freeipa/ipalib/frontend.py
Pavel Zuna dc23be6878 Make get_dn parameter list more generic. Fix Attribute name regex.
The old name regex made it impossible to have Attribute instances with
names composed of more than two words separated by underscores.
2009-06-10 11:53:14 -04:00

1064 lines
36 KiB
Python

# Authors:
# Jason Gerard DeRose <jderose@redhat.com>
#
# Copyright (C) 2008 Red Hat
# see file 'COPYING' for use and warranty information
#
# This program is free software; you can redistribute it and/or
# modify it under the terms of the GNU General Public License as
# published by the Free Software Foundation; version 2 only
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
"""
Base classes for all front-end plugins.
"""
import re
import inspect
from base import lock, check_name, NameSpace
from plugable import Plugin
from parameters import create_param, parse_param_spec, Param, Str, Flag, Password
from util import make_repr
from errors import ZeroArgumentError, MaxArgumentError, OverlapError, RequiresRoot
from constants import TYPE_ERROR
RULE_FLAG = 'validation_rule'
def rule(obj):
assert not hasattr(obj, RULE_FLAG)
setattr(obj, RULE_FLAG, True)
return obj
def is_rule(obj):
return callable(obj) and getattr(obj, RULE_FLAG, False) is True
class HasParam(Plugin):
"""
Base class for plugins that have `Param` `NameSpace` attributes.
Subclasses of `HasParam` will on one or more attributes store `NameSpace`
instances containing zero or more `Param` instances. These parameters might
describe, for example, the arguments and options a command takes, or the
attributes an LDAP entry can include, or whatever else the subclass sees
fit.
Although the interface a subclass must implement is very simple, it must
conform to a specific naming convention: if you want a namespace
``SubClass.foo``, you must define a ``Subclass.takes_foo`` attribute and a
``SubCLass.get_foo()`` method, and you may optionally define a
``SubClass.check_foo()`` method.
A quick big-picture example
===========================
Say you want the ``options`` instance attribute on your subclass to be a
`Param` `NameSpace`... then according to the enforced naming convention,
your subclass must define a ``takes_options`` attribute and a
``get_options()`` method. For example:
>>> from ipalib import Str, Int
>>> class Example(HasParam):
...
... options = None # This will be replaced with your namespace
...
... takes_options = (Str('one'), Int('two'))
...
... def get_options(self):
... return self._get_param_iterable('options')
...
>>> eg = Example()
The ``Example.takes_options`` attribute is a ``tuple`` defining the
parameters you want your ``Example.options`` namespace to contain. Your
``Example.takes_options`` attribute will be accessed via
`HasParam._get_param_iterable()`, which, among other things, enforces the
``('takes_' + name)`` naming convention. For example:
>>> eg._get_param_iterable('options')
(Str('one'), Int('two'))
The ``Example.get_options()`` method simply returns
``Example.takes_options`` by calling `HasParam._get_param_iterable()`. Your
``Example.get_options()`` method will be called via
`HasParam._filter_param_by_context()`, which, among other things, enforces
the ``('get_' + name)`` naming convention. For example:
>>> list(eg._filter_param_by_context('options'))
[Str('one'), Int('two')]
At this point, the ``eg.options`` instance attribute is still ``None``:
>>> eg.options is None
True
`HasParam._create_param_namespace()` will create the ``eg.options``
namespace from the parameters yielded by
`HasParam._filter_param_by_context()`. For example:
>>> eg._create_param_namespace('options')
>>> eg.options
NameSpace(<2 members>, sort=False)
>>> list(eg.options) # Like dict.__iter__()
['one', 'two']
Your subclass can optionally define a ``check_options()`` method to perform
sanity checks. If it exists, the ``check_options()`` method is called by
`HasParam._create_param_namespace()` with a single value, the `NameSpace`
instance it created. For example:
>>> class Example2(Example):
...
... def check_options(self, namespace):
... for param in namespace(): # Like dict.itervalues()
... if param.name == 'three':
... raise ValueError("I dislike the param 'three'")
... print ' ** Looks good! **' # Note output below
...
