UIP-1637 First pass at strong mode compliance #15

aaronlademann-wf · 2016-10-28T00:29:22Z

Ultimate problem:
dart-lang/test#2331 - OverReact should be strong-mode compliant!

How it was fixed:

Some changes were made to the react-dart wrapper.
Lots of explicit type casting.
Resolving misc. Dart analyzer warnings, errors and hints.
For _FluxComponentMixin... I made it implement BatchedRedraws, but I'm not 100% confident in that solution. Would be great to get @maxwellpeterson-wf, @evanweible-wf or @trentgrover-wf to weight in on the validity of that.
Turned on strong mode in .analysis_options.

Testing suggestions:

Verify that all tests pass
Verify that demo components behave as expected
Try it out on some of our more complex internal Workiva components to verify that regressions are not being introduced.

Potential areas of regression:
Pretty much everything 😁

FYA: @greglittlefield-wf @aaronlademann-wf @jacehensley-wf @clairesarsam-wf @joelleibow-wf @maxwellpeterson-wf, @evanweible-wf or @trentgrover-wf

aviary-wf · 2016-10-28T00:29:26Z

Raven

Number of Findings: 0

greglittlefield-wf

Awesome, glad we're getting the ball rolling on this! Noticed a few things.

Side note: I'm not super familiar with best practices around writing strong-mode-compliant code, but at first glance, some of the typing in these changes seems to be a bit redundant.

Perhaps there are some patterns/tricks we can employ to make some of this code more concise. Of course, that can be done at a later time.

greglittlefield-wf · 2016-10-28T00:38:41Z

test/test_util/react_util.dart

-/// Renders a React component or builder into a detached node and returns the associtated Dart component.
-react.Component renderAndGetComponent(dynamic component) => getDartComponent(render(component));
+/// Renders a React component or builder into a detached node and returns the associated Dart component.
+UiComponent renderAndGetComponent(dynamic component) => getDartComponent(render(component)) as UiComponent;


Why is this changed to UiComponent? It's possible to use this to render a non-over_react component, so we should leave it as react.Component.

@greglittlefield-wf I changed it because on line 335 of common_component_tests.dart, consumedProps is not available on Component.

I guess I can just do a cast to UiComponent in the test...

Done here 2804edf

greglittlefield-wf · 2016-10-28T00:40:22Z

web/src/demo_components/progress.dart

-  /// If left unspecified one will be auto-generated for you to ensure
-  /// that the [caption] element is properly linked for accessibility purposes.
-  @override
-  String get id;


Could the @virtual annotation be used on the super field to allow this? (See meta package 1.0.4 changelog)

@greglittlefield-wf same as what I found in the react-dart PR... @virtual doesn't seem to make the error go away.

I think that only works when you override a field with a field 😢

Actually that doesn't even seem to work, at least on 1.19.1

greglittlefield-wf · 2016-10-28T00:41:40Z

test/test_util/react_util.dart

@@ -287,7 +287,7 @@ bool _hasTestId(Map props, String key, String value) {

  bool first = false;

-  var results = react_test_utils.findAllInRenderedTree(root, allowInterop((descendant) {
+  var results = react_test_utils.findAllInRenderedTree(root, allowInterop((Function descendant) {


Pretty sure descendant is always either ReactComponent or Element, not Function...

+ Also removed some unnecessary parentheses

codecov-io · 2016-10-28T16:50:57Z

Current coverage is 97.28% (diff: 100%)

Merging #15 into master will increase coverage by 0.25%

@@             master        dart-lang/test#2332   diff @@
==========================================
  Files            27         27          
  Lines          1211       1209     -2   
  Methods           0          0          
  Messages          0          0          
  Branches          0          0          
==========================================
+ Hits           1175       1176     +1   
+ Misses           36         33     -3   
  Partials          0          0

Powered by Codecov. Last update 170a02e...1dbb92f

aaronlademann-wf · 2016-10-31T13:14:40Z

@jacehensley-wf @greglittlefield-wf @clairesarsam-wf this is ready for another pass.

jacehensley-wf · 2016-10-31T17:18:16Z

integrate/pubspec.lock

@@ -0,0 +1,436 @@
+# Generated by pub


I don't think this should be committed

evanweible-wf · 2016-10-31T17:40:11Z

@aaronlademann-wf over_react's SDK range includes the latest (>=1.17.1), but the travis CI setup is pinned at 1.17.1. If you upgrade to the latest version of Dart, the analyzer step fails.

