[ XText editor for OCCI Extensions ] Save reformats the code about ecore

[wpiers]

This is due to the fact that we store models in XMI, not in text. Thus the model is parsed/serialized on each save.

To avoid that, the solution would be to store models in textual syntax, in a standard XText mode.
This implies that the XText syntax must be flawless.
Also, storing models in text prevents the use of Eclipse CDO, but this might not be relevant for us.

[vincent-zurczak]

I don't know what is the interest of storing models in XMI files.
Maybe other members of the project need it.

A dirty workaround would be to hack the save process.

• Intercept the new serialized text, remove all the spaces and line breaks...
• ... (re)load the current text content, remove all the spaces and line breaks...
• ... and compare it (optimistic hypothetisis: the serializer does not modify elements order).

If it is the same, we do not write into the file.
And normally, it should be the same...

No matter how it is done, the XText editor looses its interest if it reformats all the code. 😢

PS: unlike the text format, I think almost no one will edit XMI directly.
So, if I had to choose one side, I would do generate XMI from the textual reprensentation. 🚸

[vincent-zurczak]

Then, there are elements I don't know much about... because I did not search.
As an example, maybe the XMI is used by the OCCIE modeler. How does it work when both editors are open?

[philippemerle]

In EMF, the default serialization format is XMI.
XMI files can be read by Eclipse tools we use, e.g., Sirius, Acceleo, etc.
Saving extensions in text format could be incompatible with Eclipse tools we use.
But saving extensions in text format allows one to add comments, identations, and so on.
So we should find a solution supporting text format but also compatibility with used Eclipse tools.