Datetoken

Features

TL;DR: This package allows you to store complex relative dates in string tokens.

• To define the initial/starting point in time (typically written as now), to work with dates in the past or in the future. Ideal to perform date simulations.

• Time zone configuration. Additionally to the starting point in time, TZs can be provided to as to somehow abstract the user away from localizing datetimes objects in their apps.

  As a disclaimer, if the custom datetime specified as the starting point (now's value) is tz-unaware or naive, it will be treated as an UTC one. Default now's value also fallsback to datetime.datetime.utcnow, localized to UTC.

  Now, if a time zone is specified, the nows value will be coerced to that TZ prior to applying both snap and modifier expressions. This is handy to quickly resolve tokens given any point in time (either naive or aware), a time zone and the datetoken itself.

Motivation

Have you ever needed to make an application where dates needed to be represented in a relative fashion, like background periodic tasks, datetime range pickers... in a compact and stringified format? This library enables you to persist these string tokens during the lifetime of a process or even longer, since calculations are performed in the moment of evaluation. Theses tokens are also useful when caching URLs as replacement of timestamps, which would break caching given their mutability nature.

Some common examples of relative tokens:

As you may have noticed, tokens follow a pattern:

• The word now. It means the point in the future timeline when tokens are parsed to their datetime form.

• Optionally, modifiers to add and/or subtract the future value of now can be used. Unsurprisingly, additions are set via +, while - mean subtractions. These modifiers can be chained as many times as needed. E.g: now-1M+3d+2h. Along with the arithmetical sign and the amount, the unit of time the amount refers to must be specified. Currently, the supported units are:

  • s seconds
  • m minutes
  • h hours
  • d days
  • w weeks
  • M months

• Optionally, there exist two extra modifiers to snap dates to the start or the end of any given snapshot unit. Those are:

  • / Snap the date to the start of the snapshot unit.
  • @ Snap the date to the end of the snapshot unit.

  Snapshot units are the same as arithmetical modifiers, plus the following ones:

  • bw, business week
  • mon, Monday
  • tue, Tuesday
  • wed, Wednesday
  • thu, Thursday
  • fri, Friday
  • sat, Saturday
  • sun, Sunday

  With this, we achieve a simple way to define canonical relative date ranges, such as Today or Last month. As an example of the later:

  • String representation: now-1M/M, now-1M@M
  • Being today 15 Jan 2018, the result range should be: 2018-01-01 00:00:00 / 2018-01-31 23:59:59

Installing

Install and update via either pipenv or pip

pipenv install datetoken

or

pip install datetoken

A glance into the API

You can use either the evaluator subpackage or the utils one for quicker access to simpler/common usages.

• datetoken.evaluator.eval_datetoken
  • Arguments:
    • token: {string} E.g: now-w/w+2d+8h
    • kwargs:
      • at: {datetime.datetime} custom starting point
      • tz: {str|pytz.timezone} custom time zone
  • Return:
    • datetoken.objects.Token. Model for tokens. Provides meta information such as AST nodes, and whether the token is snapped or has modifiers applied
• datetoken.evaluator.Datetoken Facade to build tokens on the fly. Supports fluent programming too.
• datetoken.utils.token_to_date:
  • Arguments:
    • token: {string} E.g: now-w/w+2d+8h
    • kwargs:
      • at: {datetime.datetime} custom starting point
      • tz: {str|pytz.timezone} custom time zone
  • Return:
    • datetime.datetime. Datetime object with the result of applying token modifiers. Always returns aware tz objects.
• datetoken.utils.token_to_utc_date: Same as token_to_date but coercing the result to UTC.

Examples

Most probably you will be dealing with simple presets such as yesterday or the last 24 hours.

>>> from datetoken.utils import token_to_date
>>> from datetime import datetime
>>> print(datetime.utcnow())
2018-10-18 14:08:47
>>> token_to_date('now-d/d')  # Start of yesterday
2018-10-17 00:00:00
>>> token_to_date('now-d@d')  # End of yesterday
2018-10-17 23:59:59

However, more complex configurations are also supported so as to provide the flexibility advanced users may need.

>>> from datetoken.utils import token_to_date
>>> print(datetime.utcnow())
2018-10-18 16:34:29+02:00
>>> token_to_date('now-M/M+w/bw')  # Starting of first business week of previous
                                   # month
2018-09-03 00:00:00

Fluent programming is also supported:

>>> import pytz
>>> from datetime import datetime
>>> from datetoken.evaluator import Datetoken
>>> then = datetime(2019, 1, 26, 12, 24, 23, tzinfo=pytz.UTC)
>>> Datetoken().at(then).on('Europe/Madrid').for_token('now/d').to_date()
datetime(2019, 1, 26, 0, 0, 0, tzinfo="<DstTzInfo 'Europe/Madrid' CET+1:00:00 STD>")

Retrieving dates in UTC is implemented too:

>>> import pytz
>>> from datetime import datetime
>>> from datetoken.evaluator import Datetoken
>>> then = datetime(2019, 1, 26, 12, 24, 23, tzinfo=pytz.UTC)
>>> Datetoken().at(then).on('Europe/Madrid').for_token('now/d').to_utc_date()
datetime(2019, 1, 25, 23, 0, 0, tzinfo="<UTC>")

>>> from datetoken.utils import token_to_utc_date
>>> then_in_madrid = datetime(2019, 1, 26, 12, 24, 23, tzinfo=pytz.timezone('Europe/Madrid'))
>>> token_to_utc_date('now/d', at=then_in_madrid)
datetime(2019, 1, 25, 23, 0, 0, tzinfo="<UTC>")

If you thought fluent programming is no longer fashionable:

...
>>> then = datetime(2019, 1, 26, 12, 24, 23, tzinfo=pytz.UTC)
>>> token = Datetoken(at=then, tz='Europe/Madrid', token='now/d')
>>> token.to_date()
datetime(2019, 1, 26, 0, 0, 0, tzinfo="<DstTzInfo 'Europe/Madrid' CET+1:00:00 STD>")

Issues

• Business week snapshots might not be reliable in timezones where weeks start in days other than Monday or week duration lasts fewer or greater than five days