GH AllenInstitute#508 - Remove np.nan values from np.diff array

np.nan values were not being dropped from the end of the np.diff array
before calling np.flatnonzero to remove non-zero values in get_stim_epoch.
This was causing get_stim_epoch to incorrectly identify the end of the
stimulus epoch as the same as the end of the sweep epoch, which was also
causing the end of the experiment epoch to be incorrect as well since it
relys on the stimulus epoch. By dropping the np.nan values from the end
of the difference array (di) in get_stim_epoch, the correct indices for
end of the stimulus epoch as well as for the experiment epoch are now being
returned. This should fix the issue with sweeps being incorrectly tagged
with 'Recording stopped before completing the experiment epoch'.

CHANGELOG.md

@@ -2,7 +2,12 @@
 All notable changes to this project will be documented in this file.
 
 ## Unreleased
-
+This pull request adds a line of code to the get_stim_epoch function in epochs.py to 
+remove np.nan values from the array of differences between data points before removing 
+all non-zero values. This change allows get_stim_epoch to find the correct ending index 
+of the stimlus epoch and the correct ending index of the experiment epoch as well since
+it relys on the stimulus epoch, which should fix the issue of sweeps being incorrectly 
+tagged with "Recording stopped before completing the experiment epoch".
 ### Added
 
 ### Changed

ipfx/epochs.py

@@ -108,6 +108,7 @@ def get_stim_epoch(i, test_pulse=True):
     """
 
     di = np.diff(i)
+    di = di[~np.isnan(di)] # != NaN
     di_idx = np.flatnonzero(di)   # != 0
 
     if test_pulse:

tests/test_epochs.py

@@ -109,6 +109,11 @@ def test_get_test_epoch(i, sampling_rate, test_epoch):
                                  [0, 0, 0, 0, 0, 0, 0],
                                  None
                              ),
+                             #   array containing np.nan
+                             (
+                                 [0, 0, 1, 1, 0, 0, 2, 2, 2, 0, 0, np.nan, np.nan],
+                                 (6, 8)
+                             ),
 
                          ]
                          )