The motivation behind this change is to improve the productivity in the angular/angular repo
without sacrificing the original goal of having better understanding of changes within
the repo.
When the minBodyLength limit was originally introduced the goal was simple: force
committers to provide more contextual information for each change coming into the
repo. Too often we found ourselves in a situation where nobody understood what
motivated some of the changes and we needed more contextual info to know if the
change was correct, desirable, and still relevant (at a later point in time).
When the limit was introduced, we needed to pick a minimum body length - given no
data, and frustration with even big changes being committed with just a words in
the subject (along the lines of "fix(core): fixing a bug"), we overcompensated
and started off with a really high bar of minBodyLength set to 100 chars.
This turned out to be impractical and created a big friction point in making valid
changes in the angular/angular repo, and in fact caused some of the refactorings
and smaller changes to be either skipped or combined into other commits which
increased the burden for code reviewers.
The evidence in the friction points can be seen in the number of PRs that fail to pass
the current lint check on the first try, but more importantly also in the "creative"
writing that some of the committers are forced to resort to in order to satisfy the
current checks. Examples:
- https://github.com/angular/angular/commit/286fbf42c65a02e6da7420cd19a2c5baff3f2656
- https://github.com/angular/angular/commit/b2816a1536075397d876ba27ce2b7dcd785d4a39
Given that we primarily care to document the motivation behind each change
(the answer to the ultimate question: WHY?), I've collected several *common* &
*valid* commit messages that are minimalistic and capture the WHY sufficiently:
```
Refactoring for readability. => 28 chars
Improving variable naming. => 26 chars
Additional test coverage. => 25 chars
Cleaning up the code. => 21 chars
Simplified the code. => 20 chars
```
These commit message bodies in addition to the commit message subject should
sufficiently satisfy the need to capture the context and motivation behind each
change without creating an undue burden on committers.
Example minimalistic commit message:
------
refactor(core): cleanup the expression parser
Simplifying the code.
----
Given this research, I'm decreasing the minBodyLenth in angular/angular to 20 chars.
The commit message quality can be additionally improved by implementing a commit message
template via `.gitmessage` that will guide the committers in following our commit message
guidelines via instructions provided in the form of in-the-flow help rather than as an after
the fact lint check.
More info: https://thoughtbot.com/blog/better-commit-messages-with-a-gitmessage-template
I'm intentionally deferring such change for a separate PR as not to complicate or delay the
minBodyLength limit decrease.
PR Close#37949
The ngtsc testing packages for file_system and logging were missing from the bazel deps rules, which means that they were not included in the releases
PR Close#37977
Often changelogs are generated from the patch branch and then
cherry-picked into the `CHANGELOG.md` file in `master` for
better access and readability. This is problematic though as
`conventional-changelog` (the tool we use for generating the
changelog), will duplicate commits when a future changelog
is generated from `master` then (i.e. for a new minor release).
This happens because conventional-changelog always generates the
changelog from the latest tag in a given branch to `HEAD`. The
tag in the patch branch does not correspond to any SHA in `master`
so the intersection of commits is not automatically omitted.
We work around this naively (until we have a better tool provided
by dev-infra), by deduping commits that are already part of the
changelog. This has proven to work as expected in the components
repo.
PR Close#37956
This commit includes a couple of minor fixes to docs related to updating
to v10:
- Fix markdown link in "Updating to Angular version 10" guide.
- Correctly display numbered list in
"Solution-style `tsconfig.json` migration" guide.
PR Close#37897
The merge script uses `git cherry-pick` for both the API merge strategy
and the autosquash strategy. It uses cherry-pick to push commits to
different target branches (e.g. into the `10.0.x` branch).
