In certain databases there is a need to set the collation for ID fields
like dag_id or task_id to something different than the database default.
This is because in MySQL with utf8mb4 the index size becomes too big for
the MySQL limits. In past pull requests this was handled
[#7570](#7570),
[#17729](#17729), but the
root_dag_id field on the dag model was missed. Since this field is used
to join with the dag_id in various other models ([and
self-referentially](https://github.com/apache/airflow/blob/451c7cbc42a83a180c4362693508ed33dd1d1dab/airflow/models/dag.py#L2766)),
it also needs to have the same collation as other ID fields.

This can be seen by running `airflow db reset` before and after applying
this change while also specifying `sql_engine_collation_for_ids` in the
configuration.

Other related PRs
[#19408](#19408)

* Add sample dag and doc for S3ListOperator

* Fix doc

* Move tree filtering inside react and add some filters

* Move filters from context to utils

* Fix tests for useTreeData

* Fix last tests.

* Add tests for useFilters

* Refact to use existing SimpleStatus component

* Additional fix after rebase.

* Update following bbovenzi code review

* Update following code review

* Fix tests.

* Fix page flickering issues from react-query

* Fix side panel and small changes.

* Use default_dag_run_display_number in the filter options

* Handle timezone

* Fix flaky test

Co-authored-by: Brent Bovenzi <brent.bovenzi@gmail.com>

This is another attempt to improve caching performance for
multi-platform images as the previous ones were undermined by a
bug in buildx multiplatform cache-to implementattion that caused
the image cache to be overwritten between platforms,
when multiple images were build.

The bug is created for the buildx behaviour at
docker/buildx#1044 and until it is fixed
we have to prpare separate caches for each platform and push them
to separate tags.

That adds a bit overhead on the building step, but for now it is
the simplest way we can workaround the bug if we do not want to
manually manipulate manifests and images.

* Add sample dag and doc for S3ListPrefixesOperator

* Fix static checks

…23612)

Generation of Breeze help requires breeze to be installed. However
if you have locally installed breeze with different dependencies
and did not run self-upgrade, the results of generation of the
images might be different (for example when different rich
version is used). This change works in the way that:
* you do not have to have breeze installed at all to make it work
* it always upgrades to latest breeze when it is not installed
* but this only happens when you actually modified some breeze code

* Add Quicksight create ingestion Hook and Operator

Co-authored-by: eladkal <45845474+eladkal@users.noreply.github.com>

…KubernetesExecutor (#23617)

* UnicodeDecodeError: 'utf-8' codec can't decode byte 0xXX in position X: invalid start byte

  File "/opt/work/python395/lib/python3.9/site-packages/airflow/hooks/subprocess.py", line 89, in run_command
    line = raw_line.decode(output_encoding).rstrip()            # raw_line ==  b'\x00\x00\x00\x11\xa9\x01\n'
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xa9 in position 4: invalid start byte

* Update subprocess.py

* Update subprocess.py

* Fix:  Exception when parsing log #20966

* Fix:  Exception when parsing log #20966 

 Another alternative is: try-catch it. 

e.g.

```
            line = ''
            for raw_line in iter(self.sub_process.stdout.readline, b''):
                try:
                    line = raw_line.decode(output_encoding).rstrip()
                except UnicodeDecodeError as err:
                    print(err, output_encoding, raw_line)
                self.log.info("%s", line)
```

* Create test_subprocess.sh

* Update test_subprocess.py

* Added shell directive and license to test_subprocess.sh

* Distinguish between raw and decoded lines as suggested by @uranusjr

* simplify test

Co-authored-by: muhua <microhuang@live.com>

Previously you had to manually add versions when changelog was
modified. But why not to get a bit more fun and get the versions
bumped automatically based on your assesment when reviewing the
provideers rather than after looking at the generated changelog.

