Strictly speaking, the schema is the drawing of the data model. We could rename the schema/ directory to model/. And the entities/ directory contains the logic/ of the application. That would get us to model/, logic/ and views/, which is even closer to something understandable without documentation. Maybe the contents of hooks/ could be moved to logic/ at the same time. Of course the old directories would be supported and warnings issued for the next two versions. | |
| priority | normal |
|---|---|
| type | enhancement |
| appeared in | <not specified> |
| done in | <not specified> |
| load | 0.500 |
| load left | 0.500 |
| closed by | <not specified> |
[register] [login]
#506805 add cube instance version string to dump filename
Comments
imo: * keep hooks * sobjects merged with views * pre|post create files move to model
After discussing this further with sylvain, I realized that the actual separation depends on where the code runs. There is code that will run on the server side, code that will run on the ui side and code that will run on both (logic). Depending on where the code runs, imports are allowed or not. Importing cubicweb.web for example, is allowed only in the code that runs on the ui side, but not on the server side or both sides.
Taking this into account, I think we should rename things to have:
Then we add tests that enforce the import restrictions described above (and we document these restrictions).
a lot of logic happens in hooks in my experience.
strong -1
I agree with acampeas: please defer this to CW 4.0 (if you insist on renaming stuff).