With providers split into bundles, the test injector is now able to
use providers for a given bundle. Suggested provider lists for tests are
available in `angular2/platform/testing/<platform>`.
Change the providers for a test suite using `setBaseTestProviders`. This
should be done once at the start of the test suite, before any test cases
run.
BREAKING CHANGE: Tests are now required to use `setBaseTestProviders`
to set up. Assuming your tests are run on a browser, setup would change
as follows.
Before:
```js
// Somewhere in test setup
import {BrowserDomAdapter} from 'angular2/src/platform/browser/browser_adapter';
BrowserDomAdapter.makeCurrent
```
After:
```js
// Somewhere in the test setup
import {setBaseTestProviders} from 'angular2/testing';
import {
TEST_BROWSER_PLATFORM_PROVIDERS,
TEST_BROWSER_APPLICATION_PROVIDERS
} from 'angular2/platform/testing/browser';
setBaseTestProviders(TEST_BROWSER_PLATFORM_PROVIDERS,
TEST_BROWSER_APPLICATION_PROVIDERS);
```
Closes#5351, Closes#5585Closes#5975
BREAKING CHANGE
Before
Previously Angular would run in dev prod mode by default, and you could enable the dev mode by calling enableDevMode.
After
Now, Angular runs in the dev mode by default, and you can enable the prod mode by calling enableProdMode.
I'm not sure what's the best way to update the shallow repo and I'm in a state where there is nothing to be
updated so I'll revisit the proper fix the next time I'm doing a release unless someone beats me to it.
Closes#5872
BREAKING CHANGE:
Previously `angular2.js`, `angular2.min.js` and `angular2.dev.js` bundles
would have zone.js and reflect-metadata pre-appended. New bundles don't
contain zone.js nor reflect-metadata - those external dependencies can
be easily loaded into a browser using `angular2-polyfills.js`
Closes#5881
Closes#5815Closes#5844
BREAKING CHANGE:
`angular2/angular2` was removed. Use the correct import from one of the barrels. E.g. `angular2/core`, `angular2/platform/browser`, `angular2/common`, …
Note: This only applies to JavaScript, Dart is not changed.
Assets defined for `templateUrl` and `styleUrls` can now be loaded
in relative to where the component file is placed so long as the
`moduleId` is set within the component annotation.
Closes#5634
Assets defined for `templateUrl` and `styleUrls` can now be loaded
in relative to where the component file is placed so long as the
`moduleId` is set within the component annotation.
Closes#5634Closes#5634
Fixes#5593
Part of #5665
BREAKING CHANGE:
Number and content of UMD bundles have changed:
- we only publish one bundle that contains: core, common, platform/browser, http, router, instrumentation and upgrade
- exported names have changed and now:
- core is exported as `ng.core`
- common is exported as `ng.common`
- platform/browser is exported as `ng.platform.browser`
- http is exported as `ng.http`
- router is exported as `ng.router`
- instrumentation is exported as `ng.instrumentation`
- upgrade is exported as `ng.upgrade`
Closes#5697
Have DomElementSchemaRegistry support namespaced elements,
so that it does not fail when directives are applied in SVG (or xlink).
Without this fix, directives or property bindings cannot be
used in SVG.
Related to #5547Closes#5653
Currently, importing from 'angular2/angular2', in addition to providing Angular tokens, brings in global-es6.d.ts. Since we are deprecating 'angular2/angular2', we need to do the same in 'angular2/core'.
move to new RxJS distribution.
BREAKING CHANGE:
RxJS imports now are via `rxjs` instead of `@reactivex/rxjs`
Individual operators can be imported `import 'rxjs/operators/map'`
- we now correctly print errors even on old Node versions
- we print error messages even when node_modules are missing or messed up
- error messages looks better
Closes#5230
This is a big change. @matsko also deserves much of the credit for the implementation.
Previously, `ComponentInstruction`s held all the state for async components.
Now, we introduce several subclasses for `Instruction` to describe each type of navigation.
BREAKING CHANGE:
Redirects now use the Link DSL syntax. Before:
```
@RouteConfig([
{ path: '/foo', redirectTo: '/bar' },
{ path: '/bar', component: BarCmp }
])
```
After:
```
@RouteConfig([
{ path: '/foo', redirectTo: ['Bar'] },
{ path: '/bar', component: BarCmp, name: 'Bar' }
])
```
BREAKING CHANGE:
This also introduces `useAsDefault` in the RouteConfig, which makes cases like lazy-loading
and encapsulating large routes with sub-routes easier.
