[GRASS-SVN] r55390 - grass/trunk/lib/python/pydispatch
svn_grass at osgeo.org
svn_grass at osgeo.org
Thu Mar 14 15:59:48 PDT 2013
Author: wenzeslaus
Date: 2013-03-14 15:59:48 -0700 (Thu, 14 Mar 2013)
New Revision: 55390
Modified:
grass/trunk/lib/python/pydispatch/dispatcher.py
grass/trunk/lib/python/pydispatch/pydispatchlib.dox
grass/trunk/lib/python/pydispatch/signal.py
Log:
pythonlib/pydispatch: API documentation
Modified: grass/trunk/lib/python/pydispatch/dispatcher.py
===================================================================
--- grass/trunk/lib/python/pydispatch/dispatcher.py 2013-03-14 22:56:25 UTC (rev 55389)
+++ grass/trunk/lib/python/pydispatch/dispatcher.py 2013-03-14 22:59:48 UTC (rev 55390)
@@ -368,7 +368,9 @@
def _removeReceiver(receiver):
"""Remove receiver from connections."""
if not sendersBack or not connections:
- # During module cleanup the mapping will be replaced with None
+ # During module cleanup the objects will be replaced with None
+ # The order of replacing many chnage, so both variables need
+ # to be checked.
return False
backKey = id(receiver)
try:
Modified: grass/trunk/lib/python/pydispatch/pydispatchlib.dox
===================================================================
--- grass/trunk/lib/python/pydispatch/pydispatchlib.dox 2013-03-14 22:56:25 UTC (rev 55389)
+++ grass/trunk/lib/python/pydispatch/pydispatchlib.dox 2013-03-14 22:59:48 UTC (rev 55390)
@@ -1,4 +1,4 @@
-/*! \page pydispatch Customized pydispatcher library
+/*! \page pydispatch Customized PyDispatcher library
by GRASS Development Team (http://grass.osgeo.org)
@@ -8,7 +8,7 @@
\section pydispatchIntro Introduction
Files dispatcher.py, errors.py, robustapply.py, robust.py and saferef.py are
-part of the original pydispatcher library.
+part of the original PyDispatcher library.
File signal.py is the object-based easy-to-use interface created for GRASS.
Signals as a concept (API) does not depend on PyDispatcher,
@@ -188,9 +188,9 @@
\section pydispatchAlternatives Alternatives
-There is wxPython' pubsub module which provides similar functionality to pydispatcher. However, it has two incompatible APIs which depends on the wxPython version and in the time of writing, GRASS would have to support both these APIs. The only possibility is to include pubsub into GRASS. Whether to use pydispatcher or pubsub usually depends on personal taste.
+There is wxPython's pubsub module which provides similar functionality to PyDispatcher. However, it has two incompatible APIs which depends on the wxPython version and in the time of writing, GRASS would have to support both these APIs. The only possibility is to include pubsub into GRASS. Whether to use PyDispatcher or pubsub usually depends on personal taste.
-There are several forks of pydispatcher, e.g. the fork in Django framework.
+There are several forks of PyDispatcher, e.g. the fork in Django framework.
They provide the same or similar functionality.
Of course it is also possible to implement this system from the scratch using e.g., observer pattern.
Modified: grass/trunk/lib/python/pydispatch/signal.py
===================================================================
--- grass/trunk/lib/python/pydispatch/signal.py 2013-03-14 22:56:25 UTC (rev 55389)
+++ grass/trunk/lib/python/pydispatch/signal.py 2013-03-14 22:59:48 UTC (rev 55390)
@@ -10,7 +10,8 @@
def _islambda(function):
- """Tests if object is a lambda function.
+ """
+ Tests if object is a lambda function.
Should work on the most of Python implementations where name of lambda
function is not unique.
@@ -26,22 +27,47 @@
class Signal(object):
"""
+
+ The signal object is created usually as a instance attribute.
+ However, it can be created anywhere.
+
>>> signal1 = Signal('signal1')
+
+ The function has to be connected to a signal in order to be called when
+ the signal is emitted. The connection can be done where the function is
+ defined (e. g., a class) but also on some other place, typically,
+ user of a class connects some signal to the method of some other class.
+
>>> def handler1():
... print "from handler1"
>>> signal1.connect(handler1)
+ Emitting of the signal is done usually only in the class which has the
+ signal as a instance attribute. Again, generally, it can be emitted
+ anywhere.
+
+ >>> signal1.emit()
+ from handler1
+
+ The signal can have parameters. These parameters are specified when
+ emitting but should be documented together with the signal (e.g., in the
+ class documentation). Parameters should be keyword arguments and handlers
+ must use these names (if the names cannot be the same, lambda function
+ can be used to overcome this problem).
+
>>> signal2 = Signal('signal2')
>>> def handler2(text):
... print "handler2: %s" % text
>>> signal2.connect(handler2)
-
- >>> signal1.emit()
- from handler1
-
>>> signal2.emit(text="Hello")
handler2: Hello
+ Do not emit the same signal with different parameters when emitting at
+ different places.
