TST: dt64 units #56261
TST: dt64 units #56261
Conversation
| @@ -130,8 +132,15 @@ def df_ref(datapath): | |||
| return df_ref | |||
|
|
|||
|
|
|||
| def adjust_expected(expected: DataFrame, read_ext: str) -> None: | |||
| def get_exp_unit(read_ext: str, engine: str | None) -> str: | |||
There was a problem hiding this comment.
So your ultimate vision here is different engines will return different bases? Or is there just considered a temporary solution?
There was a problem hiding this comment.
In #55901 this becomes
def get_exp_unit(read_ext: str, engine: str | None) -> str:
unit = "us"
if (read_ext == ".ods") ^ (engine == "calamine"):
unit = "s"
return unit
There was a problem hiding this comment.
What makes ods files or the calamine reader different here? My first thought is that is surprising behavior to have those be the only ones return seconds
There was a problem hiding this comment.
no idea. id be OK with coercing them all to micros
There was a problem hiding this comment.
What is the largest representable date for microsecond precision?
From what I see Excel only can display millseconds, but doesn't offer first class formula support for it. The ODS specification point 18.3.14 links its datetime format to an XML Schema Part 2
https://www.w3.org/TR/2004/REC-xmlschema-2-20041028/#dateTime
Which only mentions "fractional seconds" but without too much detail.
So seems like a wild west of implementation possibilities. Excel has an upper limit on dates at December 31, 9999 so maybe we just try to cover that?
There was a problem hiding this comment.
December 31, 9999 23:59:59.999999 is also the highest value supported by the stdlib datetime. microseconds can go further than that, but most of the time when we get microseconds it is because we got a stdlib datetime objects
There was a problem hiding this comment.
(btw i definitely want to have this discussion but would prefer it to happen in #55901, this PR is just aimed at diff-trimming to make that one easier)
mroeschke
added
Testing
|
Thanks @jbrockmendel |
Aimed at trimming the diff in #55901