Hi @leeoniya, thanks for the contribution! Rollup seems like a good fit for dygraphs and I'd like to incorporate this change.

My main question is whether this breaks any of the exported APIs. A few ways to check this are:

• Build dygraph.js and open the various pages in tests/.
• In particular check tests/exported-symbols.html.

open the various pages in tests/.

i opened a few and they worked ok, but since many tests require interaction or are subtle, i dont even know what to look for. for me to manually verify that all features work exactly the same with no automated testing is pretty much impossible. i think this is something only you can do :(

In particular check tests/exported-symbols.html.

this looks fine.

I pulled this branch and opened a few of the tests. This seems good. What's required to get the tests to pass?

What's required to get the tests to pass?

quite a bit more, actually.

after some tweaks to avoid utils mutation, i got auto_tests compiling/concating via Rollup.

when opening runner.html in Chrome i now get passes: 243, failures: 32. most of the failures are Cannot read property 'length' of undefined which seems like an easy fix. (can you take a look?)

once we get all tests passing, i'll try to hook up istanbul for coverage and replace phantomjs with puppeteer [1]. i don't know too much about travis and will be pretty much useless for any non-trivial tweaking of remaining shell scripts, so hopefully you can take over from there.

[1] https://www.npmjs.com/package/mocha-puppeteer

i'm going to keep digging. i'm starting to understand a bit better how the tests are structured.

passes: 330   failures: 0   duration: 8.23s

ok, i got all auto_tests passing on Windows 10 in Chrome and Firefox. the additional changes that were required:

• create a global Dygraph.setGetContext for testing & Proxy injection
• test actual, standalone built lib (don't re-bundle Dygraph into tests)
• use hackish access to RangeSelectorPlugin constructor
• fix a few off-by-one fuzzy pixel errors

hey @danvk, is this project effectively abandoned?

Sorry @leeoniya. Would you be interested in becoming an owner? #727

unfortunately i have enough of my own open source projects to maintain. the dygraphs codebase is too large for someone unfamiliar with it to start maintaining it without a very significant time investment.

like most people, i probably only personally use 20% of the functionality/options, so it's hard to justify making such a commitment to study the whole thing well enough to make well-informed choices on a project-wide scale. if you don't have time to review PRs (even as simple/ready-to-go as this one), then you certainly won't have time to answer many many questions about the details of the codebase and intricate interactions between options, etc.

@kraliq233 i'm now using uPlot and have no further plans to pursue this PR.