Skip to content

Python parser for abbreviated, ambiguous, incomplete dates in multiple languages

License

Notifications You must be signed in to change notification settings

crgz/fuzzy_dates

 
 

Repository files navigation

Python Parser for Abbreviated Dates

PypI - Version issues stars - stars stars - stars contributions - contributions

Key FeaturesHow To UseHow it worksCommon use casesOperationsRoadmap

Have you ever tried to understand a date like 11-09, št? Is the št an abbreviation of a month or a weekday? Which of those numbers represent the month or the day? This library leverages on Good Ol' Fashioned AI to parse abbreviated, ambiguous, and incomplete dates in multiple languages.

Key Features

  • Language auto-detection
  • Easily expandable into new languages (30 languages are currently supported)
  • Support for multiple date formats
  • Support for abbreviated weekdays
  • Support for abbreviated months
  • Support for ambiguous month/day numbers

How To Use

The most straightforward way to parse dates is to use the datetime.date() function, that wraps around most of the functionality of the module. This example shows a basic usage of the library to parse the date: "11-09, št":

from datetime import datetime
from fuzzy_parser.engine import Engine

start = datetime(2021, 4, 27).date()
(semantic, _) = Engine(start).when('11-09, št')
for single_date in semantic:
    print(single_date.strftime("%Y-%m-%d"))

The output will show the expected

2021-09-11

How it works

The abbreviation "št" could stand for:

  • Šeštadienis which means in Saturday in Lithuanian
  • Štvrtok which means in Thursday in Slovak

Lithuanian is spoken in Lithuania and in Latvia. Slovak is spoken in Slovakia but also by a minority in the Czech Republic. These countries use different date representations: Czech Republic, Latvia and Slovakia have the day written first because of the "little" date endianness format used as the standard in the country. Lithuania, on the other hand, uses the "big" date endianness format which means that the month is written first. The system factor in all these facts and is able to come with the right answers:

In the case of interpreting the abbreviation as a Saturday:

  • 9 of November 2024
  • 11 of September 2027

In the case of interpreting the abbreviation as a Thursday:

  • 11 of September 2025

For further details have a look at the underlying implementation. In addition, the unit tests might give an impression on how to use this library.

Common use cases

Consuming data from different sources:

  • Scraping: extract dates from different places with several formats and languages
  • IoT: consuming data coming from different sources with different date formats
  • Tooling: consuming dates from different logs / sources
  • Format transformations: when transforming dates coming from different files (PDF, CSV, etc.) to other formats (database, etc).

Operations

We are leveraging GNU Make to automate frequent actions. Please use the following command will show the available recipes to use for operating the local development environment:

make help
Click to see the available recipes
Command Description
help Print this help
synchronize Switch to the main branch, fetch changes & delete merged branches
test Run the test suite
run-time Install the packages packs required for the development environment
install Install the latest library release
uninstall Uninstall the latest library release
release Release recipe to be use from Github Actions
bump Increase the version number
build Build and check distribution packages
publish Publish the diagrams
workflow Creates the Diagrams
remove-all Remove packages and packs
clean Remove debris

The following command stores pypi.org token to "Upload packages" in the local key store to be used by the publish option:

secret-tool store --label='https://pypi.org/manage/account/token/' user ${USER} domain pypi.org

Compatibility

Tested with SWI-Prolog version 8.2.4 on Ubuntu 20.04

License

Distributed under the MIT License. See LICENSE file for more information.

Releases

No releases published

Packages

No packages published

Languages

  • Makefile 47.8%
  • Python 41.5%
  • Shell 10.7%