+ Will greatly ease the strong-mode transition

# Conflicts: # integrate/pubspec.yaml

+ Now that we’re gonna roll with a minimum SDK version of 1.19.1

aaronlademann-wf · 2016-10-31T22:50:34Z

@evanweible-wf @greglittlefield-wf I've merged in the contents of dart-lang/test#2340, which removes the need for some nasty type casting.

However - when I try to use bleeding edge sdk (1.20.1)... I get the following errors, all centered around FluxUiComponent stuff. Could one of you take a look there and see if there is a relatively painless "future-proofing" solution for that? I'm a little out-of-my-depths looking through how w_flux Store / Actions are getting typed within the transformer.

ERROR: Could not infer type parameter E, dynamic must be of type Store. ([over_react] lib/src/component_declaration/flux_component.dart:124)
ERROR: Could not infer type parameter E, dynamic must be of type Store. ([over_react] test/over_react/component_declaration/flux_component_test/handler_precedence.dart:18)
ERROR: Could not infer type parameter K, dynamic must be of type Store. ([over_react] test/over_react/component_declaration/flux_component_test/handler_precedence.dart:21)
ERROR: Could not infer type parameter E, dynamic must be of type Store. ([over_react] test/over_react/component_declaration/flux_component_test/redraw_on.dart:17)
ERROR: Could not infer type parameter K, dynamic must be of type Store. ([over_react] test/over_react/component_declaration/flux_component_test/store_handlers.dart:17)

greglittlefield-wf · 2016-11-01T15:24:12Z

@evanweible-wf My junior year HS English teacher loved that word 😄.

+ The majority of the lints this uncovers must be cast at this time, but we should have the lint on to prevent any future unnecessary casts from being added to the library.

aaronlademann-wf · 2016-11-01T15:35:17Z

ok... @evanweible-wf @greglittlefield-wf I think we're at a good place here. Final review maybe?

greglittlefield-wf

A couple small things, mostly questions. Looks really good otherwise.

greglittlefield-wf · 2016-11-01T15:29:35Z

integrate/lib/test_react_element_typing.dart

-/// > Unsound implicit cast from ReactElement<dynamic> to ReactElement<Component>
-void _reactElementTypingTest() {
+///
+///     Unsound implicit cast from ReactElement<dynamic> to ReactElement<Component>


ReactElement no longer has a generic parameter as of a breaking release or two ago, so we can probably just remove this whole file.

greglittlefield-wf · 2016-11-01T15:41:55Z

test/over_react/util/handler_chain_util_test.dart

@@ -51,7 +51,9 @@ main() {
        }

        List<TestGenericType> generateBadTypeArgs() {
-          return new List.generate(arity, (_) => new Object());
+          // ignore: Type check failed


Hm, I think this should just be List generateBadTypeArgs() {

greglittlefield-wf · 2016-11-01T15:43:18Z

test/over_react/util/handler_chain_util_test.dart

@@ -158,7 +160,7 @@ main() {
            test('calls all functions in order', () {
              var calls = [];

-              var functions = new List.generate(5, (index) {
+              List<Function> functions = new List.generate(5, (index) {


#supernit for something like this, I think it might be preferred to do

var functions = new List<Function>.generate(5, (index) {

@evanweible-wf ?

@greglittlefield-wf @evanweible-wf when I do that, I get the following test failures:

00:11 +887 -1: [Dartium Content Shell] test/over_react_test.dart: HandlerChainUtil generic chaining: CallbackUtil0Arg chainFromList() returns a function of arity 0 that calls all functions in order type 'List<Function>' is not a subtype of type 'List<Callback0Arg>' of 'callbacks' where List is from dart:core Function is from dart:core List is from dart:core package:over_react/src/util/handler_chain_util.dart 212:41 CallbackUtil.chainFromList over_react/util/handler_chain_util_test.dart 167:42 main.<fn>.<fn>.sharedTests.<fn>.<fn>.<fn>

Ah, sorry, should it be this, then?

var functions = new List<Callback0Arg>.generate(5, (index) {

I'm fine leaving it as-is if it works, though

greglittlefield-wf · 2016-11-01T15:45:00Z

test/over_react/util/rem_util_test.dart

@@ -205,7 +205,7 @@ main() {
          'correctly dispatches an event in resopnse to the first change', () async {
        expect(querySelector('#rem_change_sensor'), isNull);

-        var calls = [];
+        List<double> calls = [];


#supernit Same here: should this be

var calls = <double>[]?

