Rework DateTime.Parse/TryParse and DateTimeOffset.Parse/TryParse

[MangelMaxime]

Description

Hello @dbrattli,

The last failing test I have in Thoth.Json currently are related to DateTime/DateTimeOffset and I work on a fix for that. While working on it, I discovered that the parse function is using dateutil which is not a Python standard API and so that function doesn't work.

Fable/src/fable-library-py/fable_library/date.py

Lines 226 to 229 in 9276528

	def parse(string: str, detectUTC: bool = False) -> datetime:
	from dateutil import parser
	return parser.parse(string)

Currently, I am doing something really naive to use the native API and support at least a subset of the .NET formatting date:

def parse(string: str, detectUTC: bool = False) -> datetime:
    # from dateutil import parser

    try:
        return datetime.fromisoformat(string)
    except ValueError:
        try:
            return datetime.strptime(string, "%H:%M:%S")
        except ValueError:
            try:
                return datetime.strptime(string, "%I:%M:%S %p")
            except ValueError:
                # IMPORTANT: This format needs to be tested after %I:%M:%S %p
                # Otherwise, it happens that %H:%M:%S %p incorrectly parses some
                # of the string that can overlap with %I:%M:%S %p
                # Example: 1:5:34 PM
                try:
                    return datetime.strptime(string, "%H:%M:%S %p")
                except ValueError:
                    raise ValueError("Unsupported format by Fable")

Really ugly, I know ^^. This is just a temporary solution instead of something that crash rightnow.

If we want to use dateutil, we need to publish fable-library to PyPi (or whatever is the package repository in Python) and check if this doesn't create others issues like we discussed on Slack regarding the Rust distribution of the library.

Or embed the dateutil source code inside of Fable, like it has been done in the past for BigInt in JavaScript.

How do you envision the DateTime/DateTimeOffset parsing functions to be implemented?

[dbrattli]

@MangelMaxime I would be happy if we could get rid of the dateutil dependency. Here's a copilot refactor of your code:

def parse(string: str, detectUTC: bool = False) -> datetime:
    formats = {
        # Matches a date-time string in the format "YYYY-MM-DDTHH:MM:SS"
        r"\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}": "%Y-%m-%dT%H:%M:%S",

        # Matches a time string in 24-hour format "HH:MM:SS"
        r"\d{2}:\d{2}:\d{2}": "%H:%M:%S",

        # Matches a time string in 12-hour format with AM/PM "HH:MM:SS AM" or "HH:MM:SS PM"
        r"\d{2}:\d{2}:\d{2} [AP]M": "%I:%M:%S %p",
    }

    for pattern, format in formats.items():
        if re.fullmatch(pattern, string):
            return datetime.strptime(string, format)

    raise ValueError("Unsupported format by Fable")

[MangelMaxime]

This looks cleaner and avoid the infinite nested in the code. I will give it a try thank you