Yep, I can see why you are confused. I'm also very thankful to have your
statement below about what the actual runtime does and i'd like to think
the example i sent up just helped to get to this information out.. I do
have a "contracts" with my OpenLaszlo designs that I receive notice
anytime an attribute changes including constraint-related changes. I see
LZX as a more basic combination then what actually exists in OpenLaszlo,
whether that be for optimization sake or no, but _not_ so much that the
coding patterns can't coexist. I did get burned by a few of them
problem-solving wise, and i hope i can be there to answer questions for
others that have them.
And i seem to have quite a few definitions that are not typical. traits
and constraints - they even rhyme. guess i want to corral those words
for laszlo, since constraints are indeed an overloaded term but i like
the Laszlo usage the best.
P T Withington wrote:
Ok, you've confused me now.
You seem to be saying that you don't need events, and you've offered as proof
an implementation of a constraint system in Python, similar to OpenLaszlo's,
where you have used setters to implement a simple event system on which to
build your constraints.
I think you just have a definition of "constraint" that is different than the
accepted definition (and what OpenLaszlo implements). In OpenLaszlo, you can constrain
an attribute to the value of an expression that involves other object's attributes. The
contract of this constraint is to ensure that the constrained attribute always has the
value that would be computed by that expression. Nothing in the contract of constraints
says that there are events sent. That is an implementation detail. We could implement
constraints using a completely different mechanism (and have several experimental ones
that could significantly improve the performance of applications with many constraints).
On 2009-12-22, at 12:43, jamesr wrote:
aye, and it's better confusion because i have a portable working model
in LZX that never needs events. even if it is implemented "under the
hood", i still have yet to encounter a single piece of code that needed
it, conceptually. so i'm confused, as some are, as to why the
distinction. Yes, i understand that events don't send state like
attributes do, so when i need a "pure" event i just send true, or some
ignored value, and i teach that pattern.
Just for illustrative reasons, you'll appreciate this, In python it is
done with "setters" as well, but there is no event class.
here's the excerpt copy (2 relavent methods) of the constrain and
setattr methods on a <node> in pyroglyph, a python runtime,, evaluating
constraints, it shows the two functions on node.py that set up a
constraint, and fire that constraint when values change. as you might
interpret, there is no setAttribute call here.
def constrain(self, attr=None, func=None, set=False):
""" make a constraint, that is, add to _handlers on a target node a
pointer
to a function to call when it value changes. Setting set to
true (the default) also sets that value on the node. This is because
before the init is completed, there may not be value to set!
"""
#place a pointer to this new function as a listener on the node
pointed to by the handler
if not self._handlers.has_key(attr):
self.__dict__['_handlers'][attr] = []
self.__dict__['_handlers'][attr].append(func)
#if the constraint is made after initialization, set its value
immediately from the target
if set: func(self[attr])
and
def __setattr__(self, attr, value):
""" and.. the very heart of delegates, events, constraints, etc! """
#set the value
self.__dict__[attr] = value
#set the canvas as being dirty - that is, something changed and we
need to redraw!
#this is an optimaztion that may be removed with runtimes other then
openGL.
self.canvas.__dict__['dirty'] = True
#distribute this value to all other listeners for this attribute
if self._handlers.has_key(attr):
[listener(value) for listener in self._handlers[attr]]
P T Withington wrote:
James, I think you are conflating setters and constraints. OpenLaszlo has
setters which are always invoked by setAttribute; this is similar to the other
languages you mention.
OpenLaszlo also has the ability to constrain attribute values. There are very
few other languages that have this feature. Constraints in OpenLaszlo are
implemented by setters and events, which may be the source of the confusion.
On Dec 21, 2009, at 21:43, jamesr <[email protected]> wrote:
unfortunately, this side effect of the constraint system, bypassing events in
normal use, is not a side effect but an intended main effect in other runtimes
and part of a broader lzx concept i play with, so i'm conceptually bound to
support only using attributes for handlers. It is nice to find out they are in
use, although i've spoken with many lzx coders that don't know anything about
events, only attributes.
"
The constraint system uses a necessary, but not sufficient, condition for
sending events --
"
I have solved that problem here via explicit replication for some cases; it
doesn't remove the need for laszlo replication but compliments it. so i can't
really complain. Of course, anything using laszlo replication may not be
portable to other runtimes as-a-pattern, because they may not support the
laszlo datapathing as laszlo does it. That's by way of explaining why i care so
much.
