This makes them less likely to collide, which was a problem on some
larger instances.

A following PR will generate entity_id from the hashes before
(de)serialization, allowing for easier merging later (since entity_id is
indexed in the database).

That case was erroring on Postgres (and MySQL?) because it tried to do
`AND collection_id IN ()`

* Show collection in data reference

* Fix type check

* Revert "Fix type check"

This reverts commit 3eef7cd21c8c6c1677fc427ac72a581cd8580bb4.

* Use Collections selector instead of question.card().collection

* fetch collection in loadMetadataForCard

* Replace getCollectionFromCard with Collections selector and rename question -> card

Just added that we don't include archived items in dumps

* Add collection to card tag autocomplete results

* Remove comment

* Tidy test

* Remove unused import

* Bump width

* static pie labels

* visual tests

* Quick fix to prevent non-temporal labels from datetime formatting

* show values pie BE

* Text fix 

* add specs

* renamt show_labels to show_values

* Test fix 



* update spec

Co-authored-by: Adam James <adam.vermeer2@gmail.com>
Co-authored-by: Tim Macdonald <tim@tsmacdonald.com>

* Fix Collection created_at migrations for H2

* Fix MySQL/MariaDB set created_at for empty collections

* Fix MySQL/MariaDB one last time

Add `saml-configured`, `jwt-configured` and `google-auth-configured` settings, and some general setting refactor (#26045)

* general sso configured/enabled setings refactor

* fix ts type error and update be tests

Co-authored-by: Jeff Bruemmer <jeff.bruemmer@gmail.com>

* update new model copy

* add educational button

* remove EducationalButtonProps

* Add created_at to Collection

* Simplified migrations [ci skip]

* Add SQL for H2 and MySQL; add tests

* Sort namespaces

* Test fixes 

* Test fixes

* Test fixes 

* Make sure MySQL/MariaDB Liquibase generator sets the current timestamp function for ADD COLUMN

* Column settings correct keys. Show Axes labels logic defaults 'true'

These are two small changes related to static viz column formatting.

The show axes labels logic change fixes a faulty assumption from an earlier PR where it was assumed that if
:labels_enabled key was NOT present, then it was 'false'