>>> eg = Example2()
>>> eg._create_param_namespace('options')
** Looks good! **
>>> eg.options
NameSpace(<2 members>, sort=False)
However, if we subclass again and add a `Param` named ``'three'``:
>>> class Example3(Example2):
...
... takes_options = (Str('one'), Int('two'), Str('three'))
...
>>> eg = Example3()
>>> eg._create_param_namespace('options')
Traceback (most recent call last):
...
ValueError: I dislike the param 'three'
>>> eg.options is None # eg.options was not set
True
The Devil and the details
=========================
In the above example, ``takes_options`` is a ``tuple``, but it can also be
a param spec (see `create_param()`), or a callable that returns an iterable
containing one or more param spec. Regardless of how ``takes_options`` is
defined, `HasParam._get_param_iterable()` will return a uniform iterable,
conveniently hiding the details.
The above example uses the simplest ``get_options()`` method possible, but
you could instead implement a ``get_options()`` method that would, for
example, produce (or withhold) certain parameters based on the whether
certain plugins are loaded.
Think of ``takes_options`` as declarative, a simple definition of *what*
parameters should be included in the namespace. You should only implement
a ``takes_options()`` method if a `Param` must reference attributes on your
plugin instance (for example, for validation rules); you should not use a
``takes_options()`` method to filter the parameters or add any other
procedural behaviour.
On the other hand, think of the ``get_options()`` method as imperative, a
procedure for *how* the parameters should be created and filtered. In the
example above the *how* just returns the *what* unchanged, but arbitrary
logic can be implemented in the ``get_options()`` method. For example, you
might filter certain parameters from ``takes_options`` base on some
criteria, or you might insert additional parameters provided by other
plugins.
The typical use case for using ``get_options()`` this way is to procedurally
generate the arguments and options for all the CRUD commands operating on a
specific LDAP object: the `Object` plugin defines the possible LDAP entry
attributes (as `Param`), and then the CRUD commands intelligently build
their ``args`` and ``options`` namespaces based on which attribute is the
primary key. In this way new LDAP attributes (aka parameters) can be added
to the single point of definition (the `Object` plugin), and all the
corresponding CRUD commands pick up these new parameters without requiring
modification. For an example of how this is done, see the
`ipalib.crud.Create` base class.
However, there is one type of filtering you should not implement in your
``get_options()`` method, because it's already provided at a higher level:
you should not filter parameters based on the value of ``api.env.context``
nor (preferably) on any values in ``api.env``.
`HasParam._filter_param_by_context()` already does this by calling
`Param.use_in_context()` for each parameter. Although the base
`Param.use_in_context()` implementation makes a decision solely on the value
of ``api.env.context``, subclasses can override this with implementations
that consider arbitrary ``api.env`` values.
"""
def _get_param_iterable(self, name):
"""
Return an iterable of params defined by the attribute named ``name``.
A sequence of params can be defined one of three ways: as a ``tuple``;
as a callable that returns an iterable; or as a param spec (a `Param` or
``str`` instance). This method returns a uniform iterable regardless of
how the param sequence was defined.
For example, when defined with a tuple:
>>> class ByTuple(HasParam):
... takes_args = (Param('foo'), Param('bar'))
...
>>> by_tuple = ByTuple()
>>> list(by_tuple._get_param_iterable('args'))
[Param('foo'), Param('bar')]
Or you can define your param sequence with a callable when you need to
reference attributes on your plugin instance (for validation rules,
etc.). For example:
>>> class ByCallable(HasParam):
... def takes_args(self):
... yield Param('foo', self.validate_foo)
... yield Param('bar', self.validate_bar)
...
... def validate_foo(self, _, value, **kw):
... if value != 'Foo':
... return _("must be 'Foo'")
...
... def validate_bar(self, _, value, **kw):
... if value != 'Bar':
... return _("must be 'Bar'")
...
>>> by_callable = ByCallable()
>>> list(by_callable._get_param_iterable('args'))
[Param('foo', validate_foo), Param('bar', validate_bar)]
Lastly, as a convenience for when a param sequence contains a single
param, your defining attribute may a param spec (either a `Param`
or an ``str`` instance). For example:
>>> class BySpec(HasParam):
... takes_args = Param('foo')
... takes_options = 'bar?'