A while ago when i first found out about this i was unsure what to do, but it
is conveniently patched, so life can go on! more soon...
P T Withington wrote:
Attributes _do_ define state in LZX. LZX has always had attributes/constraints
_and_ events/handlers. It has always been the case that you could constrain
the value of one attribute to an expression that depended on the values of
other attributes and the constraint system would ensure that the constrained
value was maintained. It has always been the case that you could listen for
and send events. It _so_happens_ that the constraint system is implemented
using events to signal when a dependent of a constraint has changed. As a
side-effect, you can attach a handler to the implicit event associated with an
attribute, and you are guaranteed to get invoked any time the value of that
attribute changes. You are _not_ guaranteed to get an event if the attribute
is set to the same value. By the same token, you are not guaranteed to _not_
get an event if the constraint is recalculated and doesn't change. The
constraint system uses a necessary, but not sufficient, conditi!
on for sending events -- that is, it sends enough events to maintain the
constraints, but it does not send the minimum possible set of events. For this
reason, if you actually want an event to be sent, you should use explicit
events, not rely on a side-effect of the constraint system.
On 2009-12-21, at 12:30, jamesr wrote:
Ah it makes so much sense now. For those that are not up with the
attribute-event-handler debate, it is a semantic discussion about the
appropriate way to design flexible laszlo components. There are two dominant
patterns:
One: Attributes define state, and things respond to that state
Two: Events define interest, and things respond to that interest by retrieving
and using attributes
I'll refer to them as events and attributes below
The two seem almost interchangable but there is a difference in program design.
The biggest difference is that the attributes model is simpler to teach and
predict. The idea that there is a second event system which attribute access
almost covers leads to confusion. Another is that the attribute model is
simpler to implement in other more powerful languages (python) where setattr
and getattr remove the need for the .setAttribute syntax altogether; writing
LZX to the Attributes spec represents a more portable codebase. Events are
something pulled more from other languages and less from the lzx model to begin
with, so i would argue. When first learning laszlo I was safely able to forget
about the event tag, and i've been 100% okay without it (but for one pitfall i
hacked around using getTime() to change values every time, which i'm okay with,
for now)
The more-recently introduced condition the cautionary note PT wrote speaks
about is that there is a case when attributes will fail to send events,
requiring the use of the event tag, in one edge case. I don't know if everyone
is following, but this is something we can choose, and it was done for a
reason. I'm hoping for a flag in future version to allow a pure attribute
model, as i don't have the event tag in _any_ of production code! And also,
because that characteristic seems like it would be in a pure LZX implementation
(i.e. not javascript but something more powerful, like python or ruby)
So, i have a question for the list at large, not PT or Henry who do enough :) For
anyone reading this, when was the last time a listmember reading this used
<event> so as to work with handlers, anywhere, in any code base? Has anyone
considered this perhaps, or like me, never thought of it until it wasn't there?
Kind regards,
j.
P T Withington wrote:
Just to be clear, the idea proposed in
http://jira.openlaszlo.org/jira/browse/LPP-7816
is not yet implemented. Currently, you can only use <handler> to connect to an event
that is already declared (either explicitly using <event> or implicitly on an
attribute).
And one caution: If you create a handler on the implicit event associated with
a constrained attribute, the handler will _only_ be invoked when the attribute
changes value. Consider this example:
<canvas>
<view layout="axis: y; spacing: 5">
<attribute name="set" value="false" />
<attribute name="followset" value="${this.set}" />
<handler name="onset" args="newvalue">
Debug.info("set: %w", newvalue);
</handler>
<handler name="onfollowset" args="newvalue">
Debug.info("followset: %w", newvalue);
</handler>
<button onclick="parent.setAttribute('set', true)">setAttribute('set',
true)</button>
<button onclick="parent.setAttribute('set', new Date())">setAttribute('set', new
Date())</button>
<event name="send"/>
<handler name="send" args="eventvalue">
Debug.info("sent: %w", eventvalue);
</handler>
<button onclick="parent.send.sendEvent(true)">sendEvent(true)</button>
<button onclick="parent.send.sendEvent(new Date())">sendEvent(new
Date())</button>
</view>
</canvas>
Notice that although the handler "onset" is invoked every time you set the attribute, whether the
value changes or not, the hander "onfollowset" is _only_ invoked when the value it is constrained to
changes. Personally, if I want to send events, I prefer to be explicit about it, as shown in the second half
of the example, and declare the <event> and <hander> and use `sendEvent`, rather than relying on
the implicit events sent when attributes change.
