Use atol instead of rtol

Author: rweigel

hapiclient/test/readcompare.py

@@ -1,5 +1,4 @@
 import os
-import pickle
 import numpy as np
 
 from hapiclient import hapi
@@ -38,8 +37,8 @@ def equalNonFloats(a, b):
     allequal = True
     for name in a.dtype.names:
         if np.issubdtype(a[name].dtype, np.integer) or np.issubdtype(a[name].dtype, np.flexible):
-            # https://docs.scipy.org/doc/numpy-1.10.1/reference/arrays.scalars.html
             # Parameter type is string or integer
+            # https://docs.scipy.org/doc/numpy-1.10.1/reference/arrays.scalars.html
             if not np.array_equal(a[name], b[name]):
                 allequal = False
                 if debug: print(name + ' values differ.')
@@ -52,16 +51,19 @@ def closeFloats(a, b):
     allclose = True
     for name in a.dtype.names:
         if np.issubdtype(a[name].dtype, np.inexact):
-            # https://docs.scipy.org/doc/numpy-1.10.1/reference/arrays.scalars.html
             # Parameter is floating point number
-            if np.allclose(a[name], b[name], rtol=1e-15, atol=0.0, equal_nan=True):
-                if not np.array_equal(a[name], b[name]):
-                    if debug: print(name + ' values equal within rtol=1e-15.')
+            # See https://docs.scipy.org/doc/numpy-1.10.1/reference/arrays.scalars.html
+            atol = np.finfo(a[name].dtype.str).eps
+            if np.allclose(a[name], b[name], rtol=0.0, atol=atol, equal_nan=True):
+                if debug and not np.array_equal(a[name], b[name]):
+                    mdiff = np.max(np.abs(a[name] - b[name]))
+                    print('All values in parameter ' + name \
+                          + ' equal within absolute tolerance ' \
+                          + 'of %.2e; max |diff| = %.2e' % (atol, mdiff))
             else:
                 allclose = False
-                if debug: print(name + ' values not equal within rtol=1e-15.')
-                if debug: import pdb; pdb.set_trace()
-
+                print('All values in parameter ' + name \
+                      + ' not equal within absolute tolerance of %.2e.' % atol)
 
     return allclose
 
@@ -70,6 +72,7 @@ def closeFloats(a, b):
 logfile = os.path.realpath(__file__)[0:-2] + "log"
 with open(logfile, "w") as f: pass
 
+
 def xprint(msg):
     print(msg)
     f = open(logfile, "a")
@@ -82,12 +85,12 @@ def readcompare(server, dataset, parameters, run, opts):
     # Note that for this dataset, there are differences in
     # the numeric values that seem not to be due to issues
     # with the reader. This needs investigation.
-    
+
     dataset = 'dataset1'
     start = '1970-01-01'
 
     allpass = True
-    
+
     if run == 'short':
         stop = '1970-01-01T00:00:03' # Returns 3 time values
 
@@ -98,7 +101,7 @@ def readcompare(server, dataset, parameters, run, opts):
         stop= '1970-01-02T00:00:00' # Returns 86400 time values
 
     # Checks that all four read methods give same result.
-    # Does not check that an individual read is correct. 
+    # Does not check that an individual read is correct.
     # Do this manually.
 
     opts['format'] = 'csv'
@@ -114,66 +117,50 @@ def readcompare(server, dataset, parameters, run, opts):
         xprint('Method                total      d/l->buff  parse buff')
         xprint('___________________________________________________________')
 
-    
+
     opts['method'] = 'numpynolength'
     data, meta  = hapi(server, dataset, parameters, start, stop, **opts)
     xprint('csv; numpy; no len.  %8.4f   %8.4f   %8.4f' % \
             (meta['x_totalTime'], meta['x_downloadTime'], meta['x_readTime']))
     datalast = data 
 
-    
+
     opts['method'] = 'pandasnolength'
     data, meta  = hapi(server, dataset, parameters, start, stop, **opts)
 
-    diffs = ''
-    if np.array_equal(data, datalast):
-        diffs = '(diffs in float values <= 1e-15)'
-
-    xprint('csv; pandas; no len. %8.4f   %8.4f   %8.4f %s' % \
-            (meta['x_totalTime'], meta['x_downloadTime'], meta['x_readTime'], diffs))
+    xprint('csv; pandas; no len. %8.4f   %8.4f   %8.4f' % \
+            (meta['x_totalTime'], meta['x_downloadTime'], meta['x_readTime']))
 
     allpass = comparisonOK(data, datalast)
     datalast = data
 
-    
+
     opts['method'] = 'numpy'
     data, meta  = hapi(server, dataset, parameters, start, stop, **opts)
 
-    diffs = ''
-    if np.array_equal(data, datalast):
-        diffs = '(diffs in float values <= 1e-15)'
-
-    xprint('csv; numpy           %8.4f   %8.4f   %8.4f %s' % \
-            (meta['x_totalTime'], meta['x_downloadTime'], meta['x_readTime'], diffs))
+    xprint('csv; numpy           %8.4f   %8.4f   %8.4f' % \
+            (meta['x_totalTime'], meta['x_downloadTime'], meta['x_readTime']))
 
     allpass = comparisonOK(data, datalast)
     datalast = data
 
-    
+
     opts['method'] = 'pandas'
     data, meta  = hapi(server, dataset, parameters, start, stop, **opts)
 
-    diffs = ''
-    if np.array_equal(data, datalast):
-        diffs = '(diffs in float values <= 1e-15)'
-
-    xprint('csv; pandas          %8.4f   %8.4f   %8.4f %s' % \
-            (meta['x_totalTime'], meta['x_downloadTime'], meta['x_readTime'], diffs))
+    xprint('csv; pandas          %8.4f   %8.4f   %8.4f' % \
+            (meta['x_totalTime'], meta['x_downloadTime'], meta['x_readTime']))
 
     allpass = comparisonOK(data, datalast)
     datalast = data
-    
+
 
     opts['format'] = 'binary'
     opts['method'] = '' # Ignored
     data, meta  = hapi(server, dataset, parameters, start, stop, **opts)
 
-    diffs = ''
-    if np.array_equal(data, datalast):
-        diffs = '(diffs in float values <= 1e-15)'
-
-    xprint('binary               %8.4f   %8.4f   %8.4f %s' % \
-            (meta['x_totalTime'], meta['x_downloadTime'], meta['x_readTime'], diffs))
+    xprint('binary               %8.4f   %8.4f   %8.4f' % \
+            (meta['x_totalTime'], meta['x_downloadTime'], meta['x_readTime']))
 
     allpass = comparisonOK(data, datalast)