+
+ A handler is the standard function, lambda function, method or any other
+ callable object.
+
>>> import sys
>>> signal2.connect(lambda text:
... sys.stdout.write('lambda handler: %s\\n' % text))
@@ -49,6 +75,9 @@
handler2: Hi
lambda handler: Hi
+ The handler function can have only some of the signal parameters or no
+ parameters at all even if the signal has some.
+
>>> def handler3():
... print "from handler3"
>>> signal2.connect(handler3)
@@ -57,6 +86,10 @@
lambda handler: Ciao
from handler3
+ It is possible to use signal as a handler. By this, signals can be
+ forwarded from one object to another. In other words, one object can
+ expose signal of some object.
+
>>> signal3 = Signal('signal3')
>>> signal3.connect(handler3)
>>> signal1.connect(signal3)
@@ -64,6 +97,8 @@
from handler1
from handler3
+ It is possible to disconnect a particular handler.
+
>>> signal3.disconnect(handler3)
>>> signal1.emit()
from handler1
@@ -72,11 +107,32 @@
>>> signal2.emit(text='Hello')
lambda handler: Hello
"""
+ # TODO: use the name for debugging
def __init__(self, name):
+ """Creates a signal object.
+
+ The parameter name is used for debugging.
+ """
self._name = name
def connect(self, handler, weak=None):
"""
+ Connects handler to a signal.
+
+ Typically, a signal is defined in some class and the user of this
+ class connects to the signal::
+
+ from module import SomeClass
+ ...
+ self.someObject = SomeClass()
+ self.someObject.connect(self.someMethod)
+
+ Usually, it is not needed to set the weak parameter. This method
+ creates weak references for all handlers but for lambda functions, it
+ automaticaly creates (standard) references (otherwise, lambdas would be
+ garbage collected. If you want to force some behaviour, specify the
+ weak parameter.
+
>>> signal1 = Signal('signal1')
>>> import sys
>>> signal1.connect(lambda: sys.stdout.write('will print\\n'))
@@ -95,14 +151,29 @@
def disconnect(self, handler, weak=True):
"""
-
+ Disconnects a specified handler.
+
+ It is not necessary to disconnect object when it is deleted.
+ Underlying PyDispatcher will take care of connections to deleted
+ objects.
+
>>> signal1 = Signal('signal1')
>>> import sys
>>> signal1.connect(sys.stdout.write)
>>> signal1.disconnect(sys.stdout.write)
+
+ The weak parameter of must have the same value as for connection.
+ If you not specified the parameter when connecting,
+ you don't have to specify it when disconnecting.
+
+ Disconnecting the not-connected handler will result in error.
+
>>> signal1.disconnect(sys.stdout.flush) #doctest: +ELLIPSIS
Traceback (most recent call last):
DispatcherKeyError: 'No receivers found for signal <__main__.Signal object at 0x...> from sender _Any'
+
+ Disconnecting the non-exiting or unknown handler will result in error.
+
>>> signal1.disconnect(some_function)
Traceback (most recent call last):
NameError: name 'some_function' is not defined
@@ -111,8 +182,37 @@
dispatcher.disconnect(receiver=handler, signal=self, weak=weak)
# TODO: remove args?, make it work for args?
+ # TODO: where to put documentation
def emit(self, *args, **kwargs):
"""
+ Emits the signal which means that all connected handlers will be
+ called.
+
+ It is advised to have signals as instance attributes and emit signals
+ only in the class which owns the signal::
+
+ class Abc(object):
+ def __init__(self):
+ self.colorChanged = Signal('Abc.colorChanged')
+ ...
+ def setColor(self, color):
+ ...
+ self.colorChanged.emit(oldColor=self.Color, newColor=color)
+ ...
+
+ Documentation of an signal should be placed to the class documentation
+ or to the code (this need to be more specified).
+
+ Calling a signal from outside the class is usually not good
+ practice. The only case when it is permitted is when signal is the part
+ of some globaly shared object and permision to emit is stayed in the
+ documentation.
+
+ The parameters of the emit function must be the same as the parameters
+ of the handlers. However, handler can ommit some parameters.
+ The associated parameters shall be documented for each Signal instance.
+ Use only keyword arguments when emitting.
+
>>> signal1 = Signal('signal1')
>>> def mywrite(text):
... print text
@@ -128,6 +228,7 @@
"""
dispatcher.send(signal=self, *args, **kwargs)
+ # TODO: remove args?
def __call__(self, *args, **kwargs):
"""Allows to emit signal with function call syntax.
@@ -157,6 +258,7 @@
name for parameter when using signals), the error is much more readable
when using emit than function call. Concluding remark is that
emit behaves more predictable.
+
>>> signal1.emit(signal='Hello')
Traceback (most recent call last):
TypeError: send() got multiple values for keyword argument 'signal'
More information about the grass-commit
mailing list