Skip to content

Latest commit

 

History

History
119 lines (90 loc) · 5.35 KB

README.md

File metadata and controls

119 lines (90 loc) · 5.35 KB
Raw

timefhuman

PyPi Downloads per Month Coverage Status Build Status

Convert human-readable, date-like strings written in natural language to Python objects. Describe specific datetimes or ranges of datetimes. Supports Python3+. You can skip installation and try timefhuman directly, using the pytwiddle demo →

To start, describe days of the week or times of day in the vernacular.

>>> from timefhuman import timefhuman
>>> timefhuman('upcoming Monday noon')
datetime.datetime(2018, 8, 6, 12, 0)

Use any human-readable format with a time range, choices of times, or choices of time ranges.

>>> timefhuman('7/17 3-4 PM')
(datetime.datetime(2018, 7, 17, 15, 0), datetime.datetime(2018, 7, 17, 16, 0))
>>> timefhuman('7/17 3 p.m. - 4 p.m.')
(datetime.datetime(2018, 7, 17, 15, 30), datetime.datetime(2018, 7, 17, 16, 0))
>>> timefhuman('Monday 3 pm or Tu noon')
[datetime.datetime(2018, 8, 6, 15, 0), datetime.datetime(2018, 8, 7, 12, 0)]
>>> timefhuman('7/17 4 or 5 PM')
[datetime.datetime(2018, 7, 17, 16, 0), datetime.datetime(2018, 7, 17, 17, 0)]
>>> timefhuman('7/17 4-5 or 5-6 PM')
[(datetime.datetime(2018, 7, 17, 16, 0), datetime.datetime(2018, 7, 17, 17, 0)),
 (datetime.datetime(2018, 7, 17, 17, 0), datetime.datetime(2018, 7, 17, 18, 0))]

Parse lists of dates and times with more complex relationships.

>>> timefhuman('7/17, 7/18, 7/19 at 2')
[datetime.datetime(2018, 7, 17, 2, 0), datetime.datetime(2018, 7, 18, 2, 0), datetime.datetime(2018, 7, 19, 2, 0)]
>>> timefhuman('2 PM on 7/17 or 7/19')
[datetime.datetime(2018, 7, 17, 14, 0), datetime.datetime(2018, 7, 19, 14, 0)]

Use the vernacular to describe ranges or days.

>>> timefhuman('noon next week')  # coming soon

>>> timefhuman('today or tomorrow noon')  # when run on August 4, 2018
[datetime.datetime(2018, 8, 4, 12, 0), datetime.datetime(2018, 8, 5, 12, 0)]

Installation

Install with pip using

pip install timefhuman

Optionally, clone the repository and run python setup.py install.

You can also try timefhuman without a local installation using the twiddle.

Usage

Use the now kwarg to use different default values for the parser.

>>> import datetime
>>> now = datetime.datetime(2018, 8, 4, 0, 0)
>>> timefhuman('upcoming Monday noon', now=now)
datetime.datetime(2018, 8, 6, 12, 0)

Use a variety of different formats, even with days of the week, months, and times with everyday speech. These are structured formats. dateparser supports structured formats across languages, customs etc.

>>> from timefhuman import timefhuman
>>> now = datetime.datetime(year=2018, month=7, day=7)
>>> timefhuman('July 17, 2018 at 3p.m.')
datetime.datetime(2018, 7, 17, 15, 0)
>>> timefhuman('July 17, 2018 3 p.m.')
datetime.datetime(2018, 7, 17, 15, 0)
>>> timefhuman('3PM on July 17', now=now)
datetime.datetime(2018, 7, 17, 15, 0)
>>> timefhuman('July 17 at 3')
datetime.datetime(2018, 7, 17, 3, 0)
>>> timefhuman('7/17/18 3:00 p.m.')
datetime.datetime(2018, 7, 17, 15, 0)

Why

dateparser is the current king of human-readable-date parsing--it supports most common structured dates by trying each one sequentially (see code). However, this isn't optimal for understanding natural language:

>>> import dateparser
>>> dateparser.parse("7/7/18 3 p.m.")  # yay!
datetime.datetime(2018, 7, 7, 15, 0)
>>> dateparser.parse("7/7/18 at 3")  # :(
>>> dateparser.parse("7/17 12 PM")  # yay!
datetime.datetime(2018, 7, 7, 12, 0)
>>> dateparser.parse("7/17/18 noon")  # :(
>>> dateparser.parse("7/18 3-4 p.m.")  # :((((( Parsed July 18 3-4 p.m. as July 3 4 p.m.
datetime.datetime(2018, 7, 3, 16, 0)

To remedy this, we can replace "noon" with "12 p.m.", "next Monday" with "7/17/18", "Tu" with "Tuesday" etc. and pass the cleaned string to dateparser. However, consider the number of ways we can say "next Monday at 12 p.m.". Ignoring synonyms, we have a number of different grammars to express this:

  • 12 p.m. on Monday
  • first Monday of August 12 p.m.
  • next week Monday noon

This issue compounds when you consider listing noontimes for several different days.

  • first half of next week at noon
  • 12 p.m. on Monday Tuesday or Wednesday
  • early next week midday

The permutations--even the possible combinations--are endless. Instead of enumerating each permutation, timefhuman extracts tokens: "anytime" modifies the type from 'date' to 'range', "next week" shifts the range by 7 days, "p.m." means the string right before is a time or a time range etc. Each set of tokens is then combined to produce datetimes, datetime ranges, or datetime lists. This then allows timefhuman to handle any permutation of these modifiers. Said another way: timefhuman aims to parse unstructured dates, written in natural language.