Hi, figured we could actually use a changelog now:
completed() method would result in absurd numbers
of calls overall as it walked down the tree. We now, instead, keep track as
we bubble up changes, so the computation is limited to when data changes and
to the depth of that one branch, instead of every node. (Plus, we were already
incurring this cost, since we already bubbled out changes.)Hi, figured we could actually use a changelog now: ## 1.1.4 2017-04-21 * Fix typo in package.json ## 1.1.3 2017-04-21 * Improve documentation and limit files included in the distribution. ## 1.1.2 2016-03-15 * Add tracker group cycle detection and tests for it ## 1.1.1 2016-01-29 * Fix a typo in stream completion tracker ## 1.1.0 2016-01-29 * Rewrote completion percent computation to be low impact– no more walking a tree of completion groups every time we need this info. Previously, with medium sized tree of completion groups, even a relatively modest number of calls to the top level `completed()` method would result in absurd numbers of calls overall as it walked down the tree. We now, instead, keep track as we bubble up changes, so the computation is limited to when data changes and to the depth of that one branch, instead of _every_ node. (Plus, we were already incurring _this_ cost, since we already bubbled out changes.) * Moved different tracker types out to their own files. * Made tests test for TOO MANY events too. * Standarized the source code formatting