Those commits never point to the commits that landed in the primary
Github branch though. For the autosquash strategy the pull request number
is always included, so there is a way to go back to the source. On the other
hand though, for commits cherry-picked in the API merge strategy, the
pull request number might not always be included (due to Github's
implementation of the rebase merge method).
e.g.
https://github.com/angular/components/commit/27f52711c0618a9ae4eab4c888c8ab3245638e77
For those cases we'd want to link the cherry-picked commits to the
original commits so that the corresponding PR is easier to track
down. This is not needed for the autosquash strategy (as outlined
before), but it would have been good for consistency. Unfortunately
though this would rather complicate the strategy as the autosquash
strategy cherry-picks directly from the PR head, so the SHAs that
are used in the primary branch are not known.
PR Close#37889
As part of angular.io's responsive layout, the menu shown in the top-bar
is collapsed into the sidenav on narrow screens at the point where the
search-bar (on the right side of the top-bar) would overlap with the
menu's nav-items.
Previously, the value used as break-point would work on marketing pages,
where the hamburger button is not shown on wide screens. However, on
docs pages (where the hamburger button is always shown, pushing the menu
further to the right), the search-bar would still overlap the menu
nav-items on some resolutions.
This commit fixes it by raising the screen width threshold at a value
that ensures there is no overlap even on pages where the hamburger
button is visible alongside the top-bar menu.
Fixes#37937
PR Close#37938
As part of angular.io's responsive layout, the following rules are
applied:
- On wide screens, a menu is shown in the top-bar and the sidenav is
shown side-by-side with the docs content.
- On narrow screens, the top-menu is moved from the top-bar to the
sidenav and the sidenav is closed by default and floats over the
content when manually opened.
Previously, the break-points at which the top-menu was shown in the
top-bar and the sidenav was shown side-by-side with the content were the
same (using a single variable).
This commit decouples the two break-points to make it possible to use
different values in the future.
PR Close#37938
Use a Sass variable for the screen width break-point at which the
top-bar hamburger button is hidden/shown. This allows more easily
updating the break-point.
PR Close#37938
When loading a translation file we ask each `TranslationParser`
whether it can parse the file. Occasionally, this check can find
errors in the file that would be useful to the developer. For example
if the file has invalid XML.
This commit deprecates the previous `canParse()` method and replaces it
with a new `analyze()` method. This returns an object that includes a
boolean `canParse` and then either a `hint` if it can parse the file,
or a `diagnostics` object filled with any messages that can be used to
diagnose problems with the format of the file.
Closes #37901
PR Close#37909
In `aio/`, we have a mechanism to apply patches in a `postinstall` hook.
See `aio/tools/cli-patches/README.md` for more info.
Previously, we had to update `aio/tools/cli-patches/patch.js` to list
each `.patch` file separately. While working on #37688, I found it
helpful for the script to automatically pick up `.patch` files.
This commit updates the script to automatically pick up and apply
`.patch` files from the `aio/tools/cli-patches/` directory. If one wants
to keep a `.patch` file but not apply it, they can change the file's
extension or move it to a sub-directory (without having to update the
script).
PR Close#37896
Currently when the `plural` or `select` keywords in an ICU contain trailing spaces (e.g. `{count, select , ...}`), these spaces are also included into the key names in ICU vars (e.g. "VAR_SELECT "). These trailing spaces are not desirable, since they will later be converted into `_` symbols while normalizing placeholder names, thus causing mismatches at runtime (i.e. placeholder will not be replaced with the correct value). This commit updates the code to trim these spaces while generating an object with placeholders, to make sure the runtime logic can replace these placeholders with the right values.
PR Close#37866
CanLoad guards are processed in asynchronous manner with the following rules:
* If all guards return `true`, operator returns `true`;
* `false` and `UrlTree` values wait for higher priority guards to resolve;
* Highest priority `false` or `UrlTree` value will be returned.
`prioritizedGuardValue` uses `combineLatest` which in order subscribes to each Observable immediately (not waiting when previous one completes that `concatAll` do). So it makes some advantages in order to run them concurrently. Respectively, a time to resolve all guards will be reduced.
PR Close#37523
This PR changes the logic for determining when to skip route processing from
using the URL of the last attempted navigation to the actual resulting URL after
that transition.
