Fast, lightweight attribute-style access to tuples.
| Python |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 | from operator import itemgetter as _itemgetter
from keyword import iskeyword as _iskeyword
import sys as _sys
def namedtuple(typename, field_names, verbose=False):
"""Returns a new subclass of tuple with named fields.
>>> Point = namedtuple('Point', 'x y')
>>> Point.__doc__ # docstring for the new class
'Point(x, y)'
>>> p = Point(11, y=22) # instantiate with positional args or keywords
>>> p[0] + p[1] # indexable like a plain tuple
33
>>> x, y = p # unpack like a regular tuple
>>> x, y
(11, 22)
>>> p.x + p.y # fields also accessable by name
33
>>> d = p._asdict() # convert to a dictionary
>>> d['x']
11
>>> Point(**d) # convert from a dictionary
Point(x=11, y=22)
>>> p._replace(x=100) # _replace() is like str.replace() but targets named fields
Point(x=100, y=22)
"""
# Parse and validate the field names. Validation serves two purposes,
# generating informative error messages and preventing template injection attacks.
if isinstance(field_names, basestring):
field_names = field_names.replace(',', ' ').split() # names separated by whitespace and/or commas
field_names = tuple(field_names)
for name in (typename,) + field_names:
if not min(c.isalnum() or c=='_' for c in name):
raise ValueError('Type names and field names can only contain alphanumeric characters and underscores: %r' % name)
if _iskeyword(name):
raise ValueError('Type names and field names cannot be a keyword: %r' % name)
if name[0].isdigit():
raise ValueError('Type names and field names cannot start with a number: %r' % name)
seen_names = set()
for name in field_names:
if name.startswith('_'):
raise ValueError('Field names cannot start with an underscore: %r' % name)
if name in seen_names:
raise ValueError('Encountered duplicate field name: %r' % name)
seen_names.add(name)
# Create and fill-in the class template
numfields = len(field_names)
argtxt = repr(field_names).replace("'", "")[1:-1] # tuple repr without parens or quotes
reprtxt = ', '.join('%s=%%r' % name for name in field_names)
dicttxt = ', '.join('%r: t[%d]' % (name, pos) for pos, name in enumerate(field_names))
template = '''class %(typename)s(tuple):
'%(typename)s(%(argtxt)s)' \n
__slots__ = () \n
_fields = %(field_names)r \n
def __new__(cls, %(argtxt)s):
return tuple.__new__(cls, (%(argtxt)s)) \n
@classmethod
def _make(cls, iterable, new=tuple.__new__, len=len):
'Make a new %(typename)s object from a sequence or iterable'
result = new(cls, iterable)
if len(result) != %(numfields)d:
raise TypeError('Expected %(numfields)d arguments, got %%d' %% len(result))
return result \n
def __repr__(self):
return '%(typename)s(%(reprtxt)s)' %% self \n
def _asdict(t):
'Return a new dict which maps field names to their values'
return {%(dicttxt)s} \n
def _replace(self, **kwds):
'Return a new %(typename)s object replacing specified fields with new values'
result = self._make(map(kwds.pop, %(field_names)r, self))
if kwds:
raise ValueError('Got unexpected field names: %%r' %% kwds.keys())
return result \n\n''' % locals()
for i, name in enumerate(field_names):
template += ' %s = property(itemgetter(%d))\n' % (name, i)
if verbose:
print template
# Execute the template string in a temporary namespace
namespace = dict(itemgetter=_itemgetter)
try:
exec template in namespace
except SyntaxError, e:
raise SyntaxError(e.message + ':\n' + template)
result = namespace[typename]
# For pickling to work, the __module__ variable needs to be set to the frame
# where the named tuple is created. Bypass this step in enviroments where
# sys._getframe is not defined (Jython for example).
if hasattr(_sys, '_getframe'):
result.__module__ = _sys._getframe(1).f_globals['__name__']
return result
if __name__ == '__main__':
# verify that instances can be pickled
from cPickle import loads, dumps
Point = namedtuple('Point', 'x, y', True)
p = Point(x=10, y=20)
assert p == loads(dumps(p))
# test and demonstrate ability to override methods
class Point(namedtuple('Point', 'x y')):
@property
def hypot(self):
return (self.x ** 2 + self.y ** 2) ** 0.5
def __str__(self):
return 'Point: x=%6.3f y=%6.3f hypot=%6.3f' % (self.x, self.y, self.hypot)
for p in Point(3,4), Point(14,5), Point(9./7,6):
print p
class Point(namedtuple('Point', 'x y')):
'Point class with optimized _make() and _replace() without error-checking'
_make = classmethod(tuple.__new__)
def _replace(self, _map=map, **kwds):
return self._make(_map(kwds.get, ('x', 'y'), self))
print Point(11, 22)._replace(x=100)
import doctest
TestResults = namedtuple('TestResults', 'failed attempted')
print TestResults(*doctest.testmod())
|
Discussion
There has long been a need for named access to fields in records stored as tuples. In response, people have crafted many different versions of this recipe. I've combined the best of their approaches with a few ideas of my own. The resulting recipe has been well received, so it was proposed and accepted for inclusion in the collections module for Py2.6.
