Consider updating html5lib version dependency

[taisph]

Testing suggests html5lib pubspec dependency can be expanded to <0.12.0, which would be helpful for people trying to use angular with polymer, core_elements etc.

I've bumped the version dependency locally and run the scripts/run-test.sh script as well as basic usage tests in my own Angular-based project. No code was harmed during this experiment.

pubspec.yaml

name: angular
version: 0.13.0
authors:
- Misko Hevery <[email protected]>
- Pavel Jbanov <[email protected]>
- James deBoer <[email protected]>
- Chirayu Krishnappa <[email protected]>
- Matias Niemela <[email protected]>
- Paul Rohde <[email protected]>
- Victor Berchet <[email protected]>
description: Angular for dart.
homepage: https://angulardart.org
environment:
  sdk: '>=1.4.0'
dependencies:
  args: '>=0.10.0 <0.11.0'
  analyzer: '>=0.15.0 <0.19.0'
  barback: '>=0.13.0 <0.17.0'
  browser: '>=0.10.0 <0.11.0'
  code_transformers: '>=0.1.4+2 <0.2.0'
  collection: '>=0.9.1 <1.0.0'
  di: '>=2.0.1 <3.0.0'
  html5lib: '>=0.10.0 <0.12.0'
  intl: '>=0.8.7 <0.12.0'
  perf_api: '>=0.0.8 <0.1.0'
  route_hierarchical: '>=0.4.21 <0.5.0'
  web_components: '>=0.3.3 <0.4.0'
dev_dependencies:
  benchmark_harness: '>=1.0.0 <2.0.0'
  guinness: '>=0.1.9 <0.2.0'
  mock: '>=0.10.0 <0.12.0'
  protractor: '0.0.5'
  unittest: '>=0.10.1 <0.12.0'

pub deps

angular 0.13.0
|-- analyzer 0.18.0
|   |-- args...
|   |-- logging...
|   |-- path...
|   '-- typed_mock 0.0.4
|-- args 0.10.0+2
|   '-- collection...
|-- barback 0.14.0+3
|   |-- collection...
|   |-- path...
|   |-- source_maps...
|   '-- stack_trace 0.9.3+2
|       '-- path...
|-- benchmark_harness 1.0.4
|   '-- browser...
|-- browser 0.10.0+2
|-- code_transformers 0.1.6
|   |-- analyzer...
|   |-- barback...
|   |-- path...
|   '-- source_maps...
|-- collection 0.9.4
|-- di 2.0.1
|   |-- analyzer...
|   |-- barback...
|   |-- code_transformers...
|   '-- path 1.2.2
|-- guinness 0.1.13
|   |-- collection...
|   '-- unittest...
|-- html5lib 0.11.0+2
|   |-- csslib 0.11.0
|   |   |-- args...
|   |   |-- logging...
|   |   |-- path...
|   |   '-- source_span 1.0.0
|   |       '-- path...
|   |-- source_maps 0.9.3
|   |   '-- path...
|   '-- utf 0.9.0+1
|-- intl 0.8.10+4
|   |-- analyzer...
|   |-- path...
|   '-- meta 0.8.8
|-- mock 0.11.0+2
|   '-- matcher 0.11.0
|-- perf_api 0.0.8
|-- protractor 0.0.5
|   '-- js 0.2.2
|       '-- browser...
|-- route_hierarchical 0.4.21
|   |-- browser...
|   '-- logging 0.9.2
|-- unittest 0.11.0+3
|   |-- matcher...
|   '-- stack_trace...
'-- web_components 0.3.5+1

[vicb]

Do you have any info of potential backward incompatibilities in 0.11 ?

[taisph]

I haven't been able to find any information about breaking changes or similar. Apparently the Dart Team didn't consider updating the changelog with changes between 0.10.0 and 0.11.0 important. I also haven't located an updated repository with the 0.11.0+ source.

A diff between the two shows approximately half of the changes are removal of previously deprecated methods.

As such I can only refer to the results of the Angular test suite, the Dart Editor analysis and my own app.

[sethladd]

Maybe @jmesserly has some ideas on the diff between 0.10 and 0.11 ?

[sethladd]

Also, maybe @vsavkin can take a look at this for us? It's blocking Angular + Polymer apps.

[vicb]

We only do basic things with html5lib: parsing a doc, matching nodes based on attributes. The smoke test done by @taisph should be enough to prove it works.

Are there any other deps that would need to be upgraded to play well with polymer ?

[jmesserly]

Maybe @jmesserly has some ideas on the diff between 0.10 and 0.11 ?

I believe it was https://codereview.chromium.org//268623002
querySelector/querySelectorAll have fairly complete implementations now (previously, they only handled tag names e.g. querySelector('div'))

[jmesserly]

oh, and some of the DOM APIs were fixed to align better with dart:html and http://dom.spec.whatwg.org

[taisph]

Polymer dropped html5lib (or a dependant dep did) somewhere between 0.12.0+1 and +5 but needs code_transformers >=0.2.0 <0.3.0.