Previously, you could use `redirectTo` like this to expand a URL like `/tab` to `/tab/posts`:
@RouteConfig([
{ path: '/tab', redirectTo: '/tab/users' }
{ path: '/tab', component: TabsCmp, name: 'Tab' }
])
AppCmp { ... }
Now the recommended way to handle this is case is to use `useAsDefault` like so:
```
@RouteConfig([
{ path: '/tab', component: TabsCmp, name: 'Tab' }
])
AppCmp { ... }
@RouteConfig([
{ path: '/posts', component: PostsCmp, useAsDefault: true, name: 'Posts' },
{ path: '/users', component: UsersCmp, name: 'Users' }
])
TabsCmp { ... }
```
In the above example, you can write just `['/Tab']` and the route `Users` is automatically selected as a child route.
Closes#4728Closes#4228Closes#4170Closes#4490Closes#4694Closes#5200Closes#5475
BREAKING CHANGE
Before
import * as p from 'angular2/profile';
import * as t from 'angular2/tools';
After
import * as p from 'angular2/instrumentation';
import * as t from 'angular2/platform/browser';
All common directives, forms, and pipes have been moved out of angular2/core,
but we kept reexporting them to make transition easier.
This commit removes the reexports.
BREAKING CHANGE
Before
import {NgIf} from 'angular2/core';
After
import {NgIf} from 'angular2/common';
Closes#5362
Currently, core depends on DomRenderer, which depends on the browser.
This means that if you depend on angular2/core, you will always
pull in the browser dom adapter and the browser render, regardless
if you need them or not.
This PR moves the browser dom adapter and the browser renderer out of core.
BREAKING CHANGE
If you import browser adapter or dom renderer directly (not via angular2/core),
you will have to change the import path.
Since editors and IDEs do typechecking and show errors in place,
often there is no benefit to running type checking in our test pipeline.
This PR allows you to disable type checking:
gulp test.unit.js --noTypeChecks
This commit also makes es6 generation optional.
fix(build): removes unnecessary circular dependencies
Closes#5299
This will send bundle sizes (before and after gzip) to Google Analytics so that we can
track bundle size over time for every bundle we produce.
Closes#5294
we can now filter build graph via --project flag to speed up build performance
usage:
gulp test.unit.js --project=angular2,angular2_material
Closes#5272
this is handy to conditionally create build graph but keep mergeTree() declarative - any input tree passed into
mergeTree that is null or undefined will simply be ignored
Currently, core depends on the browser, which means that other platforms (e.g., NativeScript or webworker) cannot use the bootstrapping logic core provides.
This PR extract makes bootstrapping logic in core completely platform-independent. The browser-specific code was moved to "angular2/platforms/browser".
BREAKING CHANGE
A few private helpers (e.g., platformCommon or applicationCommon) were removed or replaced with other helpers. Look at PLATFORM_COMMON_PROVIDERS, APPLICATION_COMMON_PROVIDERS, BROWSER_PROVIDERS, BROWSER_APP_PROVIDERS to see if they export the providers you need.
Closes#5219Closes#5280
Currently, core depends on the browser, which means that other platforms (e.g., NativeScript or webworker) cannot use the bootstrapping logic core provides.
This PR extract makes bootstrapping logic in core completely platform-independent. The browser-specific code was moved to "angular2/platforms/browser".
BREAKING CHANGE
A few private helpers (e.g., platformCommon or applicationCommon) were removed or replaced with other helpers. Look at PLATFORM_COMMON_PROVIDERS, APPLICATION_COMMON_PROVIDERS, BROWSER_PROVIDERS, BROWSER_APP_PROVIDERS to see if they export the providers you need.
Closes#5219
This is part of ongoing work to make core platform-independent.
BREAKING CHANGE
All private exports from 'angular2/src/core/facade/{lang,collection,exception_handler}' should be replaced with 'angular2/src/facade/{lang,collection,exception_handler}'.
BREAKING CHANGE
All private exports from 'angular2/src/core/{directives,pipes,forms}' should be replaced with 'angular2/src/common/{directives,pipes,formis}'
Closes#5153
Previously if the URL changed in `HashLocation` mode, the router would not pick up the change.
This adds a listener in `HashLocationStrategy` for `hashchange` events to fix the problem.
