Enterprise Client Community

Ah, missed that one (...users/developers never read documentation...).

...anyway: you are definitely sure this works? Tried with some value, followed the code to verify that my parameter indeed makes it into one of the static variables of PageBrowser (so guess my parameter setting is correct), but still no effect is visible.

We have a combo field.
Click on icon opens the list - OK.
Another click on the icon should close the list. It does so indeed, but when you look a bit more precisely, the list is first closed, then reopened, and then closed again.

It looks like a field with focus is slightly shaded with a yellow color.
Where does this behavior come from, can it be influenced? (e.g. define different color, or use different border instead of background color)

Hi,
modal popup with position "centered" open at strange positions.
Reason is a "new Point(top,left)" in findLefttopReferencePoint() which should be reversed.

regards
Manfred

Hi,
any plans on this feature? looks like it did not make it into todays update.

Manfred

The idea for (5) comes from http://code.google.com/p/search-as-you-type/. Still not exactly sure what one expects.

Think have catched (4): comes when two items have identical contents - this is why we would like to display several columns, and have an id available.

(2) not sure how to best continue here, maybe we just copy your code, or we can find some useful points where to plugin our enhanced behavior.

While the new autocomplete seems to do what is needed - thank you so far

-, we would like to have some enhancements

(1) waiting time: is currently fix at 1 sec; should be more flexible, either with many more parameters for the component, or e.g. a plugin mechanism where we can provide a class that can handle the strategy as needed. Currently we want that after initial list is made available, the waiting time should decrease (100/200 ms) since users expect fast reaction of newly typed characters; our server can handle this well since the initial list is cached anyway and can be narrowed quickly.

(2) design of the list: We would like to have more control on how the list is displayed; simple thing to specify the list width; then allow several columns; or still better completely flexible design where we specify the list, and search hints or clickable links, arranged somehow as needed when e.g. nothing or too many items are found, etc.

(3) have an ID in the list data, and when an item is selected, make "selectedId" available as a dedicated property in the following (flush) roundtrip so that we can determine more clearly what was selected, also to differentiate that something was selected from the list vs. normal text input. Currently we hide some special characters in the list elements which is rather clumsy.

(4) sometimes keyboard interaction seems to be lost; arrowdown keys do no longer scroll the list

(5) when scrolling through the list with arrow up/down keys, the list contents is copied to the field. However the insertion point (caret) should not be set to the end, but rather stay where it was before, and perhaps select from there to the end, so that when user continues to type, it is as if nothing has been taken from list. Some programs even show the newly copied text in gray, and continued typing overwrites the gray text unless enter/tab/rightarrow has been pressed.

see attached screenshot;

User clicks on button, but the selection list should be positioned below the entire row of fields, since all these fields are filled when selecting an item - precisely at the left bottom of the leftmost field.

So we could either give the relative offset from the button (assuming we can calculate this correctly?), or preferably specify a different control (in this case would be the field for "länge"

to be used for your position calculation.

(how to specify a control - provide its ID (..how do we readily obtain it..), or give for each element a flag with which it can claim to be the owner of the new popup (and lets hope that only one control will claim so - this is something similar to the focus stuff).

For a combofield, the value help popup opens right below this field (...as is correct for most cases...).

We have a special situation when we want to open the list at a different location. How could this be achieved?
While it is possible to give an absolute position, this does not help here since we do not know the current location of our target control.

Hello,

works great on "field" controls, and "combofield",
but not so nice on "formattedfield" where the background text disapperars only after the field loses the focus - which is not what one expects. And when the text is cleared again, it does not reappear.

b.t.w. the same problems apply to the "mandatory()" feature.

regards
Manfred

Sorry, but this did not really help.
It even introduces more problems: apparently the direct childpanes of layeredpane do not conform to their own width/height attribute, but always take the height of their parent.
So even if t:row is not rendered, the surroundinf pane is always there preventing mouse and keyboard access to controls (which is desired behavior when a popup is open, but not at all when there is no popup).

The other problems regarding resize etc. are not solved, either.
regards
Manfred

when a popup is displayed, wo want to gray out the background of the page, see attached screen shot layer.jpg.

in a first attempt, we tried to implement this via layered panes, modifying the wp_workpagecontainer.jsp as attached. This has the issues described below - you can either solve them, or take this as a feature request and implement e.g. a new flag in ModalPopup named something like ("make-it-behave-as-in-our-screenshot"), or do both.

problem1: inserting the layered pane between row-g_23 and pane-g_24 mixes up the height calculation. While pane-g_24 seems to be resized correctly, the pane resulting from rowinclude-g_28 always gets the height of the "previous" parent height! So initially its height is nearly 0. Just sizing the PageBrowser window some pixels with the mouse brings it to an acceptable size, though it still follows the parent sizing one resize step behind.

problem2: our gray pane is a second layer pane-g_22_gray which has a semi transparent color and has a rendered property which is changed accordingly (...we have subclassed WorkPage and have a method getHasPopup() ...). When this flag is switched on, the layer appears, but again only after the window has been slightly resized. What does not work is to make this layer invisible again, by setting its rendered property to false. While it falls out of LAYEREDPANEElement, it is not taken out of the swing components, so continues to hang around. It is no more part of the resizing business, so when PageBrowser is made bigger, the shade stays around in the size when it was taken out of rendering.

problem3: there is probably some issues when the "rendered" flag is changed in the middle of a jsf roundtrip; this flag is queried rather often during the jsf life cycle phases, and it did not work when the flag would become true only in a later phase. What is the exact specification on when the rendered flag can be switched from false-to-true or true-to-false.

regards
Manfred

Hi,
just remarked the number of roundtrips that are generated when navigating in a list with the scroll bar.
In my example I have a value help list of 130 elements, and the standard idvalue help popup displays around 12 of them.
While it is probably a good idea to fetch only one page at a time, it is very strange that every single scroll click (even one line forward and back) causes a roundtrip and fetches each of the 12 items that are currently visible. And the way jsf is implemented this roundtrip checks every single property of the server on a possibly rather large screen; even if the number of transferred data may be small due to delta management, this causes lots of server load.

It would be much more appropriate if the client requested pages of data as needed, and keeping adjacent pages in memory does scrolling of its own, until it comes to a page that has not yet been requested.

And for that matter if we have just a few 100 entries, and an efficient way to transfer them (...thinking of new arraygrid...), it might be as efficient to transfer them all and have no scrolling at all.

regards
Manfred

sorry, i missed the word "constructor", so my answer was rubbish.

We have actually also had this problem, and found no simple solution - see http://www.casabacdemo.com/forum/posts/list/522.page - , and eventually we had to restructure this part of our application so that we know in advance if a particular page can be opened, and if not we do not include the jsp file in a rowdynamicinclude tag, so no constructor is ever invoked by the framework.

Manfred.

Hi,
you can make your bean a subclass of WorkpageDispatchedBean, and you inherit methods like

Code:

getWorkpageContainer().closeWorkpage(getWorkpage())

Manfred