Skip to content

Latest commit

 

History

History
229 lines (142 loc) · 7.76 KB

README.markdown

File metadata and controls

229 lines (142 loc) · 7.76 KB
title
sanetime: a sane date/time python interface

#sanetime

A sane date/time python interface

better epoch time, timezones, and deltas, django support as well

##intro

sanetime was written to DRY up all the common date/time manipulations we all do constantly in our code while offering the most simple, versatile, and intuitive client possible.

We've all learned that the only sane way to store times is using epoch time. (You have, haven't you?) Unfortunately, manipulating epoch time and timezones with the standard python toolset requires getting up to speed on a managerie of python modules and concepts: datetime, date, time, calendar, pytz, dateutils, timedelta, time tuples, localize, normalize.

sanetime seeks to bring a more sanity to the manipulations of epoch time, timezone, time delta, and time generally.

>>> from sanetime import time,delta   # a tiny taste

>>> time('2012-05-01 22:31',tz='America/New_York').millis
1335925860000

>>> str(time(tz='Europe/London'))   # now in London
'2012-05-29 15:28:05.178741 +Europe/London'

>>> (time(2012,6,1) - time('2012-05-01')).hours
744

>>> (time() + delta(h=12)).s    # epoch seconds 12 hours from now
1338344977

##time

###concept

The time class represents a moment in time, internally stored as microseconds since epoch.
A time object also has an associated timezone (UTC by default), however the timezone will never be considered during hashing, comparison or equality checks, i.e.
A moment in time experienced in America/New_York is equal to the same moment in time experienced in Europe/Dublin.

###methods

####construction

You can construct a sanetime object from epoch times, datetimes, date/time parts, or from a parseable string.

Epoch microseconds are assumed when no keyword is given. Intuitive aliases exists for kwargs, be as terse or verbose as you want (us = micros = epoch_micros = epoch_microseconds):

>>> time(1338508800000000)
SaneTime(1338508800000000,<UTC>)

>>> time(micros=1338508800000000)
SaneTime(1338508800000000,<UTC>)

>>> time(millis=1338508800000)
SaneTime(1338508800000000,<UTC>)

>>> time(seconds=1338508800)
SaneTime(1338508800000000,<UTC>)

>>> time(minutes=22308480, tz='America/New_York')
SaneTime(1338508800000000,<DstTzInfo 'America/New_York' EST-1 day, 19:00:00 STD>)

If you have the calendar parameters, then construct just as you would a datetime:

>>> time(2012,1,1)
SaneTime(1325376000000000,<UTC>)

>>> time(2012,1,1,12,30,1)
SaneTime(1325421001000000,<UTC>)

>>> time(2012,1,1,12,30,1,1, tz='America/New_York')
SaneTime(1325421001000001,<DstTzInfo 'America/New_York' EST-1 day, 19:00:00 STD>)

If you already have a datetime object, just construct from that:

>>> dt = datetime(2012,1,1)
>>> time(dt)
SaneTime(1325376000000000,<UTC>)

Or construct from a parsable string:

>>> time('January 1st, 2012 12:30:01pm')
SaneTime(1325421001000000,<UTC>)

>>> time('January 1st, 2012 12:30:01pm', tz='America/New_York')
SaneTime(1325421001000000,<DstTzInfo 'America/New_York' EST-1 day, 19:00:00 STD>)

####arithmetic

Adding any int/long assumes it to be in microseconds. You can also add any :ref:delta:

>>> time(2012,1,1) + 5
SaneTime(1325376000000005,<UTC>)

>>> time(2012,1,1) + delta(hours=5)
SaneTime(1325394000000000,<UTC>)

Subtracting two sanetimes produces a :ref:delta:

>>> time() - time(2012,1,1)  # time since new year
SaneDelta(15131339063956)

>>> abs(time() - time()).micros  # microseconds to construct a time
30

####conversion

You can easily convert to a timezone-aware datetime or to a naive datetime. They are accessed as properties.

