When do I have to call flush() in wct?

[aro...@flux.io]

With Polymer 1.0, observers are now evaluated synchronously.  If I have an element whose behavior is driven by observers on attributes, and everything appears to work in synchronous tests, is there any reason to call flush()?  That is, is it necessary that I call it before/after each test?

[Daniel Freedman]

Generally you can get away with not using flush.

You may still need flush for testing element lifecycle callbacks in polyfilled browsers, or measuring DOM, but I expect very few cornercases like that.

On Thu, Jul 16, 2015 at 10:46 AM, <aro...@flux.io> wrote:

With Polymer 1.0, observers are now evaluated synchronously.  If I have an element whose behavior is driven by observers on attributes, and everything appears to work in synchronous tests, is there any reason to call flush()?  That is, is it necessary that I call it before/after each test?

Follow Polymer on Google+: plus.google.com/107187849809354688692
---
You received this message because you are subscribed to the Google Groups "Polymer" group.
To unsubscribe from this group and stop receiving emails from it, send an email to polymer-dev...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/polymer-dev/adc4db0a-459d-4abf-ab8f-6453dae62213%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Michael Giuffrida]

It would be really nice if anyone could provide a more detailed answer to this and similar questions: It's not clear what operations are supposed to be synchronous or asynchronous (and when async operations are at task level vs. microtask level), which affects tests but also some more complicated UI logic.

I started a thread last week which didn't get any responses. I would appreciate even a simple "we make no guarantees" response if that's the case, because it would help knowing how defensive our code should be.

Michael

To view this discussion on the web visit https://groups.google.com/d/msgid/polymer-dev/CAAUAVAhtWUNOMEw8Zm2Nq5JVA0Q4ihDEtgPJ4pDjRVqZHk1ENA%40mail.gmail.com.

[Steve Orvell]

For efficiency, Polymer queues dynamic changes that provoke Shady DOM distribution (changes that fill insertion points, e.g. <content>) and template rendering (e.g. `dom-if` and `dom-repeat`) and processes them at microtask timing. In addition, the Custom Elements polyfill processes dynamic dom changes (upgrading elements and calling attached/detached) at microtask timing.

In general, calling `Polymer.dom.flush` provides a guarantee that these tasks complete synchronously. 

To view this discussion on the web visit https://groups.google.com/d/msgid/polymer-dev/CACi5S_1DeF4B%2BkgPzwpW8rt%3DH1dw%2B2%3D1C_BbGqgdzuD5ZtPpQw%40mail.gmail.com.

[Michael Giuffrida]

> In general, calling `Polymer.dom.flush` provides a guarantee that these tasks complete synchronously. 

sorry to dig up an old thread, but I was poking around in Polymer 1.0.4 today and discovered that `Polymer.dom.flush` didn't guarantee `dom-if` template rendering after changing the `if` property. Instead, Polymer.Async was necessary to ensure the `if` took effect.

`Polymer.dom.flush` does seem to work in recent versions of Polymer, but the fact that it was insufficient in a previous 1.0 release worries me. We rely on this in tests to avoid excessive async functions. I just wanted to clarify whether we can consider `Polymer.dom.flush` sufficient to stamp the template nowadays, or if it's just an implementation detail that may revert to 1.0.4 behavior at any time.

[Arthur Evans]

Hi Michael,

I know a lot of folks are out on vacation so you may not get a quick authoritative answer. So here's a semi-quick, semi-authoritative answer.

Based on a discussion earlier this week with Steve and Kevin, I believe that this is now the expected behavior of Polymer.dom.flush(), and I'm planning on adding it to the documentation. 

Specifically, this feature was added in 1.0.8 in the following PR:

https://github.com/Polymer/polymer/pull/2090

The feature is described in the PR, rather than being a side-effect of an unrelated change.

Hope this semi-answer is semi-helpful,

Arthur

To view this discussion on the web visit https://groups.google.com/d/msgid/polymer-dev/CACi5S_2OG5_0DEBLdBUexZ%3DjY3UT-4x7hR8YQvr1SeQnkg5JBA%40mail.gmail.com.

[Michael Giuffrida]

Nice! Thanks for the commit sleuthing. Having it in the docs would be great, too.

Michael