...
>>> by_spec = BySpec()
>>> list(by_spec._get_param_iterable('args'))
[Param('foo')]
>>> list(by_spec._get_param_iterable('options'))
['bar?']
For information on how an ``str`` param spec is interpreted, see the
`create_param()` and `parse_param_spec()` functions in the
`ipalib.parameters` module.
Also see `HasParam._filter_param_by_context()`.
"""
takes_name = 'takes_' + name
takes = getattr(self, takes_name, None)
if type(takes) is tuple:
return takes
if isinstance(takes, (Param, str)):
return (takes,)
if callable(takes):
return takes()
if takes is None:
return tuple()
raise TypeError(
'%s.%s must be a tuple, callable, or spec; got %r' % (
self.name, takes_name, takes
)
)
def _filter_param_by_context(self, name, env=None):
"""
Filter params on attribute named ``name`` by environment ``env``.
For example:
>>> from ipalib.config import Env
>>> class Example(HasParam):
...
... takes_args = (
... Str('foo_only', include=['foo']),
... Str('not_bar', exclude=['bar']),
... 'both',
... )
...
... def get_args(self):
... return self._get_param_iterable('args')
...
...
>>> eg = Example()
>>> foo = Env(context='foo')
>>> bar = Env(context='bar')
>>> another = Env(context='another')
>>> (foo.context, bar.context, another.context)
('foo', 'bar', 'another')
>>> list(eg._filter_param_by_context('args', foo))
[Str('foo_only', include=['foo']), Str('not_bar', exclude=['bar']), Str('both')]
>>> list(eg._filter_param_by_context('args', bar))
[Str('both')]
>>> list(eg._filter_param_by_context('args', another))
[Str('not_bar', exclude=['bar']), Str('both')]
"""
env = getattr(self, 'env', env)
get_name = 'get_' + name
if not hasattr(self, get_name):
raise NotImplementedError(
'%s.%s()' % (self.name, get_name)
)
get = getattr(self, get_name)
if not callable(get):
raise TypeError(
'%s.%s must be a callable; got %r' % (self.name, get_name, get)
)
for spec in get():
param = create_param(spec)
if env is None or param.use_in_context(env):
yield param
def _create_param_namespace(self, name, env=None):
namespace = NameSpace(
self._filter_param_by_context(name, env),
sort=False
)
check = getattr(self, 'check_' + name, None)
if callable(check):
check(namespace)
setattr(self, name, namespace)
class Command(HasParam):
"""
A public IPA atomic operation.
All plugins that subclass from `Command` will be automatically available
as a CLI command and as an XML-RPC method.
Plugins that subclass from Command are registered in the ``api.Command``
namespace. For example:
>>> from ipalib import create_api
>>> api = create_api()
>>> class my_command(Command):
... pass
...
>>> api.register(my_command)
>>> api.finalize()
>>> list(api.Command)
['my_command']
>>> api.Command.my_command # doctest:+ELLIPSIS
PluginProxy(Command, ...my_command())
"""
__public__ = frozenset((
'get_default',
'convert',
'normalize',
'validate',
'execute',
'__call__',
'args',
'options',
'params',
'params_2_args_options',
'args_options_2_params',
'output_for_cli',
))
takes_options = tuple()
takes_args = tuple()
args = None
options = None
params = None
output_for_cli = None
def __call__(self, *args, **options):
"""
Perform validation and then execute the command.
If not in a server context, the call will be forwarded over
XML-RPC and the executed an the nearest IPA server.
"""
params = self.args_options_2_params(*args, **options)
self.debug(
'raw: %s(%s)', self.name, ', '.join(self._repr_iter(**params))
)
params = self.normalize(**params)
params = self.convert(**params)
params.update(self.get_default(**params))
self.info(
'%s(%s)', self.name, ', '.join(self._repr_iter(**params))
)
self.validate(**params)
(args, options) = self.params_2_args_options(**params)
result = self.run(*args, **options)
self.debug('result from %s(): %r', self.name, result)
return result
def _repr_iter(self, **params):
"""
Iterate through ``repr()`` of *safe* values of args and options.