>>> time(2012,1,1,tz='America/Los_Angeles').datetime
datetime.datetime(2012, 1, 1, 0, 0, tzinfo=<DstTzInfo 'America/Los_Angeles' PST-1 day, 16:00:00 STD>)

>>> time(2012,1,1,tz='America/Los_Angeles').naive_datetime
datetime.datetime(2012, 1, 1, 0, 0)

There are other convenience datetime timezone conversions as well.

>>> time(2012,1,1,tz='America/Los_Angeles').utc_datetime
datetime.datetime(2012, 1, 1, 8, 0, tzinfo=<UTC>)

>>> time(2012,1,1,tz='America/Los_Angeles').utc_naive_datetime
datetime.datetime(2012, 1, 1, 8, 0)

>>> time(2012,1,1,tz='America/Los_Angeles').ny_datetime
datetime.datetime(2012, 1, 1, 3, 0, tzinfo=<DstTzInfo 'America/New_York' EST-1 day, 19:00:00 STD>)

>>> time(2012,1,1,tz='America/Los_Angeles').ny_naive_datetime
datetime.datetime(2012, 1, 1, 3, 0)

To epoch times:

>>> time(2012,1,1).minutes
22089600

>>> time(2012,1,1).seconds
1325376000

>>> time(2012,1,1).millis
1325376000000

>>> time(2012,1,1).micros
1325376000000000

long and int conversion just bring back the epoch microseconds

>>> int(time(2012,1,1))
1325376000000000

>>> long(time(2012,1,1))
1325376000000000L

#####date/time parts

You can get at any of the date parts just as you might with datetime properties. Be careful-- these properties are all singular. Do not confuse with the plural epoch possiblities of the previous section. (this ambiguity will be fixed in future versions)

>>> time().year
2012
>>> time().month
6
>>> time().day
24
>>> time().hour
3
>>> time().minute
42
>>> time().second
12
>>> time().micro
664819

###tztime The tztime class is exactly like the time object, except that timezone does factor into equality.
A moment in tztime experienced in America/New_York is not the same as the same tztime moment experienced in Europe/Dublin.

###delta The delta class represents a period of time, and provides easy access to all the different ways you might slice and dice this: micros, millis, seconds, minutes, hours, mean_days, mean_weeks, mean_months, mean_years.
There are also many different flavors of these: rounded, floored, floated, positional, rounded_positional. There is no attempt made in delta yet to be calendar aware (hence the 'mean' prefixes in some cases).

###span The span class represents a window of time ranging from one specific moment in time to another specific moment in time.
You can think of it as a start time with a delta, or as a start time and a stop time.

###django A django model field is also provided: SaneTimeField, that makes it super simple to store a sanetime. They honor the auto_add and auto_add_now features to easily turn your sanetimes into updated_at or created_at fields. And they even work with south out of the box.

###NOTE This documentation is still a work in progress.
Lots of good docs are now in place here, but there are still features lying in the code waiting documentation.
As always, the code should be treated as the ultimate source of truth.
If you haven't found what you're looking for you, you may very well find it implemented in the code.
Please feel free to help us bring it to the surface in the docs.

###design principles

  • simple: simplify usecases to single method/property
  • intuitive: easy to remember methods/properties, with guessable aliases - be as verbose (and communicative) or as terse (and efficient) as you want to be. for example t = time(); t.ms == t.millis == t.milliseconds
  • properties whenever sensible: properties are especially useful for django, cuz you can use them directly in templates without having to stage them first in the views.

###FAQ Why is everything stored internally as microseconds?

Python's datetime gives us access to microseconds, and since milliseconds would already have us cross the 32bit integer boundary, we might as well capture everything we can and take on microseconds as well. There are plenty of helpers on the time, tztime, and delta that make using epoch seconds or milis just as easy as using micros.