Just to copy it here, we discussed the issue in the triage meeting and think we should just do it, as it seems clearly broken right now. This definitely needs the release note, though. I am also wondering if the test covers which day is excluded (start or end) in the path where only one of them is a business day?

Hmmm, somehow I had misremembered it as just including the enddate, but it actually does exclude the startdate for negative values (i.e. it always excludes the larger date). I am OK with giving it a shot, but it does make me a bit less confident that nobody might be working around this in weird ways.

I guess we merge it ping the mailing list?

Hmmm, somehow I had misremembered it as just including the enddate, but it actually does exclude the startdate for negative values (i.e. it always excludes the larger date). I am OK with giving it a shot, but it does make me a bit less confident that nobody might be working around this in weird ways.

I guess we merge it ping the mailing list?

Sounds good. You want me to send an email to the mailing list?

Happy if you do :).

@Kai-Striega The suggestion has been made to move busday functionality to numpy-financial at some point. What are your thoughts about that suggestion? It might be a bit messy, as the code is in C and depends on numpy.datetime.

Counts the number of valid days between begindates and
enddates, not including the day of enddates.

If enddates specifies a date value that is earlier than the
corresponding begindates date value, the count will be negative.

The function calculates the delta for a half open bracket. Question is: what is that bracket when you have enddates < begindates.

For indexing, we always exclude the last:

In [13]: x[1:4]
Out[13]: array([1, 2, 3])

In [15]: x[4:1:-1]
Out[15]: array([4, 3, 2])

Current behavior:

In [16]: np.busday_count(np.datetime64('2023-03-12'), np.datetime64('2023-03-13'))
Out[16]: 0

Sunday to Monday, with Monday excluded. Makes sense. But:

In [17]: np.busday_count(np.datetime64('2023-03-13'), np.datetime64('2023-03-12'))
Out[17]: 0

Monday through Sunday, would have expected -1.

The current behavior (a) does not match the docstring and (b) is hard to predict, so would support fixing it: always exclude the enddates.

@Kai-Striega The suggestion has been made to move busday functionality to numpy-financial at some point. What are your thoughts about that suggestion? It might be a bit messy, as the code is in C and depends on numpy.datetime.

A few thoughts:

• I believe the functionality would fit in with numpy-financial's scope
• They could be useful for someone – but I can't think of a use case of the top of my head
• I don't have much experience working with / maintaining C, however, this seems like a good time to learn
• This would require a new release of numpy-financial. I've been planning this for a while anyway, it might be a good time to do one.

A message was sent to numpy-discussion for comments: https://mail.python.org/archives/list/numpy-discussion@python.org/thread/MGTAMT5ELASBMVMCWU2IOPR5CQLD7CUP/

Let's give this a shot then, thanks.