Asking about these out mostly to improve my own knowledge.

greglittlefield-wf · 2016-11-01T15:46:42Z

lib/src/component_declaration/component_base.dart

@@ -175,7 +166,10 @@ abstract class UiComponent<TProps extends UiProps> extends react.Component {
  @override
  TProps get props {
    var unwrappedProps = this.unwrappedProps;
-    TProps typedProps = _typedPropsCache[unwrappedProps];
+    /// Have to cast as [TProps] until we can parameterize [_typedPropsCache].


Hm, I wonder if we could reintroduce the type parameter on this Expando to avoid this cast... it seems the linked issue might only apply to mixins, so the omitted parameterization may just be carried over from when this code was part of a mixin.

These getters are used a lot, so it would be nice to avoid the perf hit of this cast if possible.

Just tried this locally and it works

@greglittlefield-wf @jacehensley-wf Expandos parameterized, casts removed in c8b41b7

evanweible-wf · 2016-11-01T15:46:12Z

.travis.yml

@@ -1,6 +1,6 @@
 language: dart
 dart:
-  - "1.17.1"
+  - "1.19.1"


There are two more spots where the analyzer is failing on 1.20.1, but they're easy fixes:

test/over_react/component_declaration/flux_component_test/handler_precedence.dart L21-25

@override getStoreHandlers() => {props.store.store1: increment}; - increment(_) { + increment(Store store) { numberOfHandlerCalls += 1; }

Same thing for test/over_react/component_declaration/flux_component_test/store_handlers.dart L16-21.

With those two changes, I think you'd be fine to bump this up to 1.20.1

Even if you're not using the variable is has to be typed?

Maybe it has something to do with overriding the method but omitting the return type, I'm not sure.

evanweible-wf · 2016-11-01T15:53:19Z

lib/src/transformer/declaration_parsing.dart

-    List topLevelVarsOnly(String annotationName, Iterable<CompilationUnitMember> declarations) {
-      var topLevelVarDeclarations = [];
+    List<CompilationUnitMember> topLevelVarsOnly(String annotationName, Iterable<CompilationUnitMember> declarations) {
+      var topLevelVarDeclarations = <CompilationUnitMember>[];


I think these should be <TopLevelVariableDeclaration>

evanweible-wf · 2016-11-01T15:56:21Z

lib/src/transformer/declaration_parsing.dart

-    List classesOnly(String annotationName, Iterable<CompilationUnitMember> declarations) {
-      var classDeclarations = [];
+    List<CompilationUnitMember> classesOnly(String annotationName, Iterable<CompilationUnitMember> declarations) {
+      var classDeclarations = <CompilationUnitMember>[];


..and these should be <ClassDeclaration> - that should fix the failing tests

evanweible-wf · 2016-11-01T16:02:51Z

@aaronlademann-wf @greglittlefield-wf I tried running tests on 1.20.1, but there were several failures that seem like they must be a bug. Same tests pass on 1.19.1:

      'http://localhost:58105/over_react/component/resize_sensor_test.dart': error: line 79 pos 51: expression is not a valid compile-time constant
              {ResizeSensorHandler onResize, int width: defaultContainerWidth, int height: defaultContainerHeight,
                                                        ^

      dart:async                                           Future.Future.microtask
      over_react/component/resize_sensor_test.dart         main.<fn>.expectResizeAfter
      over_react/component/resize_sensor_test.dart 239:32  main.<fn>.<fn>.<fn>.<async>
      ===== asynchronous gap ===========================
      dart:async                                           Future.Future.microtask
      over_react/component/resize_sensor_test.dart         main.<fn>.<fn>.<fn>

greglittlefield-wf · 2016-11-01T16:35:54Z

@evanweible-wf Looks like a bug indeed; I've got a reduced test case and am filing an SDK bug

Update: Filed this bug: dart-lang/sdk#27719

aaronlademann-wf · 2016-11-01T20:49:02Z

@greglittlefield-wf @jacehensley-wf @evanweible-wf both dependent PRs have merged, react 3.0.1 has been published.

