RF+BF: Add tolerances and data types for nib-diff, remove hypothesis dependency

.travis.yml

@@ -13,7 +13,7 @@ cache:
     - $HOME/.cache/pip
 env:
     global:
-        - DEPENDS="six numpy scipy matplotlib h5py pillow pydicom hypothesis"
+        - DEPENDS="six numpy scipy matplotlib h5py pillow pydicom"
         - OPTIONAL_DEPENDS=""
         - INSTALL_TYPE="setup"
         - EXTRA_WHEELS="https://5cf40426d9f06eb7461d-6fe47d9331aba7cd62fc36c7196769e4.ssl.cf2.rackcdn.com"
@@ -97,7 +97,7 @@ before_install:
     - source venv/bin/activate
     - python --version # just to check
     - pip install -U pip wheel  # needed at one point
-    - retry pip install nose flake8 mock hypothesis  # always
+    - retry pip install nose flake8 mock  # always
     - pip install $EXTRA_PIP_FLAGS $DEPENDS $OPTIONAL_DEPENDS
     - if [ "${COVERAGE}" == "1" ]; then
       pip install coverage;

appveyor.yml

@@ -22,7 +22,7 @@ install:
   - SET PATH=%PYTHON%;%PYTHON%\Scripts;%PATH%
 
   # Install the dependencies of the project.
-  - pip install numpy scipy matplotlib nose h5py mock hypothesis pydicom
+  - pip install numpy scipy matplotlib nose h5py mock pydicom
   - pip install .
   - SET NIBABEL_DATA_DIR=%CD%\nibabel-data
 

dev-requirements.txt

@@ -2,4 +2,3 @@
 -r requirements.txt
 nose
 mock
-hypothesis

nibabel/cmdline/diff.py

@@ -45,25 +45,41 @@ def get_opt_parser():
 
 
 def are_values_different(*values):
-    """Generically compares values, returns true if different"""
-    value0 = values[0]
-    values = values[1:]  # to ensure that the first value isn't compared with itself
+    """Generically compare values, return True if different
 
-    for value in values:
-        try:  # we sometimes don't want NaN values
-            if np.any(np.isnan(value0)) and np.any(np.isnan(value)):  # if they're both NaN
-                break
-            elif np.any(np.isnan(value0)) or np.any(np.isnan(value)):  # if only 1 is NaN
-                return True
+    Note that comparison is targetting reporting of comparison of the headers
+    so has following specifics:
+    - even a difference in data types is considered a difference, i.e. 1 != 1.0
+    - NaNs are considered to be the "same", although generally NaN != NaN  
+    """
+    value0 = values[0]
 
-        except TypeError:
-            pass
+    # to not recompute over again
+    if isinstance(value0, np.ndarray):
+        value0_nans = np.isnan(value0)
+        if not np.any(value0_nans):
+            value0_nans = None
 
+    for value in values[1:]:
         if type(value0) != type(value):  # if types are different, then we consider them different
             return True
         elif isinstance(value0, np.ndarray):
-            return np.any(value0 != value)
-
+            if value0.dtype != value.dtype or \
+               value0.shape != value.shape:
+                return True
+            # there might be NaNs and they need special treatment
+            if value0_nans is not None:
+                value_nans = np.isnan(value)
+                if np.any(value0_nans != value_nans):
+                    return True
+                if np.any(value0[np.logical_not(value0_nans)]
+                          != value[np.logical_not(value0_nans)]):
+                    return True
+            elif np.any(value0 != value):
+                return True
+        elif value0 is np.NaN:
+            if value is not np.NaN:
+                return True
         elif value0 != value:
             return True
 

nibabel/tests/test_diff.py

@@ -7,9 +7,6 @@
 from os.path import (dirname, join as pjoin, abspath)
 import numpy as np
 
-from hypothesis import given
-import hypothesis.strategies as st
-
 
 DATA_PATH = abspath(pjoin(dirname(__file__), 'data'))
 
@@ -18,51 +15,59 @@
 # TODO: MAJOR TO DO IS TO FIGURE OUT HOW TO USE HYPOTHESIS FOR LONGER LIST LENGTHS WHILE STILL CONTROLLING FOR OUTCOMES
 
