Revert "show columns added after the dashcard has been created in the viz settings (#33886)" (#34116)

* Fix the flaky test 'downgrade-dashboard-tabs-test' by bumping the
migration version we upgrade to during the test. I'm not fully sure why
it was flaky in the first place. For some reasons the rollback step
doesn't get triggered on MySQL even though liqubiase does acknowledge
that the migration got ran?? eitherway bumping the version seems to take
the flaky a way.

Fixes #32373.

Support nested transactions for app DB
---------

Co-authored-by: metamben <103100869+metamben@users.noreply.github.com>
Co-authored-by: Ngoc Khuat <qn.khuat@gmail.com>

* Lazy load EntityForm

* Replace lazy loading with require

* add a comment

* Fix typo in comment

* Kick CI

* Run backport reminder on every push to an open PR

* Kick CI again with Nemanja's magic

---------

Co-authored-by: Nemanja <31325167+nemanjaglumac@users.noreply.github.com>

Set `last_acknowledged_version` automatically for new users

This prevents new users from being told about what's new in Metabase
when…everything is new in Metabase for them

* new implementation after rebase, also works for table details

* fix: makes #28304 pass again

* fix settings undefined crash

* use already defined variable

* get columns from settings param

* fix test data

* add unit test for metabase#28304

* Reproduce #33996

* [MLv2] Add query can-run function

* Update src/metabase/lib/native.cljc

Co-authored-by: metamben <103100869+metamben@users.noreply.github.com>

* Add test for non-native query

* Fix tests - snippet-id can be optional

---------

Co-authored-by: metamben <103100869+metamben@users.noreply.github.com>

* Bump h2 to fix flaky test

Symptom:
--------

error in linter job:
```
With premium token features = #{"audit-app"}
:metabase-enterprise.audit-app.pages.dashboards/most-popular-with-avg-speed

expected: (schema= {:status (s/eq :completed), s/Keyword s/Any} (qp/process-query query))
  actual: clojure.lang.ExceptionInfo: Error reducing result rows: Error running audit query: Invalid value "NULL" for parameter "result FETCH"; SQL statement:
WITH "MOST_POPULAR" AS (SELECT "D"."ID" AS "DASHBOARD_ID", "D"."NAME" AS "DASHBOARD_NAME", COUNT(*) AS "VIEWS" FROM "VIEW_LOG" AS "VL" LEFT JOIN "REPORT_DASHBOARD" AS "D" ON "VL"."MODEL_ID" = "D"."ID" WHERE "VL"."MODEL" = 'dashboard' GROUP BY "D"."ID" ORDER BY COUNT(*) DESC LIMIT ?), "CARD_RUNNING_TIME" AS (SELECT "QE"."CARD_ID", AVG("QE"."RUNNING_TIME") AS "AVG_RUNNING_TIME" FROM "QUERY_EXECUTION" AS "QE" WHERE "QE"."CARD_ID" IS NOT NULL GROUP BY "QE"."CARD_ID"), "DASH_AVG_RUNNING_TIME" AS (SELECT "D"."ID" AS "DASHBOARD_ID", AVG("RT"."AVG_RUNNING_TIME") AS "AVG_RUNNING_TIME" FROM "REPORT_DASHBOARDCARD" AS "DC" LEFT JOIN "CARD_RUNNING_TIME" AS "RT" ON "DC"."CARD_ID" = "RT"."CARD_ID" LEFT JOIN "REPORT_DASHBOARD" AS "D" ON "DC"."DASHBOARD_ID" = "D"."ID" WHERE "D"."ID" IN (SELECT "DASHBOARD_ID" FROM "MOST_POPULAR") GROUP BY "D"."ID") SELECT "MP"."DASHBOARD_ID", "MP"."DASHBOARD_NAME", "MP"."VIEWS", "RT"."AVG_RUNNING_TIME" FROM "MOST_POPULAR" AS "MP" LEFT JOIN "DASH_AVG_RUNNING_TIME" AS "RT" ON "MP"."DASHBOARD_ID" = "RT"."DASHBOARD_ID" ORDER BY "MP"."VIEWS" DESC LIMIT ? [90008-214]

<stack traces of us catching and rethrowing>
Caused by: org.h2.jdbc.JdbcSQLDataException: Invalid value "NULL" for parameter "result FETCH"; SQL statement:
WITH "MOST_POPULAR" AS (SELECT "D"."ID" AS "DASHBOARD_ID", "D"."NAME" AS "DASHBOARD_NAME", COUNT(*) AS "VIEWS" FROM "VIEW_LOG" AS "VL" LEFT JOIN "REPORT_DASHBOARD" AS "D" ON "VL"."MODEL_ID" = "D"."ID" WHERE "VL"."MODEL" = 'dashboard' GROUP BY "D"."ID" ORDER BY COUNT(*) DESC LIMIT ?), "CARD_RUNNING_TIME" AS (SELECT "QE"."CARD_ID", AVG("QE"."RUNNING_TIME") AS "AVG_RUNNING_TIME" FROM "QUERY_EXECUTION" AS "QE" WHERE "QE"."CARD_ID" IS NOT NULL GROUP BY "QE"."CARD_ID"), "DASH_AVG_RUNNING_TIME" AS (SELECT "D"."ID" AS "DASHBOARD_ID", AVG("RT"."AVG_RUNNING_TIME") AS "AVG_RUNNING_TIME" FROM "REPORT_DASHBOARDCARD" AS "DC" LEFT JOIN "CARD_RUNNING_TIME" AS "RT" ON "DC"."CARD_ID" = "RT"."CARD_ID" LEFT JOIN "REPORT_DASHBOARD" AS "D" ON "DC"."DASHBOARD_ID" = "D"."ID" WHERE "D"."ID" IN (SELECT "DASHBOARD_ID" FROM "MOST_POPULAR") GROUP BY "D"."ID") SELECT "MP"."DASHBOARD_ID", "MP"."DASHBOARD_NAME", "MP"."VIEWS", "RT"."AVG_RUNNING_TIME" FROM "MOST_POPULAR" AS "MP" LEFT JOIN "DASH_AVG_RUNNING_TIME" AS "RT" ON "MP"."DASHBOARD_ID" = "RT"."DASHBOARD_ID" ORDER BY "MP"."VIEWS" DESC LIMIT ? [90008-214]
 at org.h2.message.DbException.getJdbcSQLException (DbException.java:646)
    org.h2.message.DbException.getJdbcSQLException (DbException.java:477)
    org.h2.message.DbException.get (DbException.java:223)
    org.h2.message.DbException.getInvalidValueException (DbException.java:298)
    org.h2.command.query.Query.getOffsetFetch (Query.java:912)
    org.h2.command.query.Select.queryWithoutCache (Select.java:768)
    org.h2.command.query.Query.queryWithoutCacheLazyCheck (Query.java:197)
    org.h2.command.query.Query.query (Query.java:512)
    ...
```

