support #to_n about bowser

[jgaskins]

I wondered if it should provide that, but I figured that most of the time you'd need it. you'd be operating in JS already, so you could do element.native in JS to get it.

It's entirely possible I've overlooked a use case for this, though. Can you explain a bit about what you're trying to do?

[fazibear]

For example, I have a method:

def inject_to_body(elem)
end

And I want to support native elements, opal-browser, and any other libs that wraps native elements somehow. So the best solution is to convert elem to native.

def inject_to_body(elem)
  elem = elem.to_n unless native?(elem)
  # inject to body
end

#to_n is some kind of opal standard I think to convert wrapped object to native.

[jgaskins]

It's not a bad idea. I was just trying to understand the use case. I haven't looked too closely at the Native mixin yet, but it may bring in more functionality than necessary. I'd rather just add the to_n method first until we can see if there's enough functionality needed to justify including the mixin.

I wanted to keep a similar API to opal-browser to make transitioning to bowser easy — so far, in upgrading Clearwater pre-1.0-beta apps (which used opal-browser) to the 1.0 beta (which uses bowser), I've only ever had to change the element attribute of the app itself, which I consider pretty good. If you need a bit more API compatibility, we can definitely look into it.

[fazibear]

BTW Why you just don't use an opal way to make a wrapper ? I mean include Native, alias_native etc ?

[jgaskins]

I prefer to remove dependencies if possible. I don't know if native is already loaded (likely, but not guaranteed), but if it's not I'd prefer not to load it for two reasons: adding it expands API surface and increases the JS payload size.

Expanding API surface isn't exactly dangerous, but it can be problematic in that people begin relying on other methods that just happen to be there because they came with the mixin. If we then swap out the mixin later for any reason, it may cause unintended regressions.

In regards to payload size, bloated JS payloads are the reason I switched from opal-browser to Bowser. Is the convenience of being able to say alias_native worth loading an entire module? I'm not sure. I'm also not sure if this one is even something to worry about, since it's entirely possible that native is likely to be loaded already by another dependency of any given Opal app.

[jgaskins]

Fixed in cd9e31a (Element) and f5fb5a8 (Event)