This method uses `parameters.Param.safe_value()` to mask passwords when
logging. Logging the exact call is extremely useful, but we obviously
don't want to log the cleartext password.
For example:
>>> class my_cmd(Command):
... takes_args = ('login',)
... takes_options=(Password('passwd'),)
...
>>> c = my_cmd()
>>> c.finalize()
>>> list(c._repr_iter(login=u'Okay.', passwd=u'Private!'))
["u'Okay.'", "passwd=u'********'"]
"""
for arg in self.args():
value = params.get(arg.name, None)
yield repr(arg.safe_value(value))
for option in self.options():
if option.name not in params:
continue
value = params[option.name]
yield '%s=%r' % (option.name, option.safe_value(value))
def args_options_2_params(self, *args, **options):
"""
Merge (args, options) into params.
"""
if self.max_args is not None and len(args) > self.max_args:
if self.max_args == 0:
raise ZeroArgumentError(name=self.name)
raise MaxArgumentError(name=self.name, count=self.max_args)
params = dict(self.__options_2_params(options))
if len(args) > 0:
arg_kw = dict(self.__args_2_params(args))
intersection = set(arg_kw).intersection(params)
if len(intersection) > 0:
raise OverlapError(names=sorted(intersection))
params.update(arg_kw)
return params
def __args_2_params(self, values):
multivalue = False
for (i, arg) in enumerate(self.args()):
assert not multivalue
if len(values) > i:
if arg.multivalue:
multivalue = True
if len(values) == i + 1 and type(values[i]) in (list, tuple):
yield (arg.name, values[i])
else:
yield (arg.name, values[i:])
else:
yield (arg.name, values[i])
else:
break
def __options_2_params(self, options):
for name in self.params:
if name in options:
yield (name, options[name])
def args_options_2_entry(self, *args, **options):
"""
Creates a LDAP entry from attributes in args and options.
"""
kw = self.args_options_2_params(*args, **options)
return dict(self.__attributes_2_entry(kw))
def __attributes_2_entry(self, kw):
if self.api.env.use_ldap2:
for name in self.params:
if self.params[name].attribute and name in kw:
value = kw[name]
if isinstance(value, tuple):
yield (name, [v for v in value])
else:
yield (name, kw[name])
else:
for name in self.params:
if self.params[name].attribute and name in kw:
if type(kw[name]) is tuple:
yield (name, [str(value) for value in kw[name]])
else:
yield (name, str(kw[name]))
def params_2_args_options(self, **params):
"""
Split params into (args, options).
"""
args = tuple(params.get(name, None) for name in self.args)
options = dict(self.__params_2_options(params))
return (args, options)
def __params_2_options(self, params):
for name in self.options:
if name in params:
yield(name, params[name])
def normalize(self, **kw):
"""
Return a dictionary of normalized values.
For example:
>>> class my_command(Command):
... takes_options = (
... Param('first', normalizer=lambda value: value.lower()),
... Param('last'),
... )
...
>>> c = my_command()
>>> c.finalize()
>>> c.normalize(first=u'JOHN', last=u'DOE')
{'last': u'DOE', 'first': u'john'}
"""
return dict(
(k, self.params[k].normalize(v)) for (k, v) in kw.iteritems()
)
def convert(self, **kw):
"""
Return a dictionary of values converted to correct type.
>>> from ipalib import Int
>>> class my_command(Command):
... takes_args = (
... Int('one'),
... 'two',
... )
...
>>> c = my_command()
>>> c.finalize()
>>> c.convert(one=1, two=2)
{'two': u'2', 'one': 1}
"""
return dict(
(k, self.params[k].convert(v)) for (k, v) in kw.iteritems()
)
def __convert_iter(self, kw):
for param in self.params():
if kw.get(param.name, None) is None:
continue
def get_default(self, **kw):
"""
Return a dictionary of defaults for all missing required values.
