Since IE11 does not support `Function#name`, we use a shim in tests that
parses the stringified function to extract the name. Previously, that
shim would cache the computed name on the function to speed up future
invocations. However, this resulted in incorrect values for functions
that "extended" other functions (such as the code generated by
TypeScript when downleveling ES2015 classes that extended other
classes).
To avoid issues such as #41416 (see also [internal discussion][1]), this
commit removes the caching of names. This is not expected to noticeably
affect performance, since (a) it is only used in tests, (b) it is only
used on browsers that do not natively support `Function#name` (i.e.
IE11) and (c) accessing function names is rare and inexpensive compared
to other operations that happen during testing.
[1]: https://angular-team.slack.com/archives/CB4UC1932/p1617285258058000
PR Close#41439
This change introduces a new hook on the `ResourceHost` interface named `transformResource`.
Resource transformation allows both external and inline resources to be transformed prior to
compilation by the AOT compiler. This provides support for tooling integrations to enable
features such as preprocessor support for inline styles.
Only style resources are currently supported. However, the infrastructure is in place to add
template support in the future.
PR Close#41307
With this change we add renovate to update dependencies in the following locations
- WORKSPACE
- integration/bazel/WORKSPACE
- package.json
- packages/**/package.json
- tools/ts-api-guardian/package.json
- aio/package.json
We also enable yarn workspaces so that dependencies in these packages are hoisting to the root and renovate doesn't created nested lock files.
Enabling auto updates is important, because quite often dependencies get out of date especially in the compiler-cli which depends on a number of external dependencies.
PR Close#41407
Introduces an **internal**, **experimental** `profiler` function, which
the runtime invokes around user code, including before and after:
- Running the template function of a component
- Executing a lifecycle hook
- Evaluating an output handler
The `profiler` function invokes a callback set with the global
`ng.ɵsetProfiler`. This API is **private** and **experimental** and
could be removed or changed at any time.
This implementation is cheap and available in production. It's cheap
because the `profiler` function is simple, which allows the JiT compiler
to inline it in the callsites. It also doesn't add up much to the
production bundle.
To listen for profiler events:
```ts
ng.ɵsetProfiler((event, ...args) => {
// monitor user code execution
});
```
PR Close#41255
Due to an issue with wombat proxy returning the login state of the generated tokens,
we will need to require a login for all `ng-dev release publish` runs to ensure npm
login has occured.
PR Close#41422
This commit removes a check for the name of the generated factory
function, which is unimportant to test the behaviour of the code.
The name of these functions is generated from the name of the class
being instantiated. In IE11, there is no `function.name` property available
and so there is a shim for it in `third_party/shims_for_IE.js`, which patches
the `Function.property.name` property.
For performance reasons this shim writes the result of the computation
to the prototype of the function. Unfortunately, this means that any class
that extends the patched class will have the same value for `name`.
PR Close#41416
The AIO search index is built in a WebWorker on the browser from a set
of page information that is downloaded as a JSON file (`search-data.json`).
We want to keep this file as small as possible while providing enough
data to generate a useful index to query against.
Previously, we only included one copy of each (non-ignored) term from each
doc but this prevents more subtle ranking of query results, since the number
of occurences of a term in a doc is lost.
This commit changes the generated file in the following ways:
- All non-ignored terms are now included in the order in which they appear
in the doc.
- The terms are indexed into a dictonary to avoid the text of the term being
repeated in every doc that contains the term.
- Each term is pre-"stemmed" using the same Porter Stemming algorith that the
Lunr search engine uses.
The web-worker has been updated to decode the new format of the file.
Now that all terms are included, it may enable some level of phrase based
matching in the future.
The size of the generated file is considerably larger than previously, but
on production HTTP servers the data is sent compressed, which reduces the
size dramatically.
PR Close#41368
This commit tries to improve the search results by processing
the query and attempting progressively less restrictive searches
until a non-zero set of pages is matched.
The new procesing includes:
* stripping off quote marks, which were causing searches to fail
* first attempting to match pages where ALL the query terms exist
* second attempting to match pages where ANY of the query terms exist
* third attempting to match pages where the title contains partial word matches
The first query attempt approximates, quite well, the idea of searching
for multi-word phrases. This is given the technical nature of the terms
and the fairly small size of the corpus.
PR Close#41368
