World: r3wp

Also, I have reports of this error ... not sure where it is coming 
from 

make object! [
   code: 902
   type: 'internal
   id: 'stack-overflow
   arg1: none
   arg2: none
   arg3: none
   near: [unless find [object! block!] type?/word get]
   where: 'into-widget
]

well, I mean I know it is coming from rebgui-edit.r and the 'into-widget 
function which has a recursive find ....

but not sure of the circumstances of how it is being triggered.

The user thinks that they tabbed on something which is consistent 
with this function's function.

Well, in the meantime I have wrapped the into-widget with an error 
try and return none if there is an error.

Unfortunately this did not help .. just shifted the error elsewhere. 
 Now getting

make object! [
   code: 902
   type: 'internal
   id: 'stack-overflow
   arg1: none
   arg2: none
   arg3: none
   near: [all [
           face/show?
           find behaviors/tabbed face/type
           not find face/options 'info
           face
       ]]
   where: 'tabbed?
]

Is it possible to rewrite the tabbing functions so that it doesn't 
use recursion?

I have recreated the stack overflow error on tabbing here http://rebgui.codeplex.com/WorkItem/View.aspx?WorkItemId=23092

Basically a table with focus, and the only widget inside a panel 
will cause a stack overflow as it tries to tab out looking for the 
next widget that can accept focus.

Good test case. I'll look at it over the weekend.

It would also be good to be able to tab into a table ... if that's 
feasible.

Since we no longer have access to the geekisp trac and issues database 
... would all of those who remember any existing issues please enter 
them into the new issue tracker.
http://rebgui.codeplex.com/WorkItem/List.aspx

Why do you no longer have access to it?

Because I presume Jaime stopped paying for it ...

If we can get the data, we can suck it into Qtask, then it will be 
there.

And Graham, this statement is not TO YOU, but rather to anyone here 
would like to "help"

it is paid service? how much does it cost? Qtask offer is fair of 
course, but if guys are used to trac service, and it is not expensive, 
maybe we could sponsor it ...

Pekr, agreed.  You are alwys welcome though to upload the data to 
Qtask's tasks, at least you guys will never lose it.

I don't mind where the source resides, as long as:

	a) It doesn't cost me anything
	b) It's accessable via standard SVN clients
	c) It's simple


Jamie very kindly sponsored the first 3 years of hosting on geekisp, 
and when that ceased Graham created a project on CodePlex (which 
I assume offers free hosting?). That seems to be working OK so far, 
but I'm happy to move across to Qtask if it meets above three source 
hosting requirements.

Uploaded build 120 which addresses the stack-overflow issue (and 
as a by-product drastically cuts the number of calls to next-field/back-field 
when tabbing) ... please test.


I'll also echo Graham's call for issue tracker submissions, as I'm 
now prioritzing fixes based on what's in there.

Still get a stack overflow with my test case .. http://rebgui.codeplex.com/WorkItem/View.aspx?WorkItemId=23092

Pekr, it cost $5 a month on geekisp.  But codeplex is better .. as 
it is free, and it is much easier for anyone to join, and add comments. 
 However i miss the diff that geekisp had ... unless it is on codeplex 
as well somewhere.

I had a trac also on geekisp, but it was plagued by trac spam, and 
the site owner wasn't willing at that time to upgrade the trac to 
block spam.  So the only way to stop spam was to block all comments 
.. so I abandoned it.

there are also websites that offer free hosting (with issue tracking 
and other typical stuff ) of OS projects for distributed vcs-s Bazaar, 
Mecurial and Git ... I stopped using svn and went to bazaar and I 
like it  ... there is comparison of various such sites here http://en.wikipedia.org/wiki/Comparison_of_free_software_hosting_facilities

for bazaar there is https://launchpad.net/... for svn you also have 
code.google

Uploaded build 121 with a completely rewritten tabbing system. Tabbing 
into/out of tables and text-lists is now supported (you need to copy 
new behavior/tabbed and behavior/cyclic values across into %ui.dat). 
Will add remaining widgets next build plus ability to exit a cyclic 
widget (e.g. press a key to exit a group-box). What keystroke/combo 
do folks want to map the later functionality to?

Can I use this new tabbing system on earlier builds?

Yes, but make sure you use the updated versions of both %ctx-rebgui-edit.r 
and %set-focus.r. I've also uploaded build 122 to correct a typo 
in back-field.


With regards to the question as to "what key should escape from cyclic", 
I've come to the conclusion that the best way of handling this is 
to just "tab through" cyclic widgets, so:

	f1: field
	panel data [f2: field f3: field]
	f4: field

would tab as f1, f2, f3, f4, f1, etc. tab-panel would operate as:

	f1: field

 tab-panel data ["a" [f2: field f3: field] "b" [f4: field f5: field]]
	f6: field

with a tab sequence of f1, f2, f3, f4, f5, f6, f1, etc


This way we don't have to worry about what keystrokes to change tabs 
or exit grouping widgets. What do you think?

Excellent ... at present I have to trap enter on the last field of 
a panel to set focus to the next panel.

stack overflow issue persists :(

In build 122, in tour.r  and in List/Tree the right hand tree is 
now missing.

In List/Table, if you click on the table,  you can tab in and out 
now if the Look & Feel/Behaviors/Tabbing .. if you add table and 
text-list

is updated as you said above.

The group-box is screwy though ... no longer contains all the fields

The tabing changes are still a WIP, it'll take a couple of builds 
to get 100% ... but at least I can grok it now!

re: list/tree missing ... a half-implemented change that I had to 
pull out several builds ago.

Uploaded build 123 with tab-thru implemented as described above (except 
that tab-panel doesn't cycle each tab). Also fixed initial text case 
(stack overflow) and a long-standing text-list bug (no cursor control 
when select-row used).

thanks .. will check it out today.

Uploaded build 124 with new request-calc requestor (as per discussion 
from 20-Apr-2009).

Ashley and others, before I go about it,  do you have any experience 
with rebgui under the present browser plugin?

There shouldn't be an issue.

THx

If I have tree data like this [ A [  B B B ] ]


how can I tell which of the B's that I am clicking on?  Do we need 
some type of index or picked?

You need a unique id, like the index.

I have unique ids .. but putting them into the tree looks ugly :(

Ashley I think we discussed this before .. is there a way to have 
a hidden id so we have both a display element in the tree, and a 
hidden id.

Yes and yes ... but the tree widget really needs to be totally rewritten 
to support that.


If I have tree data like this [ A [  B B B ] ] ...how can I tell 
which of the B's that I am clicking on?

 Isn't this the same issue with drop-lists, edit-lists, text-lists, 
 tables, radio-groups and check-groups? In what case does it make 
 sense to have user options (text strings) that appear identical but 
 have different "real" ID's behind the scenes?

HTML selects handle this by having different display and value attributes.

Ashley, it makes less sense to the user to display meaningless 10 
digit iD codes ...

But if you're asking them to select an option and the option display 
text is the same how do they know which to select? And if it doesn't 
matter then why the need to display multiple options with the same 
display text in the first place?