BUT this only causes an issue for tests run under the cloverage
linter. It's not an issue under any other test scenarios.

Reproduction:
-------------

I can get the same stack traces with the following:

```clojure
dashboards=> (clojure.java.jdbc/query {:datasource
                                       (:data-source metabase.db.connection/*application-db*)}
                                      ["select 1 limit null"])
Execution error (JdbcSQLDataException) at org.h2.message.DbException/getJdbcSQLException (DbException.java:646).
Invalid value "NULL" for parameter "result FETCH"; SQL statement:
select 1 limit null [90008-214]

dashboards=> (clojure.java.jdbc/query {:datasource
                                       (:data-source metabase.db.connection/*application-db*)}
                                      ["select 1 limit ?" nil])
Execution error (JdbcSQLDataException) at org.h2.message.DbException/getJdbcSQLException (DbException.java:646).
Invalid value "NULL" for parameter "result FETCH"; SQL statement:
select 1 limit ? [90008-214]
```

But with a normal repl, running the test does not error:

```clojure
(deftest all-queries-test
  (mt/with-test-user :crowberto
    (with-temp-objects [objects]
      (premium-features-test/with-premium-features #{:audit-app}
                             ;; limit to just the version
        (doseq [query-type #_(all-query-methods) [:metabase-enterprise.audit-app.pages.dashboards/most-popular-with-avg-speed]]
          (testing query-type
            (do-tests-for-query-type query-type objects)))))))

pages-test=> (clojure.test/run-test all-queries-test)

Testing metabase-enterprise.audit-app.pages-test

Ran 1 tests containing 1 assertions.
0 failures, 0 errors.
{:test 1, :pass 1, :fail 0, :error 0, :type :summary}
```