Docs and examples for the module can be found on the bottom of the page at: http://docs.python.org/dev/library/collections.html#module-collections
The principal features are:
Easy to type/read/modify function signature: named_tuple('Person', 'name age sex height nationality').
C-speed attribute lookup using property() and itemgetter().
Named tuples have no instance dictionary, so their instances take no more space than a regular tuple (for example, casting thousands of sql records to named tuples has zero memory overhead).
Nice docstring is helpful with an editor's tooltips.
Optional keywords in the contructor for readability and to allow arguments to be specified in arbitrary order: Person(name='susan', height=60, nationality='english', sex='f', age=30).
Key/Value style repr for clearer error messages and for usability at the interactive prompt.
Named tuples can be pickled.
Clean error messages for missing or misnamed arguments.
A method similar to str.replace() using a field name. The _replace() method is used instead of slicing for updating fields. For example use t.replace(f=newval) instead of tuple concatenations like t[:2]+newval+t[3:].
A _fields attribute exposes the field names for introspection.
An _asdict() method for converting to an equivalent dictionary.
Recipe runs on Py2.4 or later.
Note, the idea for the __module__ attribute was suggested by Michele Simionato and Louis Riviere. That attribute makes pickling possible, and it lets help() correctly identify the module where the named tuple was defined.
Thanks to Peter Kovac pointing-out deficiencies in the keyword argument checking. Because of his comments, the recipe has evolved to its current exec-style where we get all of Python's high-speed builtin argument checking for free. The new style of building and execing a template made both the __new__ and __repr__ functions faster and cleaner than in previous versions of this recipe.
At the suggestion of Martin Blais and Robin Becker, the field name spec can be either a tuple or a string. When the field names are coming from a CSV header or other automatically generated source, it is simplest to pass in a tuple of fieldnames. When the field names are already known and written-out explicitly, the string form is best because it it easier to type, easier to read, and easier to rearrange when necessary. The string form works especially well with SQL use cases because the string can be cut-and-pasted from the field spec portion of an SQL query.
The function signature with the typename separate from the field names was suggested by Louis Riviere and Jim Jewett.
The idea to use keyword arguments for _replace() was inspired by Issac Morlund's suggestion to be able to do multiple replacements in one pass.
The inspiration for the _make() classmethod came from Robin Becker and Giovanni Bajo who pointed-out an important class of use cases where existing sequences need to be cast to named tuples.
The idea for the _fields attribute was suggested by Robin Becker.
The idea for the _asdict() method was inspired by the thought that any class with name/value pairs is conceptually a mapping. Accordingly, instances of that class should be readily convertible to and from a dict.
Comments
IMO, instead of:
this one seems more pythonic:
But of course, then the NamedTuple function should name the generated class automatically.
Also, if we really require the generated class to have the name we want, the function can have a kwarg to set the class's name.
Constructor should take an iterable. I suggest the generated __new__ should take keyword arguments or an iterable in its only supported positional argument. This matches behaviour with tuple(), list() and other containers.
Syntax proposal. How about:
Pickling error. >>> from cPickle import dumps
added to the doctest fails.
some minor mods. I like this, but :)
1) allow the fields to be defined directly rather than by split
ie
2) allow a constructor __from_iterable__ and a property __field_names__
ie
Possible Mod? I really wanted to be able to take a tuple or list like Giovanni's suggestion, so I built upon Robin's suggestions and came up with this: (I'm not sure if it breaks any semantics or protocols you were trying to preserve)
D'oh! Should have been just:
Little optimisation. template +='\n'.join('\t%s = property(itemgetter(%d))' % (name, i) for i, name in enumerate(field_names))
This may be worth considering if many Nuples and/or many fields are needed.
The readability may suffer for it though ...
Variant that works in Python 2.4 and is Psyco-optimizable. This is a neat recipe. I need it to work in Python 2.4, so I modified __replace__. Also, I use Psyco extensively, and it optimizes list comprehensions well, but has a performance penalty for generators. Hence I changed the generator in __replace__ to a list comprehension.
Psyco has can't optimize anything that looks like a closure, so I added another method __frep__ for "field replace" to be used in __replace__, rather than making it a sub-function of __replace__.
Here are the changed lines for:
Unpickling in different process. Say that you have a pickled Point instance as in the example. Is it possible to unpickle it in a different process that doesn't know about the Point class yet ? In other words, can unpickling somehow automatically run "Point = namedtuple('Point', 'x, y', True)" if there is no global Point class already ?
change verbose=False to out=None. And in the code, instead of
</pre></pre>
Sign in to comment