Fuzzy Date Parser

Overview

Create a class that allows you to configure certain items like locale and start of week values and then pass in a string and get back a dictionary of the original string, the parsed date if any and any issues found during parsing.

The really hard part of identifying what text to parse inside of a string of text is not the scope of this project, but rather if given a text item that contains any of the following, parse it:

• 5 minutes from now
• 5 min from now
• 5m from now
• in 5 minutes
• 5 min
• 5 min before now
• 5 min before next week
• 5 hours from noon
• 5 hours before noon
• in 2 weeks
• 7 days before now
• next day
• tomorrow
• yesterday
• next week
• last week

Heikki suggested:

• every two weeks
• every other day
• on even days
• third Friday of every other month except December

Number parsing - the code should be able to parse "five minutes" the same as "5 minutes"

The code should also parse most known date formats if also given.

The library would use a combination of regular expressions, keywords and state tables to chunk the given text into it's various date related peices and then extract either day/month/year information or day/month/year delta information. Once this is known, the source date can have the delta information applied to it and you have your desired date.

Task List

• Convert current single-py-file tests to unit tests
• Run code thru py-lint
• Add error handling code
  • allow soft errors to return partial dates and a flag
  • ensure hard errors are flagged and return no date information
• Add fixed date handling
• Add code to sense numbers and replace them with the proper numbers: five -> 5, twenty five -> 25
• Move code to new repository and create distutils/setuptools packaging
• Documentation
• Add locale handling
  • make sure timezones are honored
  • ensure default locale info can be provided to fill in the unknown values during parsing
• Add recurrence parsing