* Prevent KubernetesJobWatcher getting stuck on resource too old

If the watch fails because "resource too old" the
KubernetesJobWatcher should not retry with the same resource version
as that will end up in loop where there is no progress.

* Reset ResourceVersion().resource_version to 0

We have now different answers posisble when generating docs, and
for testing we assume we answered randomly during the generation
of documentation.

Co-authored-by: Jed Cunningham <66968678+jedcunningham@users.noreply.github.com>

These checks are only make sense for upgrades.  Generally they exist to resolve referential integrity issues etc before adding constraints.  In the downgrade context, we generally only remove constraints, so it's a non-issue.

* Added postgres 14 to support versions(including breeze)

Add support for `RedshiftDeleteClusterOperator`. This will help to clean resources using airflow operators when needed. In the current implementation, By default, I'm waiting until the cluster is completely removed to return immediately without waiting set `wait_for_completion` param to False

- Add operator class
- Add basic unit test
- Add an example task
- Add relevant documentation

#23649)

* Added kubernetes version (1.24) in README.md(for Main version(dev)), accidentally removed in merge cnflict.

* Update README.md

Co-authored-by: Jarek Potiuk <jarek@potiuk.com>

…23497) (#23618)

* Rename cluster_policy to task_policy

* rename task_policy as example_task_policy.

…mption (#23497) (#23618)" (#23656)

This reverts commit ee342b8.

Co-authored-by: eladkal <45845474+eladkal@users.noreply.github.com>

Co-authored-by: eladkal <45845474+eladkal@users.noreply.github.com>

We just added support for Postgres 14 and K8S 1.24 and since we
did not have any changes to support either in main we are bringing
the support to 2.3 line as well.

This documentation syncs all remaining places where it should be
updated.

the glob parameters (for example `apache-airflow-providers-*`) did
not work because only fixed list of parameters was allowed.

This PR converts the package-filter parameter to stop verifying the
value passed - so autocomplete continues to work but you should
still be able to use glob.

It also removes few places where the parameters were used with
`--` separator.

TIL that absolute() is an undocumented in Pathlib and that we
should use resolve() instead.

So this is it.

We are finally able to get rid of the annoying false-positive
warnings and we have finally a chance on having warning-free
installation during docker builds.

When names are too long, pre-commit output looks very ugly and takes up 2x lines. Here I reduce max length just a little bit further so that pre-commit output renders properly on a macbook pro 16" with terminal window splitting screen horizontally.

* communicate via customevents

* Handle open group logic in wrapper

* fix tests

* Make grid action buttons sticky

* Add default toggle fn

* fix splitting task id by '.'

* fix missing dagrun ids

The existing `DatabricksSubmitRunOperator` is extended with the support for the `git_source` parameter which allows users to run notebook tasks from files committed to git repositories.

If specified, any notebook task that is part of the payload will clone the repository and check out the commit, tag, or the tip of the specified branch. This is an alternative to dev repos ([docs](https://docs.databricks.com/repos/index.html)) where the checkout/update would have to be triggered manually.

Public documentation for the feature available here: https://docs.databricks.com/dev-tools/api/latest/jobs.html (NB: as noted in the docs, the feature is currently in public preview).

The rename from #23562 missed few shell_parms usage where it
also should be replaced.

The bundle extras we have are only used for development and they
should not be used to install airflow from PyPI. This update
to documentation clarifies it.

Closes: #23692

…23687)

Several commands of Breeze depends on docker, docker compose
being available as well as breeze image. They will work
fine if you "just" built the image but they might benefit
from the image being rebuilt (to make sure all latest
dependencies are installed in the image). The common checks
done in "shell" command for that are now extracted to common
utils and run as first thing in those commands that need it.

* Add UI tests for /utils and /components

* add test for Table

* Address PR feedback

* Fix window prompt var

* Fix TaskName test from rebase

* fix lint errors

* Handle invalid date from query parameters in views.

* Add tests.

* Use common parsing helper.