For example:
>>> from ipalib import Str
>>> class my_command(Command):
... takes_args = Str('color', default=u'Red')
...
>>> c = my_command()
>>> c.finalize()
>>> c.get_default()
{'color': u'Red'}
>>> c.get_default(color=u'Yellow')
{}
"""
return dict(self.__get_default_iter(kw))
def __get_default_iter(self, kw):
"""
Generator method used by `Command.get_default`.
"""
for param in self.params():
if param.name in kw:
continue
if param.required or param.autofill:
default = param.get_default(**kw)
if default is not None:
yield (param.name, default)
def validate(self, **kw):
"""
Validate all values.
If any value fails the validation, `ipalib.errors.ValidationError`
(or a subclass thereof) will be raised.
"""
for param in self.params():
value = kw.get(param.name, None)
param.validate(value)
def run(self, *args, **options):
"""
Dispatch to `Command.execute` or `Command.forward`.
If running in a server context, `Command.execute` is called and the
actually work this command performs is executed locally.
If running in a non-server context, `Command.forward` is called,
which forwards this call over XML-RPC to the exact same command
on the nearest IPA server and the actual work this command
performs is executed remotely.
"""
if self.api.env.in_server:
return self.execute(*args, **options)
return self.forward(*args, **options)
def execute(self, *args, **kw):
"""
Perform the actual work this command does.
This method should be implemented only against functionality
in self.api.Backend. For example, a hypothetical
user_add.execute() might be implemented like this:
>>> class user_add(Command):
... def execute(self, **kw):
... return self.api.Backend.ldap.add(**kw)
...
"""
raise NotImplementedError('%s.execute()' % self.name)
def forward(self, *args, **kw):
"""
Forward call over XML-RPC to this same command on server.
"""
return self.Backend.xmlclient.forward(self.name, *args, **kw)
def finalize(self):
"""
Finalize plugin initialization.
This method creates the ``args``, ``options``, and ``params``
namespaces. This is not done in `Command.__init__` because
subclasses (like `crud.Add`) might need to access other plugins
loaded in self.api to determine what their custom `Command.get_args`
and `Command.get_options` methods should yield.
"""
self._create_param_namespace('args')
if len(self.args) == 0 or not self.args[-1].multivalue:
self.max_args = len(self.args)
else:
self.max_args = None
self._create_param_namespace('options')
def get_key(p):
if p.required:
if p.default_from is None:
return 0
return 1
return 2
self.params = NameSpace(
sorted(tuple(self.args()) + tuple(self.options()), key=get_key),
sort=False
)
super(Command, self).finalize()
def get_args(self):
"""
Iterate through parameters for ``Command.args`` namespace.
This method gets called by `HasParam._create_param_namespace()`.
Subclasses can override this to customize how the arguments are
determined. For an example of why this can be useful, see the
`ipalib.crud.Create` subclass.
"""
for arg in self._get_param_iterable('args'):
yield arg
def check_args(self, args):
"""
Sanity test for args namespace.
This method gets called by `HasParam._create_param_namespace()`.
"""
optional = False
multivalue = False
for arg in args():
if optional and arg.required:
raise ValueError(
'%s: required argument after optional' % arg.name
)
if multivalue:
raise ValueError(
'%s: only final argument can be multivalue' % arg.name
)
if not arg.required:
optional = True
if arg.multivalue:
multivalue = True
def get_options(self):
"""
Iterate through parameters for ``Command.options`` namespace.
This method gets called by `HasParam._create_param_namespace()`.
Subclasses can override this to customize how the arguments are
determined. For an example of why this can be useful, see the
`ipalib.crud.Create` subclass.
"""
for option in self._get_param_iterable('options'):
yield option
class LocalOrRemote(Command):
"""
A command that is explicitly executed locally or remotely.
This is for commands that makes sense to execute either locally or
remotely to return a perhaps different result. The best example of
this is the `ipalib.plugins.f_misc.env` plugin which returns the
key/value pairs describing the configuration state: it can be
"""
takes_options = (
Flag('server?',
doc='Forward to server instead of running locally',
),
)
def run(self, *args, **options):
"""
Dispatch to forward() or execute() based on ``server`` option.
