ENH: Initial implementation of calculation of astronomical Julian Date. by timcera · Pull Request #4041 · pandas-dev/pandas (original) (raw)

This issue (#4065) is why I haven't worked on this. What is the use to me to implement a calendar that starts in 4714 BC (proleptic gregorian), if pandas can't represent the date?

I tried to wade through the pandas code to help out with #4065, but quickly got in way over my head.

Back to this PR, it seems that the only issues are that I need to hook up travis, and possibly change the tests? Is there anything else? It is possible that I could do these things soonish.

My main concern though that I haven't got any feedback on is whether this is the right way to implement this functionality. Should it be a part of the Timestamp and Datetimeindex OR should it be a seperate utility function? Is there a pandas way? What is the big picture in how this should be implemented? Should pandas even care about calendars other than gregorian? Would other calendars make sense? Should the julian date be a completely separate object with a 'to_gregorian()' function? Should I just keep this functionality in my code? Should it be represented as a tuple (integer day, float fraction of day) which would allow representation of smaller intervals. I don't have any tests for small intervals though I include nanosecond in the calculations - don't know whether it actually does anything.

Kindest regards,
Tim