Will this error in Pandas 3

[henrydingliu]

@genedan will explicitly setting the ult_val using nanosecond, like this section of the doc, cause any problems in pandas 3?

[genedan]

Yeah. Precision is also a problem with Pandas 2. I don't think we mentioned that precision has to be nanosecond for Pandas 2. I tried running the tests with cl.options.set_option("ULT_VAL", "2050-12-31 23:59:59.9") on Pandas 2, and they failed.

That opens the question as to what to do about it. Should we:

• Allow a round number, and then adjust the precision?
• Restrict the user to enter in the correct precision.

The first is more user friendly, but harder to program. The second is easier to program but it puts the burden on the user to do the correct math, which might be okay for annual but would be annoying if they wanted a different grain.

[genedan]

Attaching the failed tests for reference:

[genedan]

I think it has to be set to something at month end. I tried various dates like:

'2050-05-31 23:04:59.999999999'

'2050-05-12 23:04:59.999999999'

Which led to similar failures, but when set to '2050-05-31 23:59:59.999999999', the tests were ok.

[henrydingliu]

we can be practical about it and only allow .set_option("ULT_VAL",) to take a tuple of (year:int,month:int). and we can document ULT_VAL accordingly in #828.

[genedan]

Sounds good. Convert to issue?

[henrydingliu]

this has now been resolved