feat(Scope): Instrument Scope to use User tags in Dart Obervatory

[mvuksano]

No description provided.

[vicb]

could you reuse a static/final tag instead of creating a new one each time ?

[jbdeboer]

+1 to @vicb's comment

[mvuksano]

@jbdeboer related to #1119, wouldn't all user code be run using scope.watch? If this is OK to assume we could expose a public version of scope.watch() which would set the correct tags when using. Internally we could use different scope.watch method which would set the correct flags for internal usage. Something like:

ScopeInternal extends Scope {
  Scope _delegate;
  ScopeInternal(this._delegate);

  ...

  runAsync(fn()) {
    var angularCodeTag = new UserTag('angular');
    var prevTag = angularCodeTag.makeCurrent();
    _delegate.runAsync(fn)
    prevTag.makeCurrent();
  }
  ...
}

Not sure if this is the right approach.. I'm still thinking this through... Any suggestions are welcome :)

[mhevery]

moved locally to the functions

[mvuksano]

@mhevery & @jbdeboer This now includes quite a bit of information about scope. Any other ideas what might be useful to instrument around here?

Here's a screenshot to see what current output might look like.

[mvuksano]

Btw, it's an interesting insight that reaction function during digest is by far the most expensive in this example.

[mhevery]

UserTags need to be stored in global variable. We can't be making new ones ever time this code runs.

[mhevery]

Nice work. Just make sure that we don't constantly create new tags.

[mhevery]

LGTM, Otherwise

[mhevery]

use List instead of Map. Initialize List size from ScopeTTL

[mvuksano]

Done

[mhevery]

Also instrument View creation. This can be done in ViewFactory

https://github.com/angular/angular.dart/blob/master/lib/core_dom/tagging_view_factory.dart#L21

[jbdeboer]

@markovuksanovic @tbosch How do we know this change doesn't cause a performance regression in production mode (when tags are not needed?)

[mvuksano]

@jbdeboer I guess the only way to answer your question is to run some benchmarks. Do you know if there are some benchmarks that are already in place which could be used for this? I've seen some benchmarks added recently.

[mhevery]

The tests I have run do not show any impact on performance. Changing tags
is very cheap, about the cost of a field read/write, and they are not
placed in critical path of cade.

On Tue, Jun 24, 2014 at 6:59 AM, Marko Vuksanovic notifications@github.com
wrote:

@jbdeboer https://github.com/jbdeboer I guess the only way to answer
your question is to run some benchmarks. Do you know if there are some
benchmarks that are already in place which could be used for this? I've
seen some benchmarks added recently.

—
Reply to this email directly or view it on GitHub
#1138 (comment).