Julian

[dridk]

Description

Could you add Julian parser like pandas ? https://pandas.pydata.org/docs/reference/api/pandas.Timestamp.to_julian_date.html

[MarcoGorelli]

thanks for the request - I'm tempted to say this is too specialised and would be better suited to a polars plugin, but curious what others think

[mkleinbort]

Is this equivalent to a fixed integer offset? If so, yes, I'd say this is a little beyond what should be available as core functionality.

[alexander-beedie]

I think this could actually be quite useful for some domains where it is a typical representation (and also perhaps areas like sqlite/excel interop, which both use Julians natively). It's also easy/lightweight to implement; in fact I'm 80% done with it after half a coffee 😅☕

Update: I take it back; it's trivial to implement naïvely, but if you dig in deeper there are issues relating to floating point deviation that can bite you in the 🍑 if you're not careful; at one point I found myself looking through source code used at NASA and figured I'd sit on this one for a bit 🤣

[MarcoGorelli]

I've added this to polars-xdt

In [2]: from datetime import datetime
   ...: import polars_xdt  # noqa: F401
   ...: df = pl.DataFrame(
   ...:     {
   ...:         "date_col": [
   ...:             datetime(2013, 1, 1, 0, 30),
   ...:             datetime(2024, 1, 7, 13, 18, 51),
   ...:         ],
   ...:     }
   ...: )
   ...: with pl.Config(float_precision=10) as cfg:
   ...:     print(df.with_columns(
   ...:         julian_date=pl.col("date_col").xdt.to_julian_date()
   ...:     ))
   ...:
shape: (2, 2)
┌─────────────────────┬────────────────────┐
│ date_col            ┆ julian_date        │
│ ---                 ┆ ---                │
│ datetime[μs]        ┆ f64                │
╞═════════════════════╪════════════════════╡
│ 2013-01-01 00:30:00 ┆ 2456293.5208333335 │
│ 2024-01-07 13:18:51 ┆ 2460317.0547569445 │
└─────────────────────┴────────────────────┘

OK to close this and for people to use that, or does it need to be in the core library?

[alexander-beedie]

OK to close this and for people to use that, or does it need to be in the core library?

@MarcoGorelli: Probably worth having (I was eyeing it for polars-sql functionality just recently, and with SQLite/Excel both using Julians there might be some other benefits). Will have to take a look at what algorithm you used; the industrial-strength version has various tricks to guard against floating point rounding deviations creeping in :))

[MarcoGorelli]

sounds good, thanks! here it is:

https://github.com/pola-rs/polars-xdt/blob/6023324a58e9d7112f14205f86b0e7d958c8f2db/src/to_julian.rs#L9-L31