def do_concat (shape1, shape2, iaxis, count=[0]):
from pygeode.axis import XAxis, YAxis, ZAxis, TAxis
from pygeode.var import Var
from pygeode.concat import concat
from var_test import varTest
import numpy as np
# Increment test counter (and generate a unique test name)
count[0] += 1
testname = "concattest%05d"%(count[0])
# Create some test data
np.random.seed(count[0])
array1 = np.random.randn(*shape1)
array2 = np.random.randn(*shape2)
# Get the # of dimensions, and assign a unique axis class for each dim
assert array1.ndim == array2.ndim
ndim = array1.ndim
axis_classes = (XAxis, YAxis, ZAxis, TAxis)[:ndim]
# Construct the first var
axes = [cls(n) for cls,n in zip(axis_classes,array1.shape)]
var1 = Var(axes = axes, values = array1, name = "myvar", atts={'a':1, 'b':2, 'c':3})
# The second var should have the same axes, except for the concatenation one
n1 = array1.shape[iaxis]
n2 = array2.shape[iaxis]
axes[iaxis] = axis_classes[iaxis](np.arange(n1, n1+n2))
var2 = Var(axes = axes, values = array2, name = "myvar", atts={'a':1, 'b':3, 'd':4})
# Try concatenating
var = concat(var1,var2)
# The expected result
expected = np.concatenate ( (array1, array2), iaxis)
# Test this
test = varTest(testname=testname, var=var, values=expected)
# Store this test
globals()[testname] = test
def do_concat (shape1, shape2, iaxis, count=[0]):
from pygeode.axis import XAxis, YAxis, ZAxis, TAxis
from pygeode.var import Var
from pygeode.concat import concat
from var_test import varTest
import numpy as np
# Increment test counter (and generate a unique test name)
count[0] += 1
testname = "concattest%05d"%(count[0])
# Create some test data
np.random.seed(count[0])
array1 = np.random.randn(*shape1)
array2 = np.random.randn(*shape2)
# Get the # of dimensions, and assign a unique axis class for each dim
assert array1.ndim == array2.ndim
ndim = array1.ndim
axis_classes = (XAxis, YAxis, ZAxis, TAxis)[:ndim]
# Construct the first var
axes = [cls(n) for cls,n in zip(axis_classes,array1.shape)]
var1 = Var(axes = axes, values = array1, name = "myvar", atts={'a':1, 'b':2, 'c':3})
# The second var should have the same axes, except for the concatenation one
n1 = array1.shape[iaxis]
n2 = array2.shape[iaxis]
axes[iaxis] = axis_classes[iaxis](np.arange(n1, n1+n2))
var2 = Var(axes = axes, values = array2, name = "myvar", atts={'a':1, 'b':3, 'd':4})
# Try concatenating
var = concat(var1,var2)
# The expected result
expected = np.concatenate ( (array1, array2), iaxis)
# Test this
test = varTest(testname=testname, var=var, values=expected)
# Store this test
globals()[testname] = test
sl = tuple([i] if isinstance(i,int) else i for i in sl)
# Apply the slices one dimension at a time.
# Avoids a 'feature' in numpy slicing when there are advanced integer
# indices. If there are 2 such things, numpy goes into a special
# mode where it zips the arrays to get a list of specific coordinates,
# then picks out those individual elements.
# E.g., you would think that x[ [0,1], [0,1] ] would give you a 2x2
# square array of the first 2 rows and first 2 columns of x, but noooo...
# what it does is give you a 1D array, consisting of elements (0,0) and
# (1,1)!!!
expected = values
for dim in range(naxes):
current_sl = [slice(None)]*dim + [sl[dim]] + [slice(None)]*(naxes-1-dim)
expected = expected[current_sl]
# Things are just about to start getting crazy-go-nuts.
# Pass the var and expected values to 'varTest', which in turn
# dynamically defines a test class (subclass of unittest.TestCase)
# to check the var for consistency.
# We then have to take this test class, and assign it to a variable
# (this has to be done dynamically, since we're looping over many tests).
# 'nosetests' will then find this file, import it, and look for any
# global variables that represent a subclass of unittest.TestCase(?),
# then invoke the corresponding tests.
testname = 'slicetest%05d'%count
globals()[testname] = varTest(testname=testname, var=slicedvar, values=expected)
count += 1
