A different way to bind

[treeowl]

I realized it's possible to write

SeqT q >>= f = SeqT $ fmap {- a bunch of gunk -} q

This automatically takes care of making >>= entirely strict in its first argument for strict monad lawfulness as mentioned in #34. Is it a good thing or a bad thing otherwise? Clearly any rewrite rules might need to be adjusted to accommodate it.

One potentially nifty thing: (s >>= f) >>= g becomes, essentially, fmap (gunk g) (fmap (gunk f) s), making it subject to an fmap/fmap fusion law we could write for Queue.

[treeowl]

The alternative definition I came up with:

  SeqT q >>= f = SeqT $ flip fmap q $ \m -> m >>= view
    (pure Empty)
    (\h t -> toView (f h) >>= view (toView (t >>= f)) (\b s -> pure (b :< (s <|> (t >>= f)))))

How ugly is that? We could even dig into that toView (t >>= f) a bit, for extra pain!