On 2009-12-18, at 17:23, cem sonmez wrote:
@P T Withington : I looked at your your solution for creating delegates.
When i need to use the delegates, i m going to use the way that you
mentioned. While looking at the developer documentation of openlaszlo, I
hadnT understand the meaning of delegates and the usage of it. it seemed to
me a bit confused :) Now, i think know more about delegates than before,
thanks
@jamesr : also thank you for the information. I worked out the problem. I
used an attrbiute in the connection.lzx
<attribute name="connectionOK" value="false" type="boolean"/>
and in the case of "NetConnection.Connect.Success", i set it to true.
Then while creating an object of sharedObject, i used like that :
<sharedObjectChat id="soChat">
<handler name="onconnectionOK" reference="conn">
Debug.info("connectionOK attribute operation, value
:%w",conn.connectionOK);
if (conn.connectionOK==true) {
//this.setAttribute("createSharedObject",true);
this.connect("chat", conn._conn, true);
this.so.client=this;
}
</handler>
</sharedObjectChat>
So itS ok now. İt is impossible to learn something wihout making mistakes :)
Best regards.
2009/12/18 P T Withington <[email protected]>
I have a proposed "solution" for the LZX programmer having to create
delegates in script in:
http://jira.openlaszlo.org/jira/browse/LPP-7816
Basically, I want you to be able to declare a handler without attaching it
to an event at declaration time (and you would later use script to register
the handler on the event it is to handle). The main purpose for this change
is that <handler /> will automatically be managed, whereas delegates must be
manually managed (and destroyed) or they can lead to memory leaks.
On 2009-12-18, at 16:03, jamesr wrote:
for the record, the two use cases for making a delegate are
1) you want to compute what object to latch onto programmatically and
2) you want to latch onto something that is not yet created at the
handler init phase.
don't see any other use, although the two above are obviously important
cem sonmez wrote:
you are exactly right. I want to make some operations when the
netconnection successful event is fired. Actually i m trying to get the
shared object on the server side, so at first i need a netconnection
instance to do that. ThatS why i thought that i should try delegates. I
havenT used delegates before, but i thought that this is the just one
solution to handle the operations depend on the related event.
The method you said, using an attribute in the connection class seems
the better way. Till now, i havenT needed the delegates. I hope using
connection status attribute in the connection class will fix the problem.
I m going to post back the results here of course.
Thanks for the reply J
2009/12/18 jamesr <[email protected] <mailto:[email protected]
I can tackle it. let's see; i'd write the following, after an
introduction (all code is untested and represents a pattern, adapt
it as you will)
Your problem: You want to do is send a "connection successful"
event that you can catch, signaling that a method is to be called
to continue using the connection. You seem to be trying to do it
by setting up the delegate manually, when i think, looking at your
code, that you have hooks to fire events on laszlo nodes that can
communicate this state change.
Solution: An event is anytime you use x.setAttribute() by the way
- that's what it does, makes sure that handlers are fired. That
said, if you declare an attribute of any laszlo node and in your
code, when you have success on your net connection, you say
"somenode.setAttribute('success, true);" then in another node you
can say, <handler name="onfoo" reference="somenode">...</handler>
and in that code you can then do what ever other steps are
required to use the connection.
Further thought if i'm wrong: can you say why it is you decided to
use manual delegates? it might make things clearer for me, i've
not used the net connection code you have there but make guesses.
.j.
cem sonmez wrote:
doesnT anyone have any idea. I got stuck here. Waiting for
someones advices.
Thanks
2009/12/18 cem sonmez <[email protected]
<mailto:[email protected]> <mailto:[email protected]
<mailto:[email protected]>>>
hi
when i try do use the delegate such like :
if( typeof this.del == "undefined" ) {
this.del = new LzDelegate(this, "connect('chat',
conn,
true)" );
} this.del.register(conn,
"netStatusHandler('NetConnection.Connect.Success')" );
}
I m getting the error on the debug like :
*soChat.connect('chat', conn, true) => (void 0) (must be a
function)*
Actually i want to do this : call the *connect* method on the
class when the netStatusHandler method of the *conn* object
has
been completed. I m not sure that am i using the delegate
correctly (as i m getting the error, of course not :)).
I have attached the relevant files.
Can anyone help me what i m missing to do.
Kind regards.
-- Cem SONMEZ
-- Cem SONMEZ
--
Cem SONMEZ
--
Cem SONMEZ
