Clarify what date formats are inferred for CSV (4 digits year, big and small endian only)

[sm-Fifteen]

Description

The try_parse_dates parameter of pl.read_csv is documented by saying "Most ISO8601-like formats can be inferred, as well as a handful of others.", with no further clarification of what it may or may not try to recognize. The user guide also mentions it, but doesn't explain what is covered either.

I believe format support should be in Polars's doc. If you look at Pandas in general, one its main pitfalls is that it tries to be automatically helpful, which is good for interactive use, but may lead to incorrect results when you're not looking. For dates especially, a wrong guess can end up completely mangling your data. Polars instead explicitly restricts what it will automatically parse to completely unambiguous formats, which is good, but if that safety isn't communicated to users, they might shy away from trying to use it.

In an unrelated discussion on the Pandas tracker with @MarcoGorelli, regarding Pandas' date handling, I found out that Polars only tried (at the time) a handful of big-endian, 4-digits-year date formats. In the time since, it has been modified to also try a number of little-endian 4-digits-year formats. This is also not currently mentionned anywhere I could find.

Link

https://docs.pola.rs/api/python/stable/reference/api/polars.read_csv.html#polars-read-csv