Skip to content

Latest commit

 

History

History
71 lines (49 loc) · 2.62 KB

README.md

File metadata and controls

71 lines (49 loc) · 2.62 KB
Raw

dtparse

Fast datetime parser for Python written in Rust. Parses 10x-15x faster than datetime.strptime.

It isn't a drop-in replacement for datetime.strptime, although they work similarly most of the time. Instead, think of it as of a library of its own.

The full list of supported specifiers (%Y, %m, %d, ...) can be found in chrono documentation. They are a bit different from Python's.

Notable differences from datetime.strptime

Required parts of the template

dtparse.parse requires hours, minutes, and the full date to be present in the template.

For datetime.strptime, all parts of the template are optional.

datetime.strptime('2018/01/02 12', '%Y/%m/%d %H')  # no minutes in the template
# datetime(2018, 1, 2, 12, 0)

dtparse.parse('2018/01/02 12', '%Y/%m/%d %H')
# ValueError: input is not enough for unique date and time

Specifiers for sub-second precision

Python uses %f as a specifier for microseconds, zero-padded on the left. Chrono's %f is different - left-aligned, optionally zero-padded to the right.

datetime.strptime('2004-12-01 13:02:47.123456', '%Y-%m-%d %H:%M:%S.%f')
# datetime(2004, 12, 1, 13, 2, 47, 123456)

dtparse.parse('2004-12-01 13:02:47.123456', '%Y-%m-%d %H:%M:%S.%f')
# datetime(2004, 12, 1, 13, 2, 47, 123)

On the other hand, Python's .%f works very close to chrono's %.f. The only difference is that chrono accepts up to 9 digits, because it parses nanoseconds, not microseconds.

datetime.strptime('2004-12-01 13:02:47.123456', '%Y-%m-%d %H:%M:%S.%f')
# datetime(2004, 12, 1, 13, 2, 47, 123456)

dtparse.parse('2004-12-01 13:02:47.123456', '%Y-%m-%d %H:%M:%S%.f')
# datetime(2004, 12, 1, 13, 2, 47, 123456)
datetime.strptime('2004-12-01 13:02:47.123456789', '%Y-%m-%d %H:%M:%S.%f')
# ValueError: unconverted data remains: 789

dtparse.parse('2004-12-01 13:02:47.123456789', '%Y-%m-%d %H:%M:%S%.f')
# datetime(2004, 12, 1, 13, 2, 47, 123456)

How does it work?

Uses chrono library under the cover. Binds Rust to Python via PyO3. Wheels are built and distributed using GitHub actions, see the workflow here.

Origins

This library was written for my talk on Rust extensions on Yandex Pytup. It is intended to be both a stable library people can use, and a reference for those interested in extending Python with Rust.