New package version requirements

conda.recipe/meta.yaml

@@ -11,19 +11,17 @@ requirements:
     - pytest
     - setuptools
     - python
-    - mock
-    - numpy ==1.11.2
-    - pandas <=0.18.0
+    - numpy >=1.11.2
+    - pandas >=0.18.0
     - numba
     - toolz
     - six
   run:
     - pytest
     - setuptools
     - python
-    - mock
-    - numpy ==1.11.2
-    - pandas <=0.18.0
+    - numpy >=1.11.2
+    - pandas >=0.18.0
     - numba
     - toolz
     - six

environment.yml

@@ -2,16 +2,15 @@ name: taxcalc-dev
 dependencies:
 - pytest
 - setuptools
-- numpy =1.11.2
-- pandas =0.18.0
+- numpy >=1.11.2
+- pandas >=0.18.0
 - numba
-- bokeh =0.12.3
 - toolz
 - six
-- ipython
-- ipython-notebook
+- bokeh >=0.12.3
+- mock
 - pep8
 - pylint
-- mock
+- coverage
 - pip:
   - pytest-pep8

taxcalc/behavior.py

@@ -183,9 +183,8 @@ def response(calc_x, calc_y):
             new_ltcg = calc_x.records.p23250 * exp_term
             ltcg_chg = new_ltcg - calc_x.records.p23250
         # calculate charitable giving effect
-        no_charity_response = \
-            calc_y.behavior.BE_charity.tolist() == \
-            [0.0, 0.0, 0.0]
+        no_charity_response = (calc_y.behavior.BE_charity.tolist() ==
+                               [0.0, 0.0, 0.0])
         if no_charity_response:
             c_charity_chg = np.zeros(calc_x.records.dim)
             nc_charity_chg = np.zeros(calc_x.records.dim)
@@ -299,23 +298,22 @@ def _update_ordinary_income(taxinc_change, calc):
                         calc.records.c04470)
         agi_m_ided = agi - ided
         # assume behv response only for filing units with positive agi_m_ided
-        delta_income = np.where(agi_m_ided > 0., taxinc_change, 0.)
+        pos = np.array(agi_m_ided > 0., dtype=bool)
+        delta_income = np.where(pos, taxinc_change, 0.)
         # allocate delta_income into three parts
-        delta_wage = np.where(agi_m_ided > 0.,
-                              delta_income * calc.records.e00200 / agi_m_ided,
-                              0.)
-        other_income = agi - calc.records.e00200
-        delta_oinc = np.where(agi_m_ided > 0.,
-                              delta_income * other_income / agi_m_ided,
-                              0.)
-        delta_ided = np.where(agi_m_ided > 0.,
-                              delta_income * ided / agi_m_ided,
-                              0.)
+        winc = calc.records.e00200
+        delta_winc = np.zeros_like(agi)
+        delta_winc[pos] = delta_income[pos] * winc[pos] / agi_m_ided[pos]
+        oinc = agi - winc
+        delta_oinc = np.zeros_like(agi)
+        delta_oinc[pos] = delta_income[pos] * oinc[pos] / agi_m_ided[pos]
+        delta_ided = np.zeros_like(agi)
+        delta_ided[pos] = delta_income[pos] * ided[pos] / agi_m_ided[pos]
         # confirm that the three parts are consistent with delta_income
-        assert np.allclose(delta_income, delta_wage + delta_oinc - delta_ided)
+        assert np.allclose(delta_income, delta_winc + delta_oinc - delta_ided)
         # add the three parts to different calc.records variables
-        calc.records.e00200 = calc.records.e00200 + delta_wage
-        calc.records.e00200p = calc.records.e00200p + delta_wage
+        calc.records.e00200 = calc.records.e00200 + delta_winc
+        calc.records.e00200p = calc.records.e00200p + delta_winc
         calc.records.e00300 = calc.records.e00300 + delta_oinc
         calc.records.e19200 = calc.records.e19200 + delta_ided
         return calc

taxcalc/cli/install_local_taxcalc_package

@@ -1,8 +1,9 @@
 #!/bin/bash
 # USAGE: ./install_local_taxcalc_package
 # ACTION: (1) uninstalls any installed taxcalc package (conda uninstall)
-#         (2) builds local taxcalc=0.0.0 package (conda build)
-#         (3) installs local taxcalc=0.0.0 package (conda install)
+#         (2) executes "conda install conda-build" (if necessary)
+#         (3) builds local taxcalc=0.0.0 package (conda build)
+#         (4) installs local taxcalc=0.0.0 package (conda install)
 # NOTE: for those with experience working with compiled languages,
 #       building a local conda package is analogous to compiling an executable
 