When running in a client context, this command is executed remotely if
``options['server']`` is true; otherwise it is executed locally.
When running in a server context, this command is always executed
locally and the value of ``options['server']`` is ignored.
"""
if options['server'] and not self.env.in_server:
return self.forward(*args, **options)
return self.execute(*args, **options)
class Object(HasParam):
__public__ = frozenset((
'backend',
'methods',
'properties',
'params',
'primary_key',
'params_minus_pk',
'params_minus',
'get_dn',
))
backend = None
methods = None
properties = None
params = None
primary_key = None
params_minus_pk = None
# Can override in subclasses:
backend_name = None
takes_params = tuple()
def set_api(self, api):
super(Object, self).set_api(api)
self.methods = NameSpace(
self.__get_attrs('Method'), sort=False
)
self.properties = NameSpace(
self.__get_attrs('Property'), sort=False
)
self._create_param_namespace('params')
pkeys = filter(lambda p: p.primary_key, self.params())
if len(pkeys) > 1:
raise ValueError(
'%s (Object) has multiple primary keys: %s' % (
self.name,
', '.join(p.name for p in pkeys),
)
)
if len(pkeys) == 1:
self.primary_key = pkeys[0]
self.params_minus_pk = NameSpace(
filter(lambda p: not p.primary_key, self.params()), sort=False
)
if 'Backend' in self.api and self.backend_name in self.api.Backend:
self.backend = self.api.Backend[self.backend_name]
def params_minus(self, *names):
"""
Yield all Param whose name is not in ``names``.
"""
if len(names) == 1 and not isinstance(names[0], (Param, str)):
names = names[0]
minus = frozenset(names)
for param in self.params():
if param.name in minus or param in minus:
continue
yield param
def get_dn(self, *args, **kwargs):
"""
Construct an LDAP DN.
"""
raise NotImplementedError('%s.get_dn()' % self.name)
def __get_attrs(self, name):
if name not in self.api:
return
namespace = self.api[name]
assert type(namespace) is NameSpace
for proxy in namespace(): # Equivalent to dict.itervalues()
if proxy.obj_name == self.name:
yield proxy.__clone__('attr_name')
def get_params(self):
"""
This method gets called by `HasParam._create_param_namespace()`.
"""
props = self.properties.__todict__()
for spec in self._get_param_iterable('params'):
if type(spec) is str:
key = spec.rstrip('?*+')
else:
assert isinstance(spec, Param)
key = spec.name
if key in props:
yield props.pop(key).param
else:
yield create_param(spec)
def get_key(p):
if p.param.required:
if p.param.default_from is None:
return 0
return 1
return 2
for prop in sorted(props.itervalues(), key=get_key):
yield prop.param
class Attribute(Plugin):
"""
Base class implementing the attribute-to-object association.
`Attribute` plugins are associated with an `Object` plugin to group
a common set of commands that operate on a common set of parameters.
The association between attribute and object is done using a simple
naming convention: the first part of the plugin class name (up to the
first underscore) is the object name, and rest is the attribute name,
as this table shows:
=============== =========== ==============
Class name Object name Attribute name
=============== =========== ==============
noun_verb noun verb
user_add user add
user_first_name user first_name
=============== =========== ==============
For example:
>>> class user_add(Attribute):
... pass
...
>>> instance = user_add()
>>> instance.obj_name
'user'
>>> instance.attr_name
'add'
In practice the `Attribute` class is not used directly, but rather is
only the base class for the `Method` and `Property` classes. Also see
the `Object` class.
"""
__public__ = frozenset((
'obj',
'obj_name',
))
__obj = None
def __init__(self):
m = re.match(
'^([a-z][a-z0-9]+)_([a-z][a-z0-9]+(?:_[a-z][a-z0-9]+)*)$',
self.__class__.__name__
)
assert m
self.__obj_name = m.group(1)
self.__attr_name = m.group(2)
super(Attribute, self).__init__()
def __get_obj_name(self):
return self.__obj_name
obj_name = property(__get_obj_name)
def __get_attr_name(self):
return self.__attr_name
attr_name = property(__get_attr_name)
def __get_obj(self):
"""
Returns the obj instance this attribute is associated with, or None
if no association has been set.
