Spans 0.3.0

Continuous set support for Python

Spans

Spans is a pure Python implementation of PostgreSQL’s range types [1]. Range types are conveinent when working with intervals of any kind. Every time you’ve found yourself working with date_start and date_end, an interval may have been what you were actually looking for.

Spans has successfully been used in production since its first release 30th August, 2013.

Here is an example on how to use ranges to determine if something happened in the 90s.

>>> from spans import daterange
>>> from datetime import date
>>> the90s = daterange(date(1990, 1, 1), date(2000, 1, 1))
>>> date(1996, 12, 4) in the90s
True
>>> date(2000, 1, 1) in the90s
False
>>> the90s.union(daterange(date(2000, 1, 1), date(2010, 1, 1)))
daterange([datetime.date(1990, 1, 1), datetime.date(2010, 1, 1))))

If you are making a booking application for a bed and breakfast hotel and want to ensure no room gets double booked:

from collections import defaultdict
from datetime import date
from spans import daterange

# Add a booking from 2013-01-14 through 2013-01-15
bookings = defaultdict(list, {
    1 : [daterange(date(2013, 1, 14), date(2013, 1, 16))]
}

def is_valid_booking(bookings, room, new_booking):
    return not any(booking.overlap(new_booking for booking in bookings[room])

print is_valid_booking(
    bookings, 1, daterange(date(2013, 1, 14), date(2013, 1, 18))) # False
print is_valid_booking(
    bookings, 1, daterange(date(2013, 1, 16), date(2013, 1, 18))) # True

The library supports ranges and sets of ranges. A range has no discontinuities between its endpoints. For some applications this is a requirement and hence the rangeset type exists.

Apart from the above mentioned overlap operation; ranges support union, difference, intersection, contains, startswith, endswith, left_of and right_of.

Built-in ranges:

• intrange

• floatrange

• strrangerange - For unicode strings

• daterange

• datetimerange

• timedeltarange

For each one of the range types a rangeset type exists as well:

• intrangeset

• floatrangeset

• strrangerangeset

• daterangeset

• datetimerangeset

• timedeltarangeset

Motivation

For a recent project of mine I started using PostgreSQL’s tsrange type and needed an equivalent in Python. These range types attempt to mimick PostgreSQL’s behavior in every way. Deviating from it is considered as a bug and should be reported.

Installation

Spans exists on PyPI.

$ pip install Spans

Documentation

Documentation is hosted on Read the Docs.

Use with Psycopg2

To use these range types with Psycopg2 the PsycoSpans library exists [2].

Custom range types

Using your own types for ranges are easy, just extend a base class and you’re good to go:

from spans.types import range_, discreterange
from spans.settypes import rangeset, discreterangeset

class intrange(discreterange):
    __slots__ = ()
    type = int
    step = 1

class intrangeset(discreterangeset):
    __slots__ = ()
    type = intrange

class floatrange(range_):
    __slots__ = ()
    type = float

class floatrangeset(rangeset):
    __slots__ = ()
    type = floatrange

For a deeper set of examples please refer to types.py and settypes.py.

[1]

http://www.postgresql.org/docs/9.2/static/rangetypes.html

[2]

https://www.github.com/runfalk/psycospans

Changelog

Version are structured like the following: <major>.<minor>.<bugfix>. The first 0.1 release does not properly adhere to this. Unless explicitly stated, changes are made by Andreas Runfalk.

Version 0.3.0

Released on 26th August, 2016

• Added documentation for __iter__()

• Fixed intersection of multiple range sets not working correctly (bug #3)

• Fixed iteration of rangeset returning an empty range when rangeset is empty (bug #4)

Warning

This change is backwards incompatible to code that expect rangesets to always return at least one set when iterating.

Version 0.2.1

Released on 27th June, 2016

• Fixed rangeset not returning NotImplemented when comparing to classes that are not sub classes of rangeset, pull request #2 (Michael Krahe)

• Updated license in setup.py to follow recommendations by PyPA

Version 0.2.0

Released on 22nd December, 2015

• Added __len__() to range sets (Michael Krahe)

• Added contains() to range sets (Michael Krahe)

• Added Sphinx style doc strings to all methods

• Added proper Sphinx documentation

• Added unit tests for uncovered parts, mostly error checking

• Added wheel to PyPI along with source distribution

• Fixed a potential bug where comparing ranges of different types would result in an infinite loop

• Changed meta class implementation for range sets to allow more mixins for custom range sets

Version 0.1.4

Released on 15th May, 2015

• Added .last property to discreterange

• Added from_date() helper to daterange

• Added more unit tests

• Improved pickle implementation

• Made type checking more strict for date ranges to prevent datetime from being allowed in daterange

Version 0.1.3

Released on 27th February, 2015

• Added offset() to some range types

• Added offset() to some range set types

• Added sanity checks to range boundaries

• Fixed incorrect __slots__ usage, resulting in __slots__ not being used on most ranges

• Fixed pickling of ranges and range sets

• Simplified creation of new rangesets, by the use of the meta class metarangeset

Version 0.1.2

Released on 13th June, 2014

• Fix for inproper version detection on Ubuntu’s bundled Python interpreter

Version 0.1.1

Released on 12th June, 2014

• Readme fixes

• Syntax highlighting for PyPI page

Version 0.1.0

Released on 30th August, 2013

• Initial release