And even more frustrating, just grabbing the sql and params and calling
the same code:

```clojure
common=> (let [driver (mdb/db-type)
               sql check/sql
               params check/params]
           (with-open [conn (.getConnection mdb.connection/*application-db*)
                       stmt (doto (sql-jdbc.execute/prepared-statement driver conn sql params)
                              foo
                              )
                       rs   (sql-jdbc.execute/execute-prepared-statement! driver stmt)]
             (into [] (clojure.java.jdbc/result-set-seq rs))))
[]
```

I can get the test to fail by running the cloverage runner with a socket
repl started and then running the test during that instrumented run:

```
clojure -J"$(socket-repl 6000)" -X:dev:ee:ee-dev:test:cloverage
```

But I'm not able to find anything interesting. It's a bug in CTEs with
parameters in h2, and it seems that only this query is sensitive.

If I inline the value it works under 2.1.214. And if I bump the version
to 2.2.224 it works. So I'm bumping the version.

* inline the offending value

* Revert "Bump h2 to fix flaky test"

This reverts commit 9cd6aed19cc4c143f79e611bb0c92218f18da365.

* [MLv2] Handle coalesce type after conversion

* Update test/metabase/lib/aggregation_test.cljc

Co-authored-by: metamben <103100869+metamben@users.noreply.github.com>

* Only update fields on converted queries

* Don't convert if base-type exists

* Guard against non-numeric field ids

* Fix test with hardcoded ids

* Allow metadata/field to return nil on not found

* Add ipaddress and mongobsonid types to equality-comparable-types

* Support is-empty not-empty for mongobsonid

* Add base-type to FE tests

* Fix e2e tests - column metadata is valid with field types filled in

---------

Co-authored-by: metamben <103100869+metamben@users.noreply.github.com>

* [MLv2] Add function for database engine

* Update src/metabase/lib/js.cljs

Co-authored-by: metamben <103100869+metamben@users.noreply.github.com>

* Use native query

---------

Co-authored-by: metamben <103100869+metamben@users.noreply.github.com>

* [MLv2] Fix template-tags snippet-id

Looking at this bug I found there were a couple problems that the
converters were not properly handling. First, the conversion to js was
copying snake case to kebab case instead of renaming. Second the conversion to
clojure was not picking up kebab case keys but using the snake case keys.

So I dropped the TemplateTag encoder and decoder and went forward with
js->clj and clj->js since the only tricky part was keeping the tag names
as strings and keeping the :type as keyword.

* Address review comments

* Make archived item rely on the `model` for its type

The `type` prop doesn't exist in the list of items.
`models` on the other hand exists and it shows the underlying item model,
which almost 1:1 matches the desired type.

Almost, because we can questions "card" in the code.
That will be accounted for in the next commit.

* Handle `card` and `dataset` models

Show "question` and "model" in the UI instead.

* Move `getTranslatedEntityName` to the common utils

* Use translated entities for the archived item tooltip

* Switch back to the implicit return

* Lowercase the translated model name

* Add missing string translation

* Add unit test for `getTranslatedEntityName`

Co-authored-by: Nemanja <31325167+nemanjaglumac@users.noreply.github.com>

'vice a versa' -> 'vice versa'

* fix multi-color selection in `ColorRangeSelector`

* unit test

* Fix back to x-ray dashboard flake

No need to load thousands of row and to rely on overly complicated automagic calculations.
The only thing this test wants to know whether or not we display "back to dashboard" button.

By greatly simplifying the underlying model, automagic dashboard becomes MUCH simpler.
The test will also be significantly faster as a result of this change.

* Make test title more accurate

* Remove the `@slow` test tag

* converts lib/utils to ts

* removed "MetabaseUtils" namespace in favor of exporting single functions

* fix tests

* fix bug

* removed unused isBase64 function

* Add custom warning message when leaving editing dashboard with unsaved changes

* Add a test for warning about unsaved changes (it does not work yet)

* Add a unit test for warning about unsaved changes

* Revert "Add a test for warning about unsaved changes (it does not work yet)"

This reverts commit 7210873994563f6fa0e1e8fd44596120cbc2d337.

* Add a test case for no changes

* Reduce number of checkNotNull calls

* Reduce number of checkNotNull calls

* Navigate to `/archive` directly

* Fix the flake by asserting on UI changes

* Clean test up

* Delete the superfluous assertion