Well, either way, thanks for bringing it to my
attention. I dig the low level things, and this is pretty cool to know
that it's that common across all UIComponents, in both 1.5 and 2.
The TextInput thing I mentioned, I didn't
build. I was using TextInputs everywhere, and my team members were like,
"Dude, what are you doing?". They said to use one they built. It
had basically 3 things; a specific look the client liked, made 2 events into
1, and maxChar as a public property. There were 2 other minor
things. I think the reason it was built was because the most common use
case was needing both enter & focusOut to fire update events, and since
they were using this in forms everywhere, they reduced a ton of code by
refactoring it this way.
In Flex 1.5, yes, this has a cost.
Inheritance in Flash Player 8 and below has a cost, mainly because of the
inheritance chain and method lookups. If a method is invoked, and it's
not on "you", it asks its parent class. This chain lookup is
expensive. So gregarious extending does cause performance
impacts.
This problem doesn't exist in Flex 2 because of
the new traits object (vs. prototype). So, now, it's more of a decision
of "should we use inteheritance" with no need to worry about performance
impacts.
As far as consolidition, I think the above
example is extremely helpful. We have a common use case for the control,
and it has a specific design, so yeah man, extending it for that simple reason
has helped a lot. It's easier to handle the look in Flex 2 because CSS
is soo much cooler now.
Well, yeah, the change thing is an issue.
However, I'm usually the one digging around in base classes. I have to
the balls to know if I do one thing wrong, I've broken like 200 uses of the
control... so it better not be wrong! Seriously, though, we do some
refactoring, but I think my boss & the rest of the team's early decisions
were good ones with forethought, so I havent' run into that problem too
much.
Naw, thanks again for letting me know,
I had no clue that property existed.
----- Original Message -----
Sent: Wednesday, July 12, 2006 3:02 AM
Subject: RE: [flexcoders] viewstack design pattern
issue
I didn't hear about
it, I just kind of came across it while working on a runtime xml double
binding solution for a future product. I tried various events but not worked
as well as that one. I only used 1.5 for about 2 months while I was
learning flex so can't comment on that one. Agreed that when you need
immediate updates it would not work, I guess everything has it's
limitations.
Do you think (any
sorry for being off topic) that when you have the same events being used in
multiple controls of the same, type that it is better to refactor that into a
separate extended control? For example buttons on a menu. If they all use the
same style, height, language tags for tool tip, and text, and all call the
same command controller. Ive been pondering this for a while, and though it
is easier to code, and cleaner to look at I thought it would bloat the final
product (I try not to go to far with regards to this stuff). What are your
experiences on this?
Im thinking for
maintenance it has to be better, but if a little change is required on a
component that uses it, you cant change that extended one, and would be
required to create another extended component, resulting in multiple extended
components all almost the same. I hope the question makes
sense.
So to extend or not
to extend? :)
back to
topic:
I wasnt questioning
your method here, just qualifying the way I do it, and found
myself questioning the methods Id chosen after reading your
post.
Jason
-----Message
d'origine-----
De :
[email protected] [mailto:[EMAIL PROTECTED]De la part
de JesterXL
Envoyé : mercredi 12 juillet 2006
02:02
À : [email protected]
Objet : Re:
[flexcoders] viewstack design pattern issue
3 things.
First, valueCommitted in Flex 1.5 does not always fire
at the appropriate
times. For example, it doesn't fire when you chang
text in a TextInput
field. Also, it doesn't fire when you hit the enter
key. Sometimes, these
are desirable events to set data. It only works
for focusOut.
So, using the default components in Flex 1.5, I have to
do:
<mx:TextInput id="my_ti" enter="doCommit()"
focusOut="doCommit()" />
function doCommit()
{
myVal =
my_ti.text;
}
vs. using an extended component that makes both
events into 1:
<view:ATextInput id="my_ti"
valueComit="doCommit()" />
See?
Second, in Flex
2, it doesn't fire until the focus goes to another control.
Finally,
neither work with change. Change is a valid event because
sometimes you
need immediate updates. ValueCommitted / ValueCommit is a
boilerplate
event. I wasn't aware of it in either versions of Flex until
you
mentioned it. It seems as long as you don't need specific events, it is
a good catch all to use. :: shrugs :: Where did you hear
about?