@@ -14,6 +15,14 @@ if [ $? -eq 1 ]; then
     ./uninstall_taxcalc_package
 fi
 
+# install conda-build package if not present
+conda list build | awk '$1~/conda-build/{rc=1}END{exit(rc)}'
+if [ $? -eq 0 ]; then
+    echo "==> Installing conda-build package"
+    conda install conda-build --yes 2>&1 > /dev/null
+    echo "==> Continue building taxcalc package"
+fi
+
 # build taxcalc conda package
 cd ../../conda.recipe/
 conda build --python 2.7 . 2>&1 | awk '$1~/BUILD/||$1~/TEST/'

taxcalc/tests/conftest.py

@@ -1,11 +1,14 @@
 import os
-
+import numpy
 import pandas
 import pytest
-
 from taxcalc import Records
 
 
+# convert all numpy warnings into errors so they can be detected in tests
+numpy.seterr(all='raise')
+
+
 @pytest.fixture(scope='session')
 def tests_path():
     return os.path.abspath(os.path.dirname(__file__))

taxcalc/tests/test_behavior.py

@@ -11,6 +11,8 @@ def test_incorrect_Behavior_instantiation():
         behv = Behavior(behavior_dict=bad_behv_dict)
     with pytest.raises(ValueError):
         behv = Behavior(num_years=0)
+    with pytest.raises(FloatingPointError):
+        np.divide(1., 0.)
 
 
 def test_correct_but_not_recommended_Behavior_instantiation():

taxcalc/utils.py

@@ -925,7 +925,7 @@ def atr_graph_data(calc1, calc2,
 
     min_avginc : float
         specifies the minimum average expanded income for a percentile to
-        be included in the graph data
+        be included in the graph data; value must be positive
 
     Returns
     -------
@@ -962,6 +962,8 @@ def atr_graph_data(calc1, calc2,
         msg = ('atr_measure="{}" is neither '
                '"itax" nor "ptax" nor "combined"')
         raise ValueError(msg.format(atr_measure))
+    # . . check min_avginc value
+    assert min_avginc > 0.
     # calculate taxes and expanded income
     calc1.calc_all()
     calc2.calc_all()
@@ -994,19 +996,18 @@ def atr_graph_data(calc1, calc2,
     avginc_series = gdfx.apply(weighted_mean, 'expanded_income')
     avgtax1_series = gdfx.apply(weighted_mean, 'tax1')
     avgtax2_series = gdfx.apply(weighted_mean, 'tax2')
-    # compute average tax rates by income percentile
-    # pylint: disable=no-member
-    # (above pylint comment eliminates bogus np.divide warnings)
-    atr1_series = np.divide(avgtax1_series, avginc_series)
-    atr2_series = np.divide(avgtax2_series, avginc_series)
+    # compute average tax rates for each included income percentile
+    included = np.array(avginc_series >= min_avginc, dtype=bool)
+    atr1_series = np.zeros_like(avginc_series)
+    atr1_series[included] = avgtax1_series[included] / avginc_series[included]
+    atr2_series = np.zeros_like(avginc_series)
+    atr2_series[included] = avgtax2_series[included] / avginc_series[included]
     # construct DataFrame containing the two atr?_series
     lines = pd.DataFrame()
-    lines['avginc'] = avginc_series
     lines['base'] = atr1_series
     lines['reform'] = atr2_series
-    # drop percentiles with average income below the specified minimum
-    lines = lines[lines['avginc'] >= min_avginc]
-    lines.drop('avginc', axis=1, inplace=True)
+    # include only percentiles with average income no less than min_avginc
+    lines = lines[included]
     # construct dictionary containing plot lines and auto-generated labels
     data = dict()
     data['lines'] = lines