Use last element as end time

[magnusuMET]

Added an additional fix for type comparisons

[avaldebe]

Is this the cleanest way to set the start/end
Also, why is does start depends on self._end_include and end depends on self._start_include?

maybe itertools.chain can improve the readability

from itertools import chain
[...]
        if self._start_include or self._startend_include:
            start = min(s for s, e in chain(self._start_include, self._startend_include))
        else:
            start = datetime.max

        if self._end_include or self._startend_include:
            end = max(e for s, e in chain(self._end_include, self._startend_include))
        else:
            end = datetime.min

[heikoklein]

I don't think the chain-version is very readable here.

[avaldebe]

I don't think the chain-version is very readable here.

Yes, at the cost of creating new lists this could be written on a simpler way

        if self._start_include or self._startend_include:
            start = min(s for s, e in self._start_include + self._startend_include)
        else:
            start = datetime.max

        if self._end_include or self._startend_include:
            end = max(e for s, e in self._end_include + self._startend_include)
        else:
            end = datetime.min

In any case, my question still stands

Also, why is does start depends on self._end_include and end depends on self._start_include?

[heikoklein]

The tuple values of (start, end) are the time-ranges of the filter, while start_include/end_include refer to measurements starttime/endtime.

[avaldebe]

Looks like includes is assumed to be an Iterable, in which case
if len(includes) == 0: is equivalent to if not includes:.

With this in mind, this section can be written as:

        idx = np.repeat(not includes, len(times1))
        for start, end in includes:
            idx |= (np.datetime64(start) <= times1) & (times2 <= np.datetime64(end))

[heikoklein]

While idx = np.repeat(not includes, len(times1)) is written using very few characters, the boolean value of includes is not related to the boolean value of idx except for a chain of action, e.g. we could also end up with idx = np.repeat(includes, ...) if we would use the idx differently.

Rather than having to add long comments, please use rather more code.

[heikoklein]

Thanks, looks good.
Please check if we could improve the code by internally storing all datetimes as np.datetime64[s] directly?

[heikoklein]

Thanks, very well that you also included a test.

[heikoklein]

I'm not sure why the conversion to np.datetime64 is needed here, or, more specific, if operations need to be done as np.datetime64 then maybe _str_list_to_datetime_list should do the conversion already?

[magnusuMET]

Probably should store dt's as numpy format, opened issue #59 for this

[heikoklein]

I don't think the chain-version is very readable here.