Closed marcelocenerine closed 6 years ago
Your benchmark seems to show that performance was slightly better before the PR, no?
sorry @julienrf , Red line is before #444 Blue line is after #444 (which is what this PR aims to revert)
Ah, thanks for the clarification!
Reverting a small change I made here. It was not needed for the purpose of that PR (turns out the anonymous class is only used internally) and can slightly increase the running time of
diff
andintersect
. This is likely caused by the fact thatMap.withDefaultValue
wraps the default value in a function causing an additional dispatch every time a key is missing.