This reverts commit dc4d4de2.

keyed off of column names that include "update" which are date, time,
or datetime columns.

This reverts commit 3e1dab30.

[Fixes #7044] Render pulse bar charts with  negative values well

Also fixes misspelled column name in Avian Singles test dataset

* Bump DB prefixes so data gets reloaded to reflect col name change [ci drivers]

* Bump test DB version from v2 to v3

[ci drivers]

* More BigQuery test fixes 

 [ci drivers]

* Fix Snowflake [ci snowflake]

Co-authored-by: Cam Saul <github@camsaul.com>

* Add a Postgres test [ci postgres]

* Mega-optimized SELECT probe query [ci drivers]

* Mega sync improvements [ci drivers]

* Even more amazing sync perf improvements [ci drivers]

* Fix bad edit [ci drivers]

* Test/lint fixes 

* Test fix  [ci drivers]

* Some driver fixes  [ci drivers]

* Address PR feedback [ci drivers]

* Test fixes for everything except for Oracle [ci drivers]

* Fix Oracle [ci drivers]

* Fix Redshift [ci redshift]

* Allow the Administrators group to sync

Allows the Administrators group to sync from LDAP or SAML providers.
This was previously not allowed to avoid a user accidentally locking
themselves out of the admin group.

Fixes metabase/metabase-enterprise#445

* test that admins are properly removed

Test that when a sync happens, if the user was previously in the administrators group and no longer is, they are properly removed from that group.

* Fingerprints reduce over table-rows-sample instead of realizing all

tests are failing because we test fingerprinters a lot by supplying
fake results and then reducing over those results. Now we are reducing
over them in the query processor and need to get a better testing
strategy.

I think i can stub out the query-processor to take the `:rff` and just
transduce over some fake supplied results but need to verify

* Hijack query-processor to use rff on fake data

* Docstring and make private `table-rows-sample-query`

* Fix tests to use query function rather than hijack qp

also was calling empty not empty? :(

* drivers check [ci drivers]

* Docstring cleanup for rff and remove outdated comment

* Make rff required for table-rows-sample

now you must reduce over the reducible row set rather than realizing
the results in memory and then working with them

* empty commit for drivers [ci drivers]

* Fix SQL for custom columns referring to aggregations. Fixes #12762

* Add test & enable cypress

* Rebase

* Typo

* Add repro for #10511

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

When a database is updated via the admin interface, any sensitive data
(private keys, passwords, etc) should be handled on an "upsert" basis -
only modify those fields if they were changed from the "protected
password" value.

This affects all database drivers.

Resolves #13442

* Fix collections/graph stack overflow

* Test fixes 

 [ci postgres] [ci mysql]

* Show engine in Task trouble shooting logs (#13608)

* Show engine in Task trouble shooting logs

in an effort to get better feedback when people raise issues about
long running sync processes, include the database engine in the logs.

Also wrapped `th` in a `tr` inside the `thead` to remove react warning

* Const intead of let on the db_id_to_engine

* Swap db id for db name

* Move comment out of jsx

* fixup! Merge branch 'release-x.37.x' into fix-collections-graph-stack-overflow

Co-authored-by: dpsutton <dan@dpsutton.com>

* [Fixes #11879] [Fixes #8410] Don't wrap email attachment filenames

* Non-ASCII characters in filenames are encoded [as per IETF RFC
  2047](https://tools.ietf.org/html/rfc2047) (c.f. #8410)
* Sufficiently-long _encoded_ filenames are sent over multiple lines
    as per [RFC 2184 section
    3](https://tools.ietf.org/html/rfc2184#section-3), for example:

    ```
    Content-Disposition: attachment;
      filename*0="=?UTF-8?Q?T=C3=A9=C5=BFting_non-ASCII_=C4=8D=C4=A5=C3=A4ract";
      filename*1="ers.xlsx?="
    ```
* Many mail clients do the right thing, but Gmail does not

This commit therefore turns off wrapping, as per [Bill Shannon's
Stack Overflow answer re. an undocumented system
property](https://stackoverflow.com/a/53344141/220529)
(Bill Shannon co-wrote the Java email library, which is used by postal
(the Clojure wrapper we use).