This should be ready for a final pass.

+ Since 99% of our contributors will never run pub get within that directory, and will therefore see a TON of errors when they first spin the project up.

jacehensley-wf

Just a few really small things

jacehensley-wf · 2016-11-01T21:01:01Z

lib/src/transformer/impl_generation.dart

@@ -364,19 +364,20 @@ class ImplGenerator {

    typedMap.node.members
        .where((member) => member is FieldDeclaration)
-        .where((FieldDeclaration member) => !member.isStatic)
-        .forEach((FieldDeclaration field) {
+        .where((member) => !(member as FieldDeclaration).isStatic) // ignore: avoid_as


Instead of this and the previous line you could just do:

typedMap.node.members .where((memeber) => member is FieldDeclaration && !member.isStatic) .forEach(...

jacehensley-wf · 2016-11-01T21:02:21Z

lib/src/transformer/impl_generation.dart

-        .where((FieldDeclaration member) => !member.isStatic)
-        .forEach((FieldDeclaration field) {
+        .where((member) => !(member as FieldDeclaration).isStatic) // ignore: avoid_as
+        .forEach((field) {


#nit Just so there is not a bunch of changes could the param be prefixed with an underscore and the new var not?

.forEach((_field) { final field = _field as FieldDeclaration; })

jacehensley-wf · 2016-11-01T21:03:11Z

test/over_react/component_declaration/component_base_test.dart

@@ -294,7 +294,7 @@ main() {

      group('modifyProps()', () {
        test('passes the provided modifier itself', () {
-          modifier(UiProps props) {
+          modifier(Map<dynamic, dynamic> props) {


#nit Unnecessary parameters

aaronlademann-wf · 2016-11-01T21:33:55Z

@jacehensley-wf feedback addressed.

jacehensley-wf · 2016-11-01T22:17:31Z

test/test_util/custom_matchers.dart

@@ -55,7 +55,9 @@ class ClassNameMatcher extends Matcher {
  }

  @override
-  bool matches(String className, Map matchState) {
+  bool matches(className, Map matchState) {
+    className = className as String; // ignore: avoid_as


We wouldn't have to do this if this were to be accepted https://github.com/dart-lang/matcher/issues/37

Update from that issue. On Dart 1.20.1 you can use @checked to override param types.

greglittlefield-wf · 2016-11-02T00:55:58Z

#shipit

+1

evanweible-wf · 2016-11-02T12:17:26Z

+1

aaronlademann-wf · 2016-11-02T14:26:28Z

@leviwith-wf this is ready for QA / merge.

leviwith-wf · 2016-11-02T22:21:55Z

QA Resource Approval: +10

tests pass (observed noted failures in current dart sdk)
provided examples behave themselves.

Merging.

Clean up diagnostics code organization

First pass at strong mode compliance

27c1a59

aaronlademann-wf added 3 commits October 27, 2016 17:30

Get rid of two more implicit type casts

f2a5cc6

Use https to see if travis likes that better

f1c6657

Get rid of two implicit casts that don’t cause warnings on my machine

17f16fa

greglittlefield-wf reviewed Oct 28, 2016

View reviewed changes

aaronlademann-wf added 4 commits October 28, 2016 08:44

Cast in the test instead of changing the type of the method

2804edf

Change Function to ReactComponent

d9071f8

Remove redundant left hand side typing

af8d5f9

+ Also removed some unnecessary parentheses

Turn on strong mode for integrate directory

2914187

aaronlademann-wf force-pushed the strong-mode-ftw branch from af32ef5 to 2914187 Compare October 28, 2016 16:42

aaronlademann-wf added 2 commits October 28, 2016 09:44

Undo pubspec path commit

3b51626

Fix implicit cast

0e7c44a

Add test coverage for string UiComponent.ref

c2299be

aaronlademann-wf added Type: enhancement analyzer-strong-mode State: dependent PR labels Oct 31, 2016

