[wruza]

Fast diffing of objects may be seen as rarely needed to a regular user, but it’s very important part of handling vdom/vnode diffs. Recently I researched few lesser web ui libraries for my own use and found out that many of them do not diff vnodes correctly, e.g. in case of RegExp or Date arguments. I don’t remember exact library names vs flaws, but my conclusion was that almost all of non-actively supported libs had it in one way or another. If you want to explore, check that first, because it’s a surprising pain point.

Seems that this particular library is aware of these potential issues (which motivated me to write this comment): https://github.com/AsyncBanana/microdiff/issues/2

[nathcd]

But vnodes have a predefined shape, whereas this library is for diffing (relatively) arbitrary objects, right? If your data isn't arbitrarily shaped, I would assume you'd get much better performance by just implementing a simple direct diff, no?

[wruza]

Yes. I should have noted for clarity that vdom diff is a special case, not a general one like that of subj. But their similarity reminded me of the experience.

[nathcd]

Ah that's my mistake, I read more into your comment than you wrote. Sorry!

It's a good point about edge cases in diffing. Dynamic value diffing in dynamically-typed languages is an interesting problem to think about.

[eyelidlessness]

And, often but probably not always, using a class instance rather than a POJO. Seems unintuitive (function call overhead), but it can signal to the JIT that the object’s type is relatively stable.

[bikeshaving]

Are these RegExps/Date objects passed as props? I’m not sure why a VDOM algorithm would have to handle these types of objects.