----- Original Message -----
From: "Jason Hawryluk" <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]ups.com>
Sent:
Tuesday, July 11, 2006 6:03 PM
Subject: RE: [flexcoders] viewstack design
pattern issue
I don't quite understand how that could be less
code?
>From the doc's
"The valueCommit event is triggered when
a user completes data entry into a
control or the value of the control is
changed programmatically"
It has nothing to do with the change
event.
>From what I gather valueCommit was designed for just that
purpose. The
focusout + enter solution only works for some
scenarios.
As every control has a valueCommit, and it's suggested to
use valueCommit
when validating the data
entered.
jason
-----Message d'origine-----
De : [EMAIL PROTECTED]ups.com
[mailto:[EMAIL PROTECTED]ups.com]De
la
part de JesterXL
Envoyé : mardi 11 juillet 2006 18:25
À : [EMAIL PROTECTED]ups.com
Objet
: Re: [flexcoders] viewstack design pattern issue
:: shrugs :: Never
used valueCommitted. Using change for a TextInput for
example causes
problems because sometimes updating every key stroke causes
updates to
happen too fast. focusOut ensures it only happens when the user
is "done"
with the TextInput. You can use enter, as well.
We have a component
in our apps that extends TextInput, and conslidates both
enter &
focusOut into a "valueComitted" event as it were. Less
coding
anyway.
----- Original Message -----
From: Jason
Hawryluk
To: [EMAIL PROTECTED]ups.com
Sent:
Tuesday, July 11, 2006 11:47 AM
Subject: RE: [flexcoders] viewstack
design pattern issue
JesterXL ? Focusout changes data on the
textinput ? What if the user just
get focus into the textinput, and does
not change it?
I'm using valueCommit on all components. Is focus out
better for this ?
Hank: Are you opening the same data multiple times,
and you require that
related screens are observed ? I.e. update one, the
other updates kind of
thing?
Jason
-----Message
d'origine-----
De : [EMAIL PROTECTED]ups.com
[mailto:[EMAIL PROTECTED]ups.com]De
la
part de JesterXL
Envoyé : mardi 11 juillet 2006 15:46
À : [EMAIL PROTECTED]ups.com
Objet
: Re: [flexcoders] viewstack design pattern issue
We use a dirty
flag. We set it to true whenever some data changes. This
data change is
detected by registering for control events that change data
(focusOut for
TextInput, selectedDate != null for DateChooser, etc.). These
events
change the data if applicable, and set the dirty flag.
Other
components can bind to the dirty flag and take appropriate
actions,
whether via a getter / setter or the binding tag.
-----
Original Message -----
From: hank williams
To: [EMAIL PROTECTED]ups.com
Sent:
Tuesday, July 11, 2006 9:20 AM
Subject: [flexcoders] viewstack design
pattern issue
I have a viewstack (actually a tabnavigator) that has
screens with data that
comes from a server.
I want to trigger the
screen to go out and get new data only when:
1. the user enters the
specific viewstack sub screen, and
2. when the data for that screen is
dirty, based on some bound data that
has triggered a change
So as
I see it, the screen needs to keep track of when the bound field
has
changed which should cause a call to get fresh data. This means that
the
screen must keep track of whether it is dirty.
This seems like
it must be a common scenario, and I am curious about how
other people
handle it. My first instinct was to create a base screen class
from which
all my screens could inherit. But I am curious whether cairngorm
already
handles this or whether this is even part of the scope of what it
tries
to cover. I am not using cairngorm right now, but may in the future.
But
understanding what is and isnt in its scope is important for me to
start
to understand.
So anyway, this is a 2 part
question.
1. how do people handle this situation
2. does cairngorm
have some pre-packaged solution to this
issue.
Hank
--
Flexcoders Mailing List
FAQ: http://groups.yahoo.com/group/flexcoders/files/flexcodersFAQ.txt
Search
Archives: http://www.mail-archive.com/flexcoders%40yahoogroups.com
Yahoo!
Groups Links
