| null text generated in editor plugin.xml [message #1399069] |
Fri, 11 July 2014 21:30  |
Eclipse User |
|
|
|
I have discovered what appears to be a bug.
When I generate my editor code, my plugin.xml gets a number of null text fragments all over the place.
I ran in debug and found that this happens during the mergePluginXML in AbstractGeneratorAdapter.
What ends up happening is that the oldExtension.content is set to null here (see arrow):
else
{
// If the new match has the same non-null generation key.
//
ExtensionData newExtension = newExtensions.get(index);
if (oldExtension.generated != null && oldExtension.generated.equals(newExtension.generated))
{
// Set up the old extension up to pull the new content.
//
--> oldExtension.content = newExtension.content;
}
// Set up the new extension to block it being pushed.
//
newExtension.content = null;
I'm not sure about the full validity of the pattern here, but the root cause is that the regex used is not foolproof.
We have this regex:
EXTENSION_POINT_PATTERN = Pattern.compile("[ \t]*<extension[^>]+point\\s*=['\"]([^'\"]+)['\"].*?(?:id|class)\\s*=\\s*['\"]([^'\"]+)['\"].*?</extension>[ \t]*(\n\r|\r\n|\n|\r)", Pattern.DOTALL);
and the extension key is based on the extension point and the id (group 1 and 2 in the regex).
But I have the following in my plugin.xml which causes duplicate matches since the id used is not the correct one, since it picks up the id from the category which is not unique:
<extension point="org.eclipse.ui.newWizards">
<!-- @generated ba -->
<category
id="org.eclipse.emf.ecore.Wizard.category.ID"
name="%_UI_Wizard_category"/>
<wizard
id="com.castortech.iris.models.ecore.ba.process.presentation.ProcessModelWizardID"
name="%_UI_ProcessModelWizard_label"
class="com.castortech.iris.models.ecore.ba.process.presentation.ProcessModelWizard"
category="org.eclipse.emf.ecore.Wizard.category.ID"
icon="icons/full/obj16/ProcessModelFile.gif">
<description>%_UI_ProcessModelWizard_description</description>
<selection class="org.eclipse.core.resources.IResource"/>
</wizard>
</extension>
So what is the solution here? Is a regex really appropriate ?
Please advise.
Cheers,
Alain
|
|
|
| Re: null text generated in editor plugin.xml [message #1399980 is a reply to message #1399069] |
Sun, 13 July 2014 09:46  |
| Eclipse User |
|
|
|
Alain,
Comments below.
On 12/07/2014 3:30 AM, Alain Picard wrote:
> I have discovered what appears to be a bug.
>
> When I generate my editor code, my plugin.xml gets a number of null
> text fragments all over the place.
I see. I know someone else mentioned that a while back, but I've never
seen that...
>
> I ran in debug and found that this happens during the mergePluginXML
> in AbstractGeneratorAdapter.
>
> What ends up happening is that the oldExtension.content is set to null
> here (see arrow):
>
> else
> {
> // If the new match has the same non-null generation key.
> //
> ExtensionData newExtension = newExtensions.get(index);
> if (oldExtension.generated != null &&
> oldExtension.generated.equals(newExtension.generated))
> {
> // Set up the old extension up to pull the new content.
> //
> --> oldExtension.content = newExtension.content;
> }
>
> // Set up the new extension to block it being pushed.
> //
> newExtension.content = null;
>
>
> I'm not sure about the full validity of the pattern here, but the root
> cause is that the regex used is not foolproof.
> We have this regex:
>
> EXTENSION_POINT_PATTERN = Pattern.compile("[
> \t]*<extension[^>]+point\\s*=['\"]([^'\"]+)['\"].*?(?:id|class)\\s*=\\s*['\"]([^'\"]+)['\"].*?</extension>[
> \t]*(\n\r|\r\n|\n|\r)", Pattern.DOTALL);
>
>
> and the extension key is based on the extension point and the id
> (group 1 and 2 in the regex).
I see.
>
> But I have the following in my plugin.xml which causes duplicate
> matches since the id used is not the correct one, since it picks up
> the id from the category which is not unique:
>
> <extension point="org.eclipse.ui.newWizards">
> <!-- @generated ba -->
> <category
> id="org.eclipse.emf.ecore.Wizard.category.ID"
> name="%_UI_Wizard_category"/>
> <wizard
> id="com.castortech.iris.models.ecore.ba.process.presentation.ProcessModelWizardID"
> name="%_UI_ProcessModelWizard_label"
> class="com.castortech.iris.models.ecore.ba.process.presentation.ProcessModelWizard"
> category="org.eclipse.emf.ecore.Wizard.category.ID"
> icon="icons/full/obj16/ProcessModelFile.gif">
> <description>%_UI_ProcessModelWizard_description</description>
> <selection class="org.eclipse.core.resources.IResource"/>
> </wizard>
> </extension>
>
>
> So what is the solution here?
Is there more than one newWizards in the same plugin.xml?
> Is a regex really appropriate ?
Processing XML would be better, but it's much harder to avoid messing up
the whole file, i.e., reordering all attributes alphabetically which DOM
serializers are always so happy to do.
> Please advise.
Please open a bugzilla with a sample that reproduces the problem.
>
> Cheers,
> Alain
|
|
|
|
Powered by
FUDForum. Page generated in 0.50803 seconds
Home