 
-@given(st.data())
-def test_diff_values_int(data):
-    x = data.draw(st.integers(), label='x')
-    y = data.draw(st.integers(min_value=x + 1), label='x+1')
-    z = data.draw(st.integers(max_value=x - 1), label='x-1')
-
-    assert not are_values_different(x, x)
-    assert are_values_different(x, y)
-    assert are_values_different(x, z)
-    assert are_values_different(y, z)
-
-
-@given(st.data())
-def test_diff_values_float(data):
-    x = data.draw(st.just(0), label='x')
-    y = data.draw(st.floats(min_value=1e8), label='y')
-    z = data.draw(st.floats(max_value=-1e8), label='z')
+def test_diff_values_int():
+    long = 10**30
+    assert not are_values_different(0, 0)
+    assert not are_values_different(1, 1)
+    assert not are_values_different(long, long)
+    assert are_values_different(0, 1)
+    assert are_values_different(1, 2)
+    assert are_values_different(1, long)
 
-    assert not are_values_different(x, x)
-    assert are_values_different(x, y)
-    assert are_values_different(x, z)
-    assert are_values_different(y, z)
 
+def test_diff_values_float():
+    assert not are_values_different(0., 0.)
+    assert not are_values_different(0., 0., 0.) # can take more
+    assert not are_values_different(1.1, 1.1)
+    assert are_values_different(0., 1.1)
+    assert are_values_different(0., 0, 1.1)
+    assert are_values_different(1., 2.)
 
-@given(st.data())
-def test_diff_values_mixed(data):
-    type_float = data.draw(st.floats(), label='float')
-    type_int = data.draw(st.integers(), label='int')
-    type_none = data.draw(st.none(), label='none')
 
-    assert are_values_different(type_float, type_int)
-    assert are_values_different(type_float, type_none)
-    assert are_values_different(type_int, type_none)
+def test_diff_values_mixed():
+    assert are_values_different(1.0, 1)
+    assert are_values_different(1.0, "1")
+    assert are_values_different(1, "1")
+    assert are_values_different(1, None)
     assert are_values_different(np.ndarray([0]), 'hey')
-    assert not are_values_different(type_none, type_none)
-
-
-@given(st.data())
-def test_diff_values_array(data):
-    a = data.draw(st.lists(elements=st.integers(min_value=0), min_size=1))
-    b = data.draw(st.lists(elements=st.integers(max_value=-1), min_size=1))
-    c = data.draw(st.lists(elements=st.floats(min_value=1e8), min_size=1))
-    d = data.draw(st.lists(elements=st.floats(max_value=-1e8), min_size=1))
-    # TODO: Figure out a way to include 0 in lists (arrays)
-
-    assert are_values_different(a, b)
-    assert are_values_different(c, d)
-    assert not are_values_different(a, a)
+    assert not are_values_different(None, None)
+
+
+def test_diff_values_array():
+    from numpy import NaN, array, inf
+    a_int = array([1, 2])
+    a_float = a_int.astype(float)
+
+    assert are_values_different(a_int, a_float)
+    assert are_values_different(a_int, a_int, a_float)
+    assert are_values_different(np.arange(3), np.arange(1, 4))
+    assert are_values_different(np.arange(3), np.arange(4))
+    assert are_values_different(np.arange(4), np.arange(4).reshape((2, 2)))
+    # no broadcasting should kick in - shape difference
+    assert are_values_different(array([1]), array([1, 1]))
+    assert not are_values_different(a_int, a_int)
+    assert not are_values_different(a_float, a_float)
+
+    # NaNs - we consider them "the same" for the purpose of these comparisons
+    assert not are_values_different(NaN, NaN)
+    assert not are_values_different(NaN, NaN, NaN)
+    assert are_values_different(NaN, NaN, 1)
+    assert are_values_different(1, NaN, NaN)
+    assert not are_values_different(array([NaN, NaN]), array([NaN, NaN]))
+    assert not are_values_different(array([NaN, NaN]), array([NaN, NaN]), array([NaN, NaN]))
+    assert not are_values_different(array([NaN, 1]), array([NaN, 1]))
+    assert are_values_different(array([NaN, NaN]), array([NaN, 1]))
+    assert are_values_different(array([0, NaN]), array([NaN, 0]))
+    # and some inf should not be a problem
+    assert not are_values_different(array([0, inf]), array([0, inf]))
+    assert are_values_different(array([0, inf]), array([inf, 0]))