Kefir was changed from a dependency to a peer dependency to avoid the possibility of having multiple versions of Kefir in a project.
Switched to using infestines instead of ramda. This means that the lift
function has now faster, but also much more limited currying support. Note that
you can implement lift with Ramda's currying with a 1-liner:
const lift = fn => R.curryN(fn.length, (...xs) => K(...xs, fn))Previously duplicates were skipped with Ramda's equals. This is ideal in the
sense that unnecessary updates are eliminated. Now duplicates are skipped with
Ramda's identical. There are two main reasons for the change:
-
Ramda's
equalsis very slow, because it uses Ramda's slow currying technique and because it handles arbitrary object graphs. When either dealing with a large number of properties or with properties that are large objects, Ramda'sequalsbecomes a bottleneck. -
In practise, when embedding properties to VDOM that are computed with lenses from a state atom,
identicalis enough to eliminate almost all unnecessary updates and can be implemented so that it works several orders magnitude faster than Ramda'sequals.
In cases where you really want a more thorough equality check, you can
explicitly use Kefir's skipDuplicates.