Make the update function return the model. about relm HOT 3 OPEN

Your first suggestion was the way used in the very first versions of relm and I switch to the more idiomatic &mut Model and it will probably come back to be an attribute to self.

In any case, neither of these solutions solve the issue with efficiently updating the view.

Are you worried about performance?

One solution that I thought to produce a more efficient version is currently described in a comment.

This idea can be schematized as follow:

            *
          /   \
        A       A
          \    /
            1

where 1 updates the properties bound to the model attribute A.
This is the case of a condition: this will result in generating one call to set_property() instead of 2.
I believe the same logic can be applied to loops so that the view is only updated after the loop.

However, it is complex to implement, so it will probably take some time before I do it.

from relm.

Are you worried about performance?

No, i'm not worried about performance, i just thought that it was too magical and slightly hard to know was going under the hood.

from relm.

If you're not worried about performance, then I believe you want to view to be updated as a whole with the new model, right?
I think recreating the view each time is a huge cost, so it will degrade the performance in a visible way when creating complex view.

One solution (that is currently impossible because not implemented in GTK+ yet) is to switch to GTK Scene Graph. As the API is not yet available, I'm not sure how it will look, but that will fix this issue, I think, amongs others.

from relm.