The core logic is still from the release script in bin/release.
The change here mostly wiring it up to GitHub Action so that it can be trigged
upon a new git tag (during the release process).

* Increase the request timeout for Percy model editor spec

* Update e2e/test/visual/models/editor.cy.spec.js

Co-authored-by: Tim Macdonald <tim@metabase.com>

---------

Co-authored-by: Tim Macdonald <tim@metabase.com>

* MLv2: validate references inside queries

* Validate references

* Skip join-alias validation if we have a Card source query

* Test fixes 



* Use `metabase.mbql.util.match` for match stuff

---------

Co-authored-by: Case Nelson <case@metabase.com>

* initial changes

* add tests

* remove namespaces

* Update test/metabase/api/user_test.clj

Co-authored-by: Noah Moss <32746338+noahmoss@users.noreply.github.com>

---------

Co-authored-by: Noah Moss <32746338+noahmoss@users.noreply.github.com>

Resolves #29687

* Move `BreakoutStep` to its own directory

* Rearrange props

* Add unit tests

Models based on native queries were not producing any SELECT clauses previously. This fixes that and puts both native and mbql query aliases under test. Additionally, I'm reducing the enum count threshold for pseudo-ddls to 10. This seems low, but models on stats are quite large so we need a lower number. In upcoming work we should make this value adaptive.

Fixes #30108.

* [MLv2] Clean query to remove any parts that fail schema validation

* Fix metric query generation

* Make test parallel

* Fix joins schema

When the FE is rendering implicitly joinable columns, it doesn't use the
target table name. Instead it uses the name of the FK field on the main
table. This is useful because there might be multiple FKs to the same
table with different meanings (eg. both `supplier_id` and `customer_id`
point to `PEOPLE`). In this example the `:fk_reference_name`s for the
column groups would be `Supplier` and `Customer`.

Fixes #30109

Many settings are inappropriate for transfer between instances, so exported settings are limited to a subset of settings
related to appearance and UI behavior of the instance.

Co-authored-by: Emmad Usmani <emmadusmani@berkeley.edu>

* Add repro test

* Fix limit was not properly set for preview queries

This reverts commit 9ebf7eec.

There are now four families of FE builds:

- `build` does a one-shot dev build.
- `build-watch` watches for changes and does dev rebuilds; no server.
- `build-hot` is `build-watch` plus a development web server.
- `build-release` is a one-shot release mode build.

Each family has `:cljs` and `:js` variants, to build just one part.

* MLv2: Add `display-name` style parameter with `:default` and `:long` options

* Add test for #29947

* Fix typos

* Update JS unit tests

* initial changes

* add saml api

* address changes

* git restore enterprise/backend/test/metabase_enterprise/sso/api/saml_test.clj

* update comment

* update to have /api

* remove dbg

* fix alias

* change deps

* update test

* add test

* update test

* split tests

* add paren

* [MLv2] Remove and replace fields clause

* Add docstring

* Delete old persistence schemas in redshift

Reminder that recently a change occurred that populates a cache_info
table in each persisted schema:

```sql
-- postgres/redshift
test-data=# select * from metabase_cache_424a9_379.cache_info ;
       key        |                value
------------------+--------------------------------------
 settings-version | 1
 created-at       | 2023-03-29T14:16:24.849866Z
 instance-uuid    | 407e4ba8-2bab-470f-aeb5-9fc63fd18c4e
 instance-name    | Metabase Test
(4 rows)
```

So for each cache schema, we can classify it as
- old style (a commit before this change. more and more infrequent)
- new style recent
- new style expired

And we can delete them accordingly in startup:
```
2023-04-11 20:09:03,402 INFO data.redshift :: Dropping expired cache schema: metabase_cache_0149c_359
2023-04-11 20:09:04,733 INFO data.redshift :: Dropping expired cache schema: metabase_cache_0149c_70
2023-04-11 20:09:05,557 INFO data.redshift :: Dropping expired cache schema: metabase_cache_0149c_71
2023-04-11 20:09:06,284 INFO data.redshift :: Dropping expired cache schema: metabase_cache_0149c_90
...
2023-04-11 20:20:33,271 INFO data.redshift :: Dropping expired cache schema: metabase_cache_fe4a7_90
2023-04-11 20:20:34,284 INFO data.redshift :: Dropping old cache schema without `cache_info` table: metabase_cache_8f4b8_358
2023-04-11 20:20:35,076 INFO data.redshift :: Dropping old cache schema without `cache_info` table: metabase_cache_8f4b8_69
...
```

It's possible this will at first cause a few flakes if we are unlucky
enough to drop a cache schema without `cache_info` for an instance that
is running tests at that point. But the `cache_info` table has been
backported so the chances of that become smaller each day.

I've let a week elapse from that change before committing this so
hopefully it is not an issue in practice.

Number of queries:
Makes a single query to get all schemas, then for each schema makes a
query to classify it. This can be unified into a single query with some
shenanigans like

```clojure
(sql/format {:select [:schema :created-at]
             :from   {:union-all
                      (for [schema schemas]
                        {:select [[[:inline schema] :schema]
                                  [{:select [:value]
                                    :from   [(keyword schema "cache_info")]
                                    :where  [:= :key [:inline "created-at"]]}
                                   :created-at]]})}}
            {:dialect :ansi})
```

But i found that this query is extremely slow and does not offer any
benefit over the simpler, repeated queries. And as we run this on each
commit now, the number of schemas will be far lower and it will be on
the order of 5-10 schemas (and therefore queries) and therefore not an
issue.

* Ngoc's suggestions

- docstring for `delete-old-schemas!`
- combine nested `doseq`
- use java-time over interop with java.time