"""
return self.__obj
obj = property(__get_obj)
def set_api(self, api):
self.__obj = api.Object[self.obj_name]
super(Attribute, self).set_api(api)
class Method(Attribute, Command):
"""
A command with an associated object.
A `Method` plugin must have a corresponding `Object` plugin. The
association between object and method is done through a simple naming
convention: the first part of the method name (up to the first under
score) is the object name, as the examples in this table show:
============= =========== ==============
Method name Object name Attribute name
============= =========== ==============
user_add user add
noun_verb noun verb
door_open_now door open_now
============= =========== ==============
There are three different places a method can be accessed. For example,
say you created a `Method` plugin and its corresponding `Object` plugin
like this:
>>> from ipalib import create_api
>>> api = create_api()
>>> class user_add(Method):
... def run(self):
... return 'Added the user!'
...
>>> class user(Object):
... pass
...
>>> api.register(user_add)
>>> api.register(user)
>>> api.finalize()
First, the ``user_add`` plugin can be accessed through the ``api.Method``
namespace:
>>> list(api.Method)
['user_add']
>>> api.Method.user_add() # Will call user_add.run()
'Added the user!'
Second, because `Method` is a subclass of `Command`, the ``user_add``
plugin can also be accessed through the ``api.Command`` namespace:
>>> list(api.Command)
['user_add']
>>> api.Command.user_add() # Will call user_add.run()
'Added the user!'
And third, ``user_add`` can be accessed as an attribute on the ``user``
`Object`:
>>> list(api.Object)
['user']
>>> list(api.Object.user.methods)
['add']
>>> api.Object.user.methods.add() # Will call user_add.run()
'Added the user!'
The `Attribute` base class implements the naming convention for the
attribute-to-object association. Also see the `Object` and the
`Property` classes.
"""
__public__ = Attribute.__public__.union(Command.__public__)
extra_options_first = False
extra_args_first = False
def __init__(self):
super(Method, self).__init__()
class Property(Attribute):
__public__ = frozenset((
'rules',
'param',
'type',
)).union(Attribute.__public__)
klass = Str
default = None
default_from = None
normalizer = None
def __init__(self):
super(Property, self).__init__()
self.rules = tuple(
sorted(self.__rules_iter(), key=lambda f: getattr(f, '__name__'))
)
self.kwargs = tuple(
sorted(self.__kw_iter(), key=lambda keyvalue: keyvalue[0])
)
kw = dict(self.kwargs)
self.param = self.klass(self.attr_name, *self.rules, **kw)
def __kw_iter(self):
for (key, kind, default) in self.klass.kwargs:
if getattr(self, key, None) is not None:
yield (key, getattr(self, key))
def __rules_iter(self):
"""
Iterates through the attributes in this instance to retrieve the
methods implementing validation rules.
"""
for name in dir(self.__class__):
if name.startswith('_'):
continue
base_attr = getattr(self.__class__, name)
if is_rule(base_attr):
attr = getattr(self, name)
if is_rule(attr):
yield attr
class Application(Command):
"""
Base class for commands register by an external application.
Special commands that only apply to a particular application built atop
`ipalib` should subclass from ``Application``.
Because ``Application`` subclasses from `Command`, plugins that subclass
from ``Application`` with be available in both the ``api.Command`` and
``api.Application`` namespaces.
"""
__public__ = frozenset((
'application',
'set_application'
)).union(Command.__public__)
__application = None
def __get_application(self):
"""
Returns external ``application`` object.
"""
return self.__application
application = property(__get_application)
def set_application(self, application):
"""
Sets the external application object to ``application``.
"""
if self.__application is not None:
raise AttributeError(
'%s.application can only be set once' % self.name
)
if application is None:
raise TypeError(
'%s.application cannot be None' % self.name
)
object.__setattr__(self, '_Application__application', application)
assert self.application is application