Closes#5013
Refactor EventEmitter and Async Facade to match ES7 Observable semantics, properly use RxJS typedefs, make EventEmitter inherit from RxJS Subject. Closes#4149.
BREAKING CHANGE:
- consumers of EventEmitter no longer need to call .toRx()
- EventEmitter is now generic and requires a type - e.g. `EventEmitter<string>`
- EventEmitter and Observable now use the `.subscribe(generatorOrNext, error, complete)` method instead of `.observer(generator)`
- ObservableWrapper uses `callNext/callError/callComplete` instead of `callNext/callThrow/callReturn`
Since the very first npm install is called while node_modules is empty, we need to ignore it, but we can track
the start timestamp and record the install even once the installation is completed.
We've had issues such as the one I documented: https://github.com/Microsoft/TypeScript/issues/5187
This tslint check prevents this happening again.
This change also updates to the newest tslint which gets typings from npm.
Closes#4970
- fixes wrapping for object literal keys called `template`.
- spacing in destructuring expressions.
- changes to keep trailing return types of functions closer to their
function declaration.
- better formatting of string literals.
Closes#4828
The directory contains code authored in a style that makes it transpilable to dart. As such, these are not idiomatic examples of Angular 2 usage.
The main purpose of this directory is to enable experimentation with Angular within the angular/angular repository.
Closes#4342Closes#4639
Adds test adapters for TypeScript and JavaScript only, exported
as part of the test_lib module. These work with the Jasmine test
framework, and allow use of the test injector within test blocks
via the `inject` function.
See #4572, #4177, #4035, #2783
This includes the TestComponentBuilder. It allows using the
test injector with Jasmine bindings, and waits for returned
promises before completing async test blocks.
This is pretty experimental, but the goal is to track the performance
of our build over time so that we can more easily track perf regressions.
Currently it's integrated only with gulp tasks, but I'd like to expand it
to tracking travis jobs, protractor/benchpress test runs, npm installs, etc.
No PII is being collected. And the data is collected via a Google Analytics
property owned by the Angular team account.
Closes#4672
This allows TypeScript to produce an API surface which matches the Dart semantics.
I found these with:
gulp build.js.dev && find dist/js/dev/es5/angular2/src -name "*.d.ts" -exec grep -H -n '^ *_' {} \;
Closes#4638
mock-fs is currently incompatible with node 4.x, but a fix is in progress
https://github.com/tschaub/mock-fs/issues/59
Since we are currently not actively developing the affected broccoli plugins,
the risk of disabling these tests is low, especially in the light of
improvements we get from node 4.x.
Closes#3605
BREAKING CHANGE:
- we don't mark an element as bound any more if it only contains text bindings
E.g. <div>{{hello}}</div>
This changes the indices when using `DebugElement.componentViewChildren` / `DebugElement.children`.
- `@Directive.compileChildren` was removed,
`ng-non-bindable` is now builtin and not a directive any more
- angular no more adds the `ng-binding` class to elements with bindings
- directives are now ordered as they are listed in the View.directives regarding change detection.
Previously they had an undefined order.
- the `Renderer` interface has new methods `createProtoView` and `registerComponentTemplate`. See `DomRenderer` for default implementations.
- reprojection with `ng-content` is now all or nothing per `ng-content` element
- angular2 transformer can't be used in tests that modify directive metadata.
Use `angular2/src/transform/inliner_for_test` transformer instead.
This change moves many APIs to the angular2/core export.
This change also automatically adds FORM_BINDINGS in
the application root injector.
BREAKING CHANGE:
Many dependencies that were previously exported from specific
APIs are now exported from angular2/core. Affected exports, which
should now be included from angular2/core include:
angular2/forms
angular2/di
angular2/directives
angular2/change_detection
angular2/bootstrap (except for dart users)
angular2/render
angular2/metadata
angular2/debug
angular2/pipes
Closes#3977
BREAKING CHANGE
Stop supporting http module in Dart. This is because Dart has a
well developed http package which should be used by Dart
customers instead.
BREAKING CHANGE:
Instead of configuring pipes via a Pipes object, now you can configure them by providing the pipes property to the View decorator.
@Pipe({
name: 'double'
})
class DoublePipe {
transform(value, args) { return value * 2; }
}
@View({
template: '{{ 10 | double}}'
pipes: [DoublePipe]
})
class CustomComponent {}
Closes#3572
This requires delicate handling of type definitions which collide, because
we use TypeScript-provided lib.d.ts for --target=es5 and lib.es6.d.ts for
--target=es6.
