1
# Check that identity and shape specs work
trans = t.Identity()
b = trans.resample(a, method='nearest')
assert (b.shape == (7, 5))
assert (np.all(b == a))
b = trans.resample(a, shape=[5, 5], method='nearest')
assert (b.shape == (5, 5))
assert (np.all(a[0:5, 0:5] == b))
# Try scaling up by a factor of 2.5 -- this should make a 3x3 square
# of ones
trans = t.Scale(2.5, post=[2, 2], pre=[-2, -2])
b = trans.resample(a, method='neares')
assert (b.shape == (7, 5))
checkval = np.zeros([7, 5])
checkval[1:4, 1:4] = 1
assert (np.all(b == checkval))
# Check that anisotropy works in the correct direction
trans = t.Scale([1, 2.5], post=[2, 2], pre=[-2, -2])
b = trans.resample(a, method='nearest')
assert (b.shape == (7, 5))
checkval = np.zeros([7, 5])
checkval[1:4, 2] = 1
assert (np.all(b == checkval))
def test_013_remap():
# remap is all about scientific coordinates, so we have to gin up some
# FITS headers.
# The test article (a) is a simple asymmetric cross.
a = np.zeros([7, 7])
a[1:5, 3] = 1
a[3, 0:5] = 1
ahdr = {
'SIMPLE': 'T',
'NAXIS': 2,
'NAXIS1': 7,
'NAXIS2': 7,
'CRPIX1': 4,
'CRPIX2': 4,
'CRVAL1': 0,
'CRVAL2': 0,
'CDELT1': 1,
'CDELT2': 1,
'CUNIT1': 'pixel',
'CUNIT2': 'pixel',
'CTYPE1': 'X',
'CTYPE2': 'Y'
}
trans = t.Identity()
b = trans.remap({'data': a, 'header': ahdr}, method='nearest')
assert (np.all(b['data'] == a))
# This tests actual transformation and also broadcast since
# PlusOne_ is 1D and a is 2D
#
# The autoscaling ought to completely undo the plus-one, so this should
# be a no-op except for incrementing CRVAL1.
trans = t.PlusOne_()
b = trans.remap({'data': a, 'header': ahdr}, method='nearest')
assert (np.all(a == b['data']))
assert (b['header']['CRPIX1'] == 4)
assert (b['header']['CRPIX2'] == 4)
assert (b['header']['CRVAL1'] == 1)
assert (b['header']['CRVAL2'] == 0)
assert (b['header']['CDELT1'] == 1)
assert (b['header']['CDELT2'] == 1)
# This again tests autoscaling - scale(3) should expand everything,
# but autoscaling should put it back.
trans = t.Scale(3, dim=2)
b = trans.remap({'data': a, 'header': ahdr}, method='nearest')
assert (np.all(a == b['data']))
assert (b['header']['CRPIX1'] == 4)
assert (b['header']['CRPIX2'] == 4)
assert (b['header']['CRVAL1'] == 0)
assert (b['header']['CRVAL2'] == 0)
print(f"CDELT1 is {b['header']['CDELT1']}")
assert (np.isclose(b['header']['CDELT1'], 3, atol=1e-10))
assert (np.isclose(b['header']['CDELT2'], 3, atol=1e-10))
# Test manual scaling of the output by setting the output_range
# This *should* be a no-op
# Note: output range is to/from the *edge* of the outermost pixel,
# so the span is the full span of the image
trans = t.Identity()
b = trans.remap({
'data': a,
'header': ahdr
},
method='nearest',
output_range=[[-3.5, 3.5], [-3.5, 3.5]])
assert (np.all(a == b['data']))
assert (b['header']['CRPIX1'] == 4)
assert (b['header']['CRPIX2'] == 4)
assert (b['header']['CRVAL1'] == 0)
assert (b['header']['CRVAL2'] == 0)
print(f"CDELT1 is {b['header']['CDELT1']}")
assert (np.isclose(b['header']['CDELT1'], 1, atol=1e-10))
assert (np.isclose(b['header']['CDELT2'], 1, atol=1e-10))
b = trans.remap({
'data': a,
'header': ahdr
},
method='nearest',
output_range=[[-3.5 / 3, 3.5 / 3], [-3.5 / 3, 3.5 / 3]])
expando = np.array([[0, 0, 1, 1, 1, 0, 0], [0, 0, 1, 1, 1, 0, 0],
[1, 1, 1, 1, 1, 1, 1], [1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1], [0, 0, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 0, 0]])
assert (np.all(np.isclose(expando, b['data'], atol=1e-10)))
assert (b['header']['CRPIX1'] == 4)
assert (b['header']['CRPIX2'] == 4)
assert (b['header']['CRVAL1'] == 0)
assert (b['header']['CRVAL2'] == 0)
assert (np.isclose(b['header']['CDELT1'], 1 / 3.0, atol=1e-10))
assert (np.isclose(b['header']['CDELT2'], 1 / 3.0, atol=1e-10))
trans = t.Scale(3, dim=2)
b = trans.remap({
'data': a,
'header': ahdr
},
method='nearest',
output_range=[[-3.5, 3.5], [-3.5, 3.5]])
assert (np.all(np.isclose(expando, b['data'], atol=1e-10)))
assert (np.isclose(b['header']['CDELT1'], 1, atol=1e-10))
assert (np.isclose(b['header']['CDELT2'], 1, atol=1e-10))
def test_003_identity_apply():
a = t.Identity()
b = np.array([1, 2])
c = a.apply(b)
assert (np.all(b == c))
def test_002_identity_constructor():
a = t.Identity()
assert (f"{a}" == "Transform( Identity )")
def test_008_wrap():
a = t.PlusOne_()
b = t.Identity()
c = t.Wrap(b, a)
assert (f"{c}" ==
"Transform( ( (Inverse _PlusOne) o (Identity) o (_PlusOne) ) )")