svelte

This is very WIP. The idea is to refactor scheduling to make it batch-centric — at the moment we have this weird impedance mismatch where effect dirtiness is 'global' (i.e. is a free-floating function, that may or may not push an effect root to the global array, and the effect's dirtiness is a flag on the effect itself), but traversal and flushing happens inside . This is an artifact of how the scheduling logic evolved to accommodate async reactivity, but it turns out to add complexity in various places. The thesis of this work is that if we mark effects _within a batch_, and then only do the walk-upwards-until-we-find-a-root bit when we're about to do the traversal, we will be able to simplify/robustify things a bit without negative performance impacts. We're a ways out from that: the first step is to prevent any cases where an effect is being scheduled outside of a batch, which is easier said than done. Working on it Before submitting the PR, please make sure you do the following [ ] It's really useful if your PR references an issue where it is discussed ahead of time. In many cases, features are absent for a reason. For large changes, please create an RFC: https://github.com/sveltejs/rfcs [ ] Prefix your PR title with , , , or . [ ] This message body should clearly illustrate what problems it solves. [ ] Ideally, include a test that fails without this PR but passes with it. [ ] If this PR changes code within , add a changeset (). Tests and linting [ ] Run the tests with and lint the project with