def get_light_times_from_behavior_file(session=None, logfile=None):
"""Return time light goes on and off in logfile from session"""
if session is not None:
lines = BeWatch.db.get_logfile_lines(session)
elif logfile is not None:
lines = TrialSpeak.read_lines_from_file(logfile)
else:
raise ValueError("must provide either session or logfile")
# They turn on in ERROR (14), INTER_TRIAL_INTERVAL (13),
# and off in ROTATE_STEPPER1 (2)
parsed_df_by_trial = TrialSpeak.parse_lines_into_df_split_by_trial(lines)
light_on = TrialSpeak.identify_state_change_times(
parsed_df_by_trial, state1=[13, 14], show_warnings=False)
light_off = TrialSpeak.identify_state_change_times(
parsed_df_by_trial, state0=2)
return light_on, light_off
def check_ardulines(logfile):
"""Error check the log file.
Here are the things that would be useful to check:
* File can be loaded with the loading function without error.
* Reported lick times match reported choice
* Reported choice and stimulus matches reported outcome
* All lines are (time, arg, XXX) where arg is known
* All state transitions are legal, though this is a lot of work
Diagnostics to return
* Values of params over trials
* Various stats on time spent in each state
* Empirical state transition probabilities
Descriptive metrics of each trial
* Lick times locked to response window opening
* State transition times locked to response window opening
"""
# Make sure the loading functions work
lines = TrialSpeak.read_lines_from_file(logfile)
pldf = TrialSpeak.parse_lines_into_df(lines)
plst = TrialSpeak.parse_lines_into_df_split_by_trial(lines)
def get_logfile_lines(session):
"""Look up the logfile for a session and return it"""
# Find the filename
bdf = get_behavior_df()
rows = bdf[bdf.session == session]
if len(rows) != 1:
raise ValueError("cannot find unique session for %s" % session)
filename = rows.irow(0)['filename']
# Read lines
lines = TrialSpeak.read_lines_from_file(filename)
# Split by trial
#~ splines = split_by_trial(lines)
return lines
if SHOW_WEBCAM:
cmd = 'xdotool search --name %s windowmove %d %d' % (
window_title, video_window_position[0], video_window_position[1])
while os.system(cmd) != 0:
# Should test here if it's been too long and then give up
print "Waiting for webcam window"
time.sleep(.5)
while True:
## Chat updates
# Update chatter
chatter.update(echo_to_stdout=ECHO_TO_STDOUT)
# Read lines and split by trial
# Could we skip this step if chatter reports no new device lines?
logfile_lines = TrialSpeak.read_lines_from_file(logfilename)
splines = TrialSpeak.split_by_trial(logfile_lines)
# Run the trial setting logic
# This try/except is no good because it conflates actual
# ValueError like sending a zero
#~ try:
translated_trial_matrix = ts_obj.update(splines, logfile_lines)
#~ except ValueError:
#~ raise ValueError("cannot get any lines; try reuploading protocol")
## Update UI
if RUN_UI:
ui.update_data(logfile_lines=logfile_lines)
ui.get_and_handle_keypress()
if SHOW_WEBCAM:
cmd = 'xdotool search --name %s windowmove %d %d' % (
window_title, video_window_position[0], video_window_position[1])
while os.system(cmd) != 0:
# Should test here if it's been too long and then give up
print "Waiting for webcam window"
time.sleep(.5)
while True:
## Chat updates
# Update chatter
chatter.update(echo_to_stdout=ECHO_TO_STDOUT)
# Read lines and split by trial
# Could we skip this step if chatter reports no new device lines?
logfile_lines = TrialSpeak.read_lines_from_file(logfilename)
splines = TrialSpeak.split_by_trial(logfile_lines)
# Run the trial setting logic
# This try/except is no good because it conflates actual
# ValueError like sending a zero
#~ try:
translated_trial_matrix = ts_obj.update(splines, logfile_lines)
#~ except ValueError:
#~ raise ValueError("cannot get any lines; try reuploading protocol")
## Update UI
if RUN_UI:
ui.update_data(logfile_lines=logfile_lines)
ui.get_and_handle_keypress()
def generate_mplayer_guesses_and_sync(metadata,
user_results=None, guess=(1., 0.), N=4, pre_time=10):
"""Generates best times to check video, and potentially also syncs.
metadata : a row from bv_files to sync. Needs to specify the following:
'filename' : behavioral filename
'guess_vvsb_start'
'duration_video'
'filename_video'
The fit is between these datasets:
X : time of retraction from behavior file, minus the test_guess_vvsb
in the metadata.
Y : user-supplied times of retraction from video
The purpose of 'initial_guess' is to generate better guesses for the user
to look in the video, but the returned data always use the combined fit
that includes any initial guess. However, test_guess_vvsb is not
accounted for in the returned value.
N times to check in the video are printed out. Typically this is run twice,
once before checking, then check, then run again now specifying the
video times in `user_results`.
If the initial guess is very wrong, you may need to find a large
gap in the video and match it up to trials info manually, and use this
to fix `guess` to be closer.
"""
initial_guess = np.asarray(guess)
# Load trials info
trials_info = TrialMatrix.make_trial_matrix_from_file(metadata['filename'])
splines = TrialSpeak.load_splines_from_file(metadata['filename'])
lines = TrialSpeak.read_lines_from_file(metadata['filename'])
parsed_df_split_by_trial = \
TrialSpeak.parse_lines_into_df_split_by_trial(lines)
# Insert servo retract time
trials_info['time_retract'] = TrialSpeak.identify_servo_retract_times(
parsed_df_split_by_trial)
# Apply the delta-time guess to the retraction times
test_guess_vvsb = metadata['guess_vvsb_start'] #/ np.timedelta64(1, 's')
trials_info['time_retract_vbase'] = \
trials_info['time_retract'] - test_guess_vvsb
# Choose test times for user
video_duration = metadata['duration_video'] / np.timedelta64(1, 's')
test_times, test_next_times = generate_test_times_for_user(
trials_info['time_retract_vbase'], video_duration,
initial_guess=initial_guess, N=N)
# Print mplayer commands
for test_time, test_next_time in zip(test_times, test_next_times):
pre_test_time = int(test_time) - pre_time
print 'mplayer -ss %d %s # guess %0.1f, next %0.1f' % (pre_test_time,
metadata['filename_video'], test_time, test_next_time)
# If no data provided, just return
if user_results is None:
return {'test_times': test_times}
if len(user_results) != N:
print "warning: len(user_results) should be %d not %d" % (
N, len(user_results))
return {'test_times': test_times}
# Otherwise, fit a correction to the original guess
new_fit = np.polyfit(test_times.values, user_results, deg=1)
resids = np.polyval(new_fit, test_times.values) - user_results
# Composite the two fits
# For some reason this is not transitive! This one appears correct.
combined_fit = np.polyval(np.poly1d(new_fit), np.poly1d(initial_guess))
# Diagnostics
print os.path.split(metadata['filename'])[-1]
print os.path.split(metadata['filename_video'])[-1]
print "combined_fit: %r" % np.asarray(combined_fit)
print "resids: %r" % np.asarray(resids)
return {'test_times': test_times, 'resids': resids,
'combined_fit': combined_fit}