FIX merge_events when replace_events=False

[teonbrooks]

this was brought on the listserv by Lucy MacGregor.
When we create the merged events, we don't get rid of the redundant copy when replace_events=False.

Below you can replicate the error.

"""
========================================================
Sample script to merge events
========================================================

Reads sample data, finds events from event file, merges events to create new events whilst retaining the originals.

Desired behaviour: create *multiple* new event codes whilst retaining originals.

Current behaviour:  If 1 new event is created, for every sample point there are now 2 events: in
cases where original event is merged this results in 1 original and 1 new event as expected. However, in cases
where the original event code isn't merged there is a repetition of the original event code. If
n new events are created, for every sample point there are now n^2 events (presumably because
replace_events=False). 

8.3.2016
Lucy.MacGregor@mrc-cbu.cam.ac.uk

"""

import mne
from mne import io
from mne.datasets import sample

print(__doc__)

data_path = sample.data_path()

###############################################################################
# Set parameters
raw_fname = data_path + '/MEG/sample/sample_audvis_filt-0-40_raw.fif'
event_fname = data_path + '/MEG/sample/sample_audvis_filt-0-40_raw-eve.fif'


#   Setup for reading the raw data
raw = io.Raw(raw_fname)
events = mne.read_events(event_fname)

# merge events
events=mne.event.merge_events(events, [1,2], 12, replace_events=False)
events=mne.event.merge_events(events, [3,4], 34, replace_events=False)
events=mne.event.merge_events(events, [1,2,3,4], 1234, replace_events=False)


event_id = dict(A= 1, B= 2, C=3, D=4, E=5, F=12, G=34, H=1234)

[agramfort]

can you add a test?

now we should recommend using dict hierarchical event ids.

travis is not happy

[teonbrooks]

test to come. wanted to make sure that this is right corrected behavior. travis breaks because the current test for this fails.

now we should recommend using dict hierarchical event ids.

agreed.

[agramfort]

@teonbrooks closing in favor of #3076

it was indeed pretty buggy...

[jona-sassenhagen]

We probably need to feature hierarchical event_ids better. Put it more prominently in some of the examples.