Hi @PyGeek03. Do you mind giving us some examples of inputs and expected outputs?

Hi @PyGeek03. Do you mind giving us some examples of inputs and expected outputs?

These examples are from the Microsoft documentation page:

• For the 1900 date system (starts counting from January 1, 1900):
  Input: 0. Output: January 1, 1900
  Input: 34519. Output: July 4, 1994
  Input: 35981. Output: July 5, 1998

• For the 1904 date system (starts counting from January 1, 1904):
  Input: 0. Output: January 1, 1904
  Input: 34519. Output: July 5, 1998
  Input: 35981. Output: July 6, 2002

An interesting issue is that for compatibility with Lotus 1-2-3, Excel considers 1900 to be a leap year, which is wrong. So the difference between the 2 date systems is not 4 years, but 4 years 1 day.

Hi @PyGeek03, I was looking a little more into the documentation. Is the input ever a fraction, or float? I just want to make sure there aren't any weird edge cases I'm missing before I propose a solution. Some of our dependencies don't play nicely with floats.

From the source code of xlrd.xldate, that seems possible. I reckon we can just write some wrapper functions around xlrd.xldate, and add it as a dependency?

Hey @PyGeek03. I was playing around with some possible solutions and it looks like we could just create a wrapper that uses our built-in shift function with a start date of either January 1, 1900 (or January 1, 1904). Both xlrd and Arrow use datetime as dependencies, so the behaviour from Arrow would be presumably the same as you are experiencing with xlrd. Also, this way avoids creating an additional dependency for Arrow.

Here is a pretty basic example of how this would work.

import arrow

base_date = arrow.get('1900-01-01')
print(base_date)

new_date = base_date.shift(days=1)
print(new_date)

@jadchaar what are your thoughts? The implementation of this would be very similar to how xlrd does it (https://github.com/python-excel/xlrd/blob/master/xlrd/xldate.py#L130).

Also, our built-in shift function does support fractions/float inputs, so this wouldn't be an issue.

Hey @PyGeek03. I was playing around with some possible solutions and it looks like we could just create a wrapper that uses our built-in shift function with a start date of either January 1, 1900 (or January 1, 1904). Both xlrd and Arrow use datetime as dependencies, so the behaviour from Arrow would be presumably the same as you are experiencing with xlrd. Also, this way avoids creating an additional dependency for Arrow.

Here is a pretty basic example of how this would work.

import arrow

base_date = arrow.get('1900-01-01')
print(base_date)

new_date = base_date.shift(days=1)
print(new_date)

@jadchaar what are your thoughts? The implementation of this would be very similar to how xlrd does it (https://github.com/python-excel/xlrd/blob/master/xlrd/xldate.py#L130).

Also, our built-in shift function does support fractions/float inputs, so this wouldn't be an issue.

I think a native shift solution would work nicely. Since this is a niche use case, we probably want to avoid adding an additional dependency.