FWD » Core Development » User Interface

Also I looked over P2J project code in order to understand how to implements P4 widgets on P2J and I found a class com.goldencode.p2j.ui.RectangleWidget. Seems that someone already started the implementation for RECTANGLE.

That is some stubbed out support on the server side. I think it is not complete. There is no client-side implementation yet.

Do you have a manual or a tutorial for developers which explains how to implements P4GL widgets on P2J?

Unfortunately, no. The best thing to do is write a simple testcase that uses a DEFINE BUTTON and uses that button in a frame. Then convert it and review the generated frame definition/business logic. Then follow the code in the runtime on both the server side and the client side for the button widget.

Today I did a lot of RECTANGLE tests using the customer's P4GL environment.
Also I converted a small procedure working with BUTTON and FRAME widgets.
Bellow are the results of my researches after this tests.
Please let me know if what I stated bellow are true or false.

1. RECTANGLE widget looks like BUTTON widget except:

- Rectangles are readonly widgets.
   - Rectangles does not accept user inputs except mouse events to move or resize.
     ChUI implementation does not support mouse events so we can state that rectangles have no inputs.
   - Rectangles cannot gain the focus. We cannot navigate to rectangles using key navigation, no tabstop properties are defined for rectangles.
   - No action triggers could be associated with rectangles.
   - Optional rectangle could have a border (edge).
   - Rectangles can be used only on the FOREGROUND or BACKGROUND of a FRAME widget.

2. Server side implementation reside in package: com.goldencode.p2j.ui
Example:
@LegacyResource(resource = LegacyResource.BUTTON)
public class ButtonWidget extends ControlEntity

3. Client side implementation reside in package: com.goldencode.p2j.ui.client
Example:
public abstract class Button extends AbstractWidget

4. ChUI implementation reside in package: com.goldencode.p2j.ui.chui
Example:
public class ButtonImpl extends Button

5. The following classes are used in creation process of widgets:

- com.goldencode.p2j.ui.client.ChuiWidgetFactory

- com.goldencode.p2j.ui.client.UiUtils 

Example from: UiUtils.reconstructComponent(com.goldencode.p2j.ui.ComponentConfig)
         case ComponentConfig.BUTTON :
            clazz = Button.class;
            break;

6. ComponentConfig - is used to store widget attributes.

7. In order to generate Java code for RECTANGLE widgets we have to write some rules.

I think this is generally correct. Since we will be implementing the GUI as well, please do plan to support the mouse events.

I think it is possible to support mouse events on some ChUI terminals. We have not implemented that support yet and we don't plan to do it soon.

Doing more test cases conversions I think that I found a bug in conversion process. Let me explain:
Trying to convert procedures using BUTTON widgets I found:

1. Statements like: DEFINE BUTTON butt LABEL "Test". are converted without errors.

2. Adding other options for button like SIZE phrase DEFINE BUTTON butt LABEL "Test" SIZE 10 BY 1. are not converted.
An exception is throw (see StringHelper.sweep):

EXPRESSION EXECUTION ERROR:
---------------------------
tw.graft("set_widget_attribute", null, lastId,
"attribute", setter,
"instance", wname,
"vartype", atype,
"varvalue", aval)
^ { No replacement value found for key: instance }

Doing some debugs I found that "wname" is null. Here is the stack trace:

Conversion [Java Application]
com.goldencode.p2j.convert.ConversionDriver at localhost:57613
Thread [main] (Suspended (breakpoint at line 427 in TemplateWorker$Template))
TemplateWorker$Template.graftAt(String, Aast, Aast, int, Object[]) line: 427
TemplateWorker$Template.graftAt(String, Aast, long, int, Object[]) line: 309
TemplateWorker$Template.graft(String, Aast, long, Object[]) line: 249
CE4564.execute() line: not available [local variables unavailable]
Expression.execute() line: 323

Expression.execute() try to execute:

tw.graft("set_widget_attribute", null, lastId,
"attribute", setter,
"instance", wname,
"vartype", atype,
"varvalue", aval)

After the call to CE4564.execute() the "wname" is null which means "instance" = null

The rule for SIZE phrase is already defined in rules/convert/frame_generator,xml line 1583: <function name="process_button_options">

The problem is on line 1602: <action>wname = node.getAnnotation("javaname")</action> where node.getAnnotation returns a null value for "wname".
I fixed this bug by replacing the "node" with "butdef" as follow: <action>wname = butdef.getAnnotation("javaname")</action>

Having this fix in place all defined options (attributes) for BUTTON widget are converted without error.

I found the same bug for RECTANGLE widget as well in line 1711. I provided the same fix in line 1711 and the RECTANGLE widget attributes are now converted without error.

Code Review 0416a

This is really good. Have you tested it on all 3 ChUI client types?

1. In frame_generator.xml function process_rect_options please change variable names butdef and bchild to something more rectangle-related (e.g. rectdef and rchild).

2. In BaseConfig, the fgcolor and bgcolor should probably be Color instances, not int.

3. In BaseConfig, why are the getters/setters for the new color attributes protected instead of public?

4. The BaseEntity copyright date needs update.

5. In BaseEntity, the setBgcolor(int64 bgcolor), getBgcolor(), setFgcolor(int64 fgcolor) and getFgcolor() should be implemented instead of just throwing UnsupportedOperationException.

6. In BaseEntity, the getTooltip(), getWidthPixels and getHeightPixels methods are missing.

Yes I tested on all 3 ChUI client types in Linux and in Windows OS.

Code Review 0430a

Everything looks good. The only open question is this:

3. In BaseConfig, why are the getters/setters for the new color attributes protected instead of public?

Code Review 0430b

It looks good. Please get it regression tested (both conversion and runtime).

I restarted the regression tests both CTRL-C part and MAIN part.
This time CTRL-C part took around 58 minutes.
Also are less failures compared to previous tests.