* (Hopefully) fix race condition in dashboard drill Cypress spec

Suggested by @flamber

* Skip flaky Cypress tests (per Nemanja)

* Only set `mail.mime.splitlongparameters` once

* with-log-level: allow passing logger to change level of

* Clean up metabase.api.card-test

* Sane version of with-model-cleanup macro

* Test fixes 

* Test fixes

* Beautiful JUnit fix 

* Clean up metabase.api.alert-test

* Don't delete Alert :card if not passed in by PUT /api/alert/:id endpoint

* Cleanup metabase.api.dashboard-test

* More test cleanup 

* Test fixes 

* Unskip cypress

* Make null behaviour correction work with `joined-field`s

* Fix cypress test

* Make linter happy

* Cypress fix

* Fix Cypress test

- added regex pattern for weird singular/plural issue
- added a different positive assertion at the end

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

This changes *only* the CSV/JSON/XLSX downloads to use a HTTP 200
response code instead of a HTTP 202.

Resolves #11974

with assistance from https://github.com/plumatic/schema/issues/120

(the only particularly interesting bit beyond the obvious walk is to
remove any "open" matchers and use the s/Any)

* Fixing of TIME breakouts in MySQL

Reworks :hour and :minute breakouts to function on MySQL 5.5+ and
MariaDB 10.2+

Older versions of MySQL and newer versions of MariaDB do not accept a
date of 0000-00-00, which date_format was generating. In the cases where
we're attempting to truncate the date and normalize them all to group by
hours or minutes, this will cast the TIME to DATETIME, which will make
it possible to convert it back to the proper type after truncation.

Adds an additional two test cases for grouping on a TIME column (both
:hour and :minute)

Resolves #12846

[ci mysql]

* Switch to hx/cast instead of a custom cast defn

[ci mysql]

* Mark state fields by fingerprint fixes #2735 [ci drivers]

* fix tests for pg, ns sort, and docstring lint

* Expectations -> clojure.test

* minor alignment in test

* Allow fingerprint classifiers to override prev classifier special_type

if another classifier threw on a special type due to name or the
number of items in the collection, this should override it as it is
introspecting values.

States previously were marked ":type/State" based on the column
name. Now use a fingerprinter looking for state names. Owing to the
fact that there might be more things in the state column (US
territory, canadian states, Washington DC, etc) lowered the threshold
for recognizing a state column from 95% of values down to 70% of
values.

* Cleanup [ci drivers]

- remark about original field in metadata
- docstring on `can-edit-special-field?`
- cleanup and check for string in `state?`
- add tests that nils and other types don't break `state?`

* Include test file

When your app db is H2, this test will fail with a NumberFormatException
trying to parse the "not-an-integer" but the MySQL JDBC driver does not
behave the same. It is sufficient to test exception conditions return a
{} query description only under the H2 app-db, since it's a catch-all
handler.

* catch query description errors and improve client error message

* add test

* MBQL: handle nils in SQL filters more intuitively. #13332 [ci drivers]

* Refactor [ci drivers]

* Correctly handle value literals [ci drivers]

* Make `correct-null-behaviour` work with `Field` instances [ci drivers]

* Alphabet is hard

* Convert metabase.query-processor-test.explicit-joins-test to new style

* Backport test from #13317

* Enable test

* Use "table - field" style for join aliases

* Wow! Don't use table - field alias if they're basically the same thing!

* Lint fixes

* Obfuscate database details in the APIs

This will stop us from sending sensitive info (passwords, keys) via the
API for authenticated administrators. Note that non-administrator users
were never privvy to these fields.

* Use cleaner form suggested by Cam

* make sensitive field list a const

following template too closely with the `(parent-thunk)` call and
forgot it was wrapped in its own thunk so i was calling it before we
were using the result set.