Because guards may prevent navigation and reset the browser URL, the raw
URL of the previous transition may not match the actual URL of the
browser at the end of the navigation process. For that reason, we need to use
`urlAfterRedirects` instead.
Other notes:
These checks in scheduleNavigation were added in https://github.com/angular/angular/commit/eb2ceff4ba19a09a5f400f81e371ec176878cd37
The test still passes and, more surprisingly, passes if the checks are removed
completely. There have likely been changes to the navigation handling that
handle the test in a different way. That said, it still appears to be important
to keep the checks there in some capacity because it does affect how many
navigation events occur. This addresses an issue that came up in #16710: https://github.com/angular/angular/issues/16710#issuecomment-634869739
This also partially addresses #13586 in fixing history for imperative
navigations that are cancelled by guards.
PR Close#37716
After recent correction to left nav TOC, the link to this new page was temporarily removed. This restores it to next, because the page is not yet available in stable.
PR Close#37855
The current code is missing a single quote at the end of the import.
(cherry picked from commit e13171ea2960dd0fa0666cb964b53799d2883e3a)
PR Close#37854
Incremental compilation allows for the output state of one compilation to be
reused as input to the next compilation. This involves retaining references
to instances from prior compilations, which must be done carefully to avoid
memory leaks.
This commit fixes such a leak with a complicated retention chain:
* `TrackedIncrementalBuildStrategy` unnecessarily hangs on to the previous
`IncrementalDriver` (state of the previous compilation) once the current
compilation completes.
In general this is unnecessary, but should be safe as long as the chain
only goes back one level - if the `IncrementalDriver` doesn't retain any
previous `TrackedIncrementalBuildStrategy` instances. However, this does
happen:
* `NgCompiler` indirectly causes retention of previous `NgCompiler`
instances (and thus previous `TrackedIncrementalBuildStrategy` instances)
through accidental capture of the `this` context in a closure created in
its constructor. This closure is wrapped in a `ts.ModuleResolutionCache`
used to create a `ModuleResolver` class, which is passed to the program's
`TraitCompiler` on construction.
* The `IncrementalDriver` retains a reference to the `TraitCompiler` of the
previous compilation, completing the reference chain.
The final retention chain thus looks like:
* `TrackedIncrementalBuildStrategy` of current program
* `.previous`: `IncrementalDriver` of previous program
* `.lastGood.traitCompiler`: `TraitCompiler`
* `.handlers[..].moduleResolver.moduleResolutionCache`: cache
* (via `getCanonicalFileName` closure): `NgCompiler`
* `.incrementalStrategy`: `TrackedIncrementalBuildStrategy` of previous
program.
The closure link is the "real" leak here. `NgCompiler` is creating a closure
for `getCanonicalFileName`, delegating to its
`this.adapter.getCanonicalFileName`, for the purposes of creating a
`ts.ModuleResolutionCache`. The fact that the closure references
`NgCompiler` thus eventually causes previous `NgCompiler` iterations to be
retained. This is also potentially problematic due to the shared nature of
`ts.ModuleResolutionCache`, which is potentially retained across multiple
compilations intentionally.
This commit fixes the first two links in the retention chain: the build
strategy is patched to not retain a `previous` pointer, and the `NgCompiler`
is patched to not create a closure in the first place, but instead pass a
bound function. This ensures that the `NgCompiler` does not retain previous
instances of itself in the first place, even if the build strategy does
end up retaining the previous incremental state unnecessarily.
The third link (`IncrementalDriver` unnecessarily retaining the whole
`TraitCompiler`) is not addressed in this commit as it's a more
architectural problem that will require some refactoring. However, the leak
potential of this retention is eliminated thanks to fixing the first two
issues.
PR Close#37835
tphobe9312
Igor Minar
Alex Rickabaugh
Olivier Combe
Paul Gschwendtner
George Kalpakas
Pete Bacon Darwin
Andrew Kushnir
Ruben Schuckit
Omar Hasan
Dmitrij Kuba
Andrew Scott
Judy Bogart
Celilsemi Sam Erkiner
Alan Agius
Sonu Kapoor