jacehensley-wf mentioned this pull request Oct 31, 2016

Make OverReact strong-mode compliant #14

Closed

2 tasks

jacehensley-wf reviewed Oct 31, 2016

View reviewed changes

Use the typedef from react-dart

fc1b96e

aaronlademann-wf added 3 commits October 31, 2016 11:42

Don’t commit integrate/pubspec.lock

b76c072

Ignore all pubspec.lock files

fe69d03

Bump min sdk to 1.19.1

46639f2

+ Will greatly ease the strong-mode transition

aaronlademann-wf mentioned this pull request Oct 31, 2016

Bump minimum sdk version from 1.17.1 to 1.19.1 #23

Merged

aaronlademann-wf added 2 commits October 31, 2016 15:46

Merge 'aaronlademann-wf/sdk_1.19.1' into strong-mode-ftw

047a8a9

# Conflicts: # integrate/pubspec.yaml

Remove unnecessary casts

9d6a136

+ Now that we’re gonna roll with a minimum SDK version of 1.19.1

Enable avoid_as lint

1b4a80a

+ The majority of the lints this uncovers must be cast at this time, but we should have the lint on to prevent any future unnecessary casts from being added to the library.

greglittlefield-wf reviewed Nov 1, 2016

View reviewed changes

evanweible-wf reviewed Nov 1, 2016

View reviewed changes

aaronlademann-wf added 8 commits November 1, 2016 10:57

Add casting to fix broken VM tests

d205ea9

Fix the vm test errors a better way

5547f8d

Fix analysis error when sdk version is > 1.20.1

c4dd932

Update react dependency to ^3.0.1

f211b71

Remove unnecessary test file

ba742e8

Address CR feedback

dd869c5

Parameterize Expando

c8b41b7

Fix some formatting anomalies

98917e9

aaronlademann-wf removed the State: dependent PR label Nov 1, 2016

Exclude integrate/ directory from analyzer

9a91b1d

+ Since 99% of our contributors will never run pub get within that directory, and will therefore see a TON of errors when they first spin the project up.

jacehensley-wf reviewed Nov 1, 2016

View reviewed changes

Address CR feedback

1dbb92f

jacehensley-wf reviewed Nov 1, 2016

View reviewed changes

rm-astro-wf changed the title ~~First pass at strong mode compliance~~ UIP-1637 First pass at strong mode compliance Nov 2, 2016

aaronlademann-wf added the State: ready for QA label Nov 2, 2016

leviwith-wf merged commit 6e6308e into Workiva:master Nov 2, 2016

greglittlefield-wf added a commit that referenced this pull request Jun 19, 2020

Merge pull request #15 from Workiva/clean-up-diagnostics

58ea216

Clean up diagnostics code organization

UIP-1637 First pass at strong mode compliance #15

UIP-1637 First pass at strong mode compliance #15

Conversation

aaronlademann-wf commented Oct 28, 2016 • edited Loading

aviary-wf commented Oct 28, 2016

Raven

greglittlefield-wf left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

aaronlademann-wf Oct 28, 2016 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

codecov-io commented Oct 28, 2016 • edited Loading

Current coverage is 97.28% (diff: 100%)

aaronlademann-wf commented Oct 31, 2016

Choose a reason for hiding this comment

evanweible-wf commented Oct 31, 2016

aaronlademann-wf commented Oct 31, 2016 • edited Loading

greglittlefield-wf commented Nov 1, 2016

aaronlademann-wf commented Nov 1, 2016

greglittlefield-wf left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

evanweible-wf commented Nov 1, 2016 • edited Loading

greglittlefield-wf commented Nov 1, 2016 • edited Loading

aaronlademann-wf commented Nov 1, 2016

jacehensley-wf left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

aaronlademann-wf commented Nov 1, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

greglittlefield-wf commented Nov 2, 2016

evanweible-wf commented Nov 2, 2016

aaronlademann-wf commented Nov 2, 2016

leviwith-wf commented Nov 2, 2016

aaronlademann-wf commented Oct 28, 2016 •

edited

Loading

aaronlademann-wf Oct 28, 2016 •

edited

Loading

codecov-io commented Oct 28, 2016 •

edited

Loading

aaronlademann-wf commented Oct 31, 2016 •

edited

Loading

evanweible-wf commented Nov 1, 2016 •

edited

Loading

greglittlefield-wf commented Nov 1, 2016 •

edited

Loading