We need to include our polyfill typings only in the --target=es5 case,
and the usages have to be consistent with lib.es6.d.ts.
Also starting with this change we now typecheck additional modules,
so this fixes a bunch of wrong typings which were never checked before.
Fixes#3178
This change also makes us compliant with 1.6.0-dev compiler,
so we can do some experiments with apps that use 1.6 features
and compile against Angular.
We should probably add a travis build for 1.6 so we stay compatible
with both versions.
Static binary component tree of depth 10, i.e. 1024 components.
Current numbers for `pureScriptTime` are:
JavaScript:
Baseline: 27.10+-9%
Ng2: 26.84+-8%
Ng1: 55.30+-14%
Dart:
Baseline: 30.13+-4%
Ng2: 45.94+-3%
Ng1: 128.88+-10%
I.e. in JS we are same speed as baseline right now!
Some background: We had a recent change in the compiler that merges components into their parents already during compilation (#2529). This made Ng2 2x faster in this benchmark (before the Ng2 JS time was 49.59+-14%ms).
Closes#3196
Don't precompile Dart2JS for pull requests, instead serve the dart
sources with pub serve. We were already testing with Dartium so
all we lose is some test coverage of defects exposed only by the
Dart2JS transpiler.
This still runs the dart transformer.
Fixes#3030
This fixes several minor indentation issues (instanceof precendence,
type declaration specificity, template string length calculation).
This should also fix some flip-flop situations with template strings.
Travis creates an empty node_modules directory when the cache is empty which confuses
our current script into thinking that it's ok to require fs-extra. While this is rare,
it's better not to depend on anything in node_modules when purging it, so I reimplemented
recorsive delete that we use to purse node_modules.
This change causes the build system to write a log file into the tmp folder after each build.
This file contains command line arguments that tell karma about all the added/changed/removed files
from the last build. Karma can then use this list instead of doing internal globbing which can be
very expensive especially for hte dart builds that contain thousands of files.
Closes#2437
One of the non-angular broccoli plugins returns a weird object. We can't
assume that all trees meet the contract that we expect them to meet, so
we do a typecheck before storing the result of the rebuild.
Closes#2662
BREAKING CHANGES:
- host actions don't take an expression as value any more but only a method name,
and assumes to get an array via the EventEmitter with the method arguments.
- Renderer.setElementProperty does not take `style.`/... prefixes any more.
Use the new methods `Renderer.setElementAttribute`, ... instead
Part of #2476Closes#2637
This implementation only works in JavaScript, while the Observable transpilation
story gets worked out. Right now, the service just makes a simple request,
and returns an Observable of Response.
Additional functionality will be captured in separate issues.
Fixes#2028
Once we add support for addedPaths then this implementation will be significantly faster than the original.
In the meantime we benefit from having stable output directory which solves issues with certain files disappearing
during rebuild of a tree that contains flatten and mergeTree plugins.
Closes#2418
It is very noisy, especially when multiple trees are used. Since the
tree differ is fairly quick, it's not measuring the real costs of a
plugin anyhow.
Closes#2378
clang-format 1.0.17 substantially improves formatting for fat arrow functions
and array literal detection. It also fixes a number of minor formatting issues.
It was confusing because the test.unit.dart task does ignore the initial, and
the logger was hardcoded to always ignore the first task, leading to the appearance
that a run was happening twice for no reason.
Also, fixed the "fake ignoreInitial" handling to not rely on a fake event, which
is not necessary.
Closes#2101
Use the `ProtoViewDto` created by the render `Compiler` to create a
`ChangeDetectorDefinition`.
From there, generate a subclass of `AbstractChangeDetector` for each
`ChangeDetectorDefinition`.
Run some basic unit tests for the dynamic and JIT change detectors on
pre-generated change detectors.
Previously we assumed that all input and ouput paths for broccoli trees are immutable, that turned out to be
incorrect.
By adding a tree stabilizer plugin in front of each diffing plugin, we ensure that the input trees
are stable. The stabilization is done via symlinks which is super cheap on platforms that support
symlinks. On Windows we currently copy the whole input directory, which is far from ideal. We should
investagate if using move operation on Windows is ok in the future to improve performance.
Closes#2051