recordtype 1.4

Description:

recordtype 1.4

NOTE
In Python 3.7 a better solution than recordtype is to use dataclasses.
I have switched active development to my namedlist project. It has a better
implementation that makes it easier to modify the code. It should be
fully compatible with recordtype.
And in general, the attrs project
(https://attrs.readthedocs.io/en/stable/) is an improvement to both
recordtype and namedlist.


Overview
recordtype provides a factory function, named
recordtype.recordtype. It is similar to collections.namedtuple, with
the following differences:

recordtype instances are mutable.
recordtype supports per-field default values.
recordtype supports an optional default value, to be used by all
fields do not have an explicit default value.



Typical usage
You can use recordtype like a mutable namedtuple:
>>> from recordtype import recordtype

>>> Point = recordtype('Point', 'x y')
>>> p = Point(1, 3)
>>> p.x = 2
>>> assert p.x == 2
>>> assert p.y == 3
Or, you can specify a default value for all fields:
>>> Point = recordtype('Point', 'x y', default=3)
>>> p = Point(y=2)
>>> assert p.x == 3
>>> assert p.y == 2
Or, you can specify per-field default values:
>>> Point = recordtype('Point', [('x', 0), ('y', 100)])
>>> p = Point()
>>> assert p.x == 0
>>> assert p.y == 100
You can also specify a the per-field defaults with a mapping, instead
of an interable. Note that this is only useful with an ordered
mapping, such as an OrderedDict:
>>> from collections import OrderedDict
>>> Point = recordtype('Point', OrderedDict((('y', 0),
... ('x', 100))))
>>> p = Point()
>>> assert p.x == 100
>>> assert p.y == 0
The default value will only be used if it is provided and a per-field
default is not used:
>>> Point = recordtype('Point', ['x', ('y', 100)], default=10)
>>> p = Point()
>>> assert p.x == 10
>>> assert p.y == 100
If you use a mapping, the value NO_DEFAULT is convenient to specify
that a field uses the default value:
>>> from recordtype import NO_DEFAULT
>>> Point = recordtype('Point', OrderedDict((('y', NO_DEFAULT),
... ('x', 100))),
... default=5)
>>> p = Point()
>>> assert p.x == 100
>>> assert p.y == 5


Creating types

Specifying Fields
Fields can be specified as in namedtuple: as either a string specifing
the field names, or as a iterable of field names. These two uses are
equivalent:
>>> Point = recordtype('Point', 'x y')
>>> Point = recordtype('Point', ['x', 'y'])
If using a string, commas are first converted to spaces. So these are
equivalent:
>>> Point = recordtype('Point', 'x y')
>>> Point = recordtype('Point', 'x,y')


Specifying Defaults
Per-field defaults can be specified by supplying a 2-tuple (name,
default_value) instead of just a string for the field name. This is
only supported when you specify a list of field names:
>>> Point = recordtype('Point', [('x', 0), ('y', 0)])
>>> p = Point(3)
>>> assert p.x == 3
>>> assert p.y == 0
In addition to, or instead of, these per-field defaults, you can also
specify a default value which is used when no per-field default value
is specified:
>>> Point = recordtype('Point', 'x y z', default=0)
>>> p = Point(y=3)
>>> assert p.x == 0
>>> assert p.y == 3
>>> assert p.z == 0

>>> Point = recordtype('Point', [('x', 0), 'y', ('z', 0)], default=4)
>>> p = Point(z=2)
>>> assert p.x == 0
>>> assert p.y == 4
>>> assert p.z == 2
In addition to supplying the field names as an iterable of 2-tuples,
you can also specify a mapping. The keys will be the field names, and
the values will be the per-field default values. This is most useful
with an OrderedDict, as the order of the fields will then be
deterministic. The module variable NO_DEFAULT can be specified if you
want a field to use the per-type default value instead of specifying
it with a field:
>>> Point = recordtype('Point', OrderedDict((('x', 0),
... ('y', NO_DEFAULT),
... ('z', 0),
... )),
... default=4)
>>> p = Point(z=2)
>>> assert p.x == 0
>>> assert p.y == 4
>>> assert p.z == 2


Writing to values
The objects retured by the factory function are fully writable, unlike
the tuple-derived classes returned by namedtuple:
>>> Point = recordtype('Point', 'x y')
>>> p = Point(1, 2)
>>> p.y = 4
>>> assert p.x == 1
>>> assert p.y == 4


Specifying __slots__
By default, the returned class sets __slots__, which is initialized to
the field names. While this decreases memory usage by eliminating the
instance dict, it also means that you cannot create new instance
members.
To change this behavior, specify use_slots=False when creating the
recordtype:
>>> Point = recordtype('Point', 'x y', use_slots=False)
>>> p = Point(0, 1)
>>> p.z = 2
>>> assert p.x == 0
>>> assert p.y == 1
>>> assert p.z == 2


Additional class members
recordtype classes contain these members:

_asdict(): Returns a dict which maps field names to their
corresponding values.
_source: A string with the pure Python source code used to create
the recordtype class. The source makes the recordtype
self-documenting. It can be printed, executed using exec(), or saved
to a file and imported.
_fields: Tuple of strings listing the field names. Useful for introspection.



Renaming invalid field names
This functionality is identical to namedtuple. If you specify
rename=True, then any invalid field names are changed to _0, _1,
etc. Reasons for a field name to be invalid are:

Zero length strings.
Containing characters other than alphanumerics and underscores.
A conflict with a Python reserved identifier.
Beginning with a digit.
Beginning with an underscore.
Using the same field name more than once.

For example:
>>> Point = recordtype('Point', 'x x for', rename=True)
>>> assert Point._fields == ('x', '_1', '_2')


Mutable default values
Be aware of creating mutable default values. Due to the way Python
handles default values, each instance of a recordtype will share the
default. This is especially problematic with default values that are
lists. For example:
>>> A = recordtype('A', [('x', [])])
>>> a = A()
>>> a.x.append(4)
>>> b = A()
>>> assert b.x == [4]
This is probably not the desired behavior.



Creating and using instances
Because the type returned by recordtype is a normal Python class, you
create instances as you would with any Python class.


Change log

1.4 2022-09-22 Eric V. Smith

For Python 3.10 compatibility, try to import collections.abc.Mapping.
Reformat with black.



1.3 2018-08-03 Eric V. Smith

Python 3 support (thanks Jakob Stasiak).
Reformat with black.



1.2 <unreleased> Eric V. Smith

Switch README.txt to support doctest.
Add tests for mutable default values.
Add warning for mutable default values.



1.1 2011-11-14 Eric V. Smith

No API or code changes.
Fixed project URL in setup.py.
Fixed license description in setup.py.



1.0 2011-10-23 Eric V. Smith

Stabilize API, move to version 1.0.
Added “python setup.py test” support to run unittests.
Improve documentation examples.
Renamed “default_default” to just “default”.
Expose “NO_DEFAULT”.
Support mapping objects in addition to strings and lists for field_names.
Add tests for iterables, not just lists, for field_names.



0.2 2011-10-13 Eric V. Smith

Fix a typo in the documentation, no code changes.



0.1 2011-10-12 Eric V. Smith

Initial release.