* Add type hint.

* Remove unwanted error check.

* Fix extra_links endpoint.

Move top margin to each breadcrumb component to make sure that there is no overlap when the header wraps with long names.

We don't really need the alembic revision id in the filename.  having version instead is much more useful.  having both of them takes up too much space.

* Add typing for airflow/configuration.py

The configuraiton.py did not have typing information and it made
it rather difficult to reason about it-especially that it went
a few changes in the past that made it rather complex to
understand.

This PR adds typing information all over the configuration file

Co-authored-by: Wojciech Januszek <januszek@google.com>

…23674)

fix #23411

As per https://www.mongodb.com/brand-resources

main branch is broken since #23630 needed rebase before merge
as #23730 added the missing example dag

…23286)

when StandardTaskRunner runs tasks with exec

Issue: #23540

The "wait for image" step lacked --tag-as-latest which made the
subsequent "fix-ownership" step run sometimes far longer than
needed - because it rebuilt the image for fix-ownership case.

Also the "fix-ownership" command has been changed to just pull
the image if one is missing locally rather than build. This
command might be run in an environment where the image is missing
or any other image was build (for example in jobs where an image
was build for different Python version) in this case the command
will simply use whatever Python version is available (it does
not matter), or in case no image is available, it will pull the image
as the last resort.

If you tried to expand via xcom into a non-templated field without
explicitly setting the upstream task dependency, the scheduler would
crash because the upstream task dependency wasn't being set
automatically. It was being set only for templated fields, but now we do
it for both.

The function is deprecated and raises warnings #10192
Replacing the usage with `get_connection()`

* Maintain grid selection on filter upstream

The grid view selection was being cleared when clicking "Filter Upstream". The selection should persist.

Also, added a left margin to the "Reset root" button

* fix linting

Same as #19508 but for Sqlite as described in https://docs.sqlalchemy.org/en/14/dialects/sqlite.html#connect-strings to be able to create a Sqlalchemy engine from the URI itself.

Without this, it currently fails with the following error due to how we create URI in Connections. An absolute path is denoted by starting with a slash, means you need four slashes:

```
url = sqlite://%2Ftmp%2Fsqlite.db

    def create_connect_args(self, url):
        if url.username or url.password or url.host or url.port:
>           raise exc.ArgumentError(
                "Invalid SQLite URL: %s\n"
                "Valid SQLite URL forms are:\n"
                " sqlite:///:memory: (or, sqlite://)\n"
                " sqlite:///relative/path/to/file.db\n"
                " sqlite:////absolute/path/to/file.db" % (url,)
            )
E           sqlalchemy.exc.ArgumentError: Invalid SQLite URL: sqlite://%2Ftmp%2Fsqlite.db
E           Valid SQLite URL forms are:
E            sqlite:///:memory: (or, sqlite://)
E            sqlite:///relative/path/to/file.db
E            sqlite:////absolute/path/to/file.db
```

Cache preparation on CI used default (Python 3.7) version of the
image. It had an influence on time of "full build needed" only and
for users who wanted to build breeze image for Python version
different than default Python 3.7.

It had no big influence on "main" builds" because in main we are
build images with "upgrade-to-newer-dependencies" which takes
longer anyway.

After #23775 I noticed that there is yet another small improvement
area in the CI buld speed. Currently build-ci-image builds and push
only "commit-tagged" images, but "fix-ownership" requires
the "latest" image to run.

This PR adds --tag-as-latest option also to build-image and
build-prod-image commands - similarly as for the pull-image and
pull-prod-image. This will retag the "commit" images as latest in the
build-ci-images step and allow to save 1m on pulling the latest image
before fix-ownership (bringing it back to 1s overhead)

Saying "**the** PostgreSQL database" confused me. I thought it was implying that a user could/should connect to the airflow metadata db

This is required to use other Datadog tenants like app.datadoghq.eu

…or (#23800)