to_evenly() assigns wrong instants

[andbiz]

signal_unevenly = ph.UnevenlySignal(np.array([1,2,3,4,5,6,8,9,10]),
                                      sampling_freq = 100, signal_nature = '', start_time = 0,
                                      x_values = np.array([1,2,3,4,5,6,7,9,10]),
                                      x_type='instants')
signal_unevenly.get_times()[0] # = 1 <=OK

signal_evenly = signal_unevenly.to_evenly()
signal_evenly.get_times()[0] # = 0 <= KO

[Alebat]

Start time is set to zero so the conversion sets the evenly signal's start time to zero.

Changes: now the start time of the new evenly is set to the first unevenly' sample not caring about the unevenly start time.

Alternatives: pad the signal holding before the first unev. sample till the start time. The absence of signal between start_time and 1st unev sample can maybe be a problem. In this case reopen the issue.

[andbiz]

The bug has reappeared:

import pyphysio as ph
signal_unevenly = ph.UnevenlySignal(np.array([1,2,3,4,5,6,8,9,10]),
                                      sampling_freq = 100, signal_nature = '', start_time = 0,
                                      x_values = np.array([1,2,3,4,5,6,7,9,10]),
                                      x_type='instants')
print(signal_unevenly.get_times()[0]) # = 1 <=OK

signal_evenly = signal_unevenly.to_evenly()
print(signal_evenly.get_times()[0]) # = 0 <= KO

[Alebat]

You set the start_time to be 0:

the difference with test_issues.testissue22 (code posted in the first comment) is that you are setting the start_time=0 instead of None, comparison:

• None: the start_time becomes the time of the first sample by the following lines

  # Get indices removing start_time
  if start_time is None:
      start_time = x_values[0]

• 0: start_time is just checked to be less than the first sample as follows and remains zero

  assert start_time <= x_values[0], "More than one sample at or before start_time"

But, it should work in any case, the new evenly should always have as start_time the first unevenly sample as it is not defined before. Should it be NaN instead and start from the original start_time?

• [ ] Yes

I found a commit cf9e676 that changes the start_time of the new signal from the first sample time to the signal's start_time. Who is the author? :smiley: It should be the first time the next start_time, is it correct? I edited :+1:

Also updated issue test to test both None and 0 but it does not work...

[Alebat]

Done