[jmesserly]

polymer.dart still depends on html5lib. https://github.com/dart-lang/bleeding_edge/blob/master/dart/pkg/polymer/pubspec.yaml#L15
example usage https://github.com/dart-lang/bleeding_edge/blob/master/dart/pkg/polymer/lib/src/build/import_inliner.dart#L16

[taisph]

@jmesserly Ok. It complained about it pre 0.12.0+1 but not on 0.12.0+5 and up. Maybe I missed something.

[taisph]

I don't know if it is helpful at all but I did some more smoke testing: I widened code_transformers, html5lib and web_components on current angular.dart master branch (commit dba6727) and ran pub upgrade with dart-sdk 1.6.0-dev.9.6; changes currently necessary for using latest polymer, core_elements and paper_elements with angular. I then ran scripts/run-test.sh:

4 tests, 11 assertions, 0 failures, 0 skipped
Chrome 37.0.2062 (Linux): Executed 1581 of 1581 (2 FAILED) (10.769 secs / 9.757 secs)
Chrome 36.0.1985 (Linux): Executed 1581 of 1581 (2 FAILED) (4 mins 10.92 secs / 4 mins 4.261 secs)
Firefox 31.0.0 (Ubuntu): Executed 1581 of 1581 (2 FAILED) (9 mins 2.955 secs / 8 mins 35.728 secs)
TOTAL: 6 FAILED, 4737 SUCCESS

Failing tests:

Chrome 37.0.2062 (Linux) with ElementProbe enabled dte.compiler components should select on element FAILED
        Test failed: Caught Expected: 'INNER()'
          Actual: 'INNER(SHADOW-CONTENT)'
           Which: is different.
        Expected: INNER()
          Actual: INNER(SHADOW-CON ...
                        ^
         Differ at offset 6

pubspec changes:

diff --git a/pubspec.yaml b/pubspec.yaml
index db2a3c5..2dc7945 100644
--- a/pubspec.yaml
+++ b/pubspec.yaml
@@ -17,10 +17,10 @@ dependencies:
   analyzer: '>=0.15.0 <0.19.0'
   barback: '>=0.13.0 <0.17.0'
   browser: '>=0.10.0 <0.11.0'
-  code_transformers: '>=0.1.4+2 <0.2.0'
+  code_transformers: '>=0.1.4+2 <0.3.0'
   collection: '>=0.9.1 <1.0.0'
   di: '>=2.0.1 <3.0.0'
-  html5lib: '>=0.10.0 <0.11.0'
+  html5lib: '>=0.10.0 <0.13.0'
   intl: '>=0.8.7 <0.12.0'
   perf_api: '>=0.0.8 <0.1.0'
   route_hierarchical: '>=0.4.22 <0.5.0'
@@ -30,4 +30,4 @@ dev_dependencies:
   mock: '>=0.10.0 <0.12.0'
   protractor: '0.0.5'
   unittest: '>=0.10.1 <0.12.0'
-  web_components: '>=0.3.3 <0.4.0'
+  web_components: '>=0.3.3 <0.6.0'

pub deps after upgrade:

angular 0.13.0
|-- analyzer 0.18.0
|   |-- args...
|   |-- logging...
|   |-- path...
|   '-- typed_mock 0.0.4
|-- args 0.10.0+2
|   '-- collection...
|-- barback 0.15.0
|   |-- collection...
|   |-- path...
|   |-- source_span...
|   |-- pool 1.0.1
|   |   '-- stack_trace...
|   '-- stack_trace 1.0.2
|       '-- path...
|-- benchmark_harness 1.0.4
|   '-- browser...
|-- browser 0.10.0+2
|-- code_transformers 0.2.1
|   |-- analyzer...
|   |-- barback...
|   |-- path...
|   |-- source_span...
|   '-- source_maps 0.10.0
|       |-- path...
|       '-- source_span...
|-- collection 0.9.4
|-- di 2.0.2
|   |-- analyzer...
|   |-- barback...
|   |-- code_transformers...
|   '-- path 1.3.0
|-- guinness 0.1.14
|   |-- collection...
|   '-- unittest...
|-- html5lib 0.12.0
|   |-- csslib 0.11.0+2
|   |   |-- args...
|   |   |-- logging...
|   |   |-- path...
|   |   '-- source_span...
|   |-- source_span 1.0.0
|   |   '-- path...
|   '-- utf 0.9.0+1
|-- intl 0.11.5
|   |-- analyzer...
|   '-- path...
|-- mock 0.11.0+2
|   '-- matcher 0.11.1
|-- perf_api 0.0.9
|-- protractor 0.0.5
|   '-- js 0.2.2
|       '-- browser...
|-- route_hierarchical 0.4.22
|   |-- browser...
|   '-- logging 0.9.2
|-- unittest 0.11.0+5
|   |-- matcher...
|   '-- stack_trace...
'-- web_components 0.5.0+1

(Updated: Added missing web_components change and Firefox didn't fail during the first couple of runs but did on the last run.)

[vicb]

@taisph I believe the failures might be related to guiness 0.1.14, see #1370

[vsavkin]

@taisph The failing test is incorrect. It was passing before because of a bug in Guinness, which was fixed in Guinness 0.1.14. The test was fixed in a branch that should be merged into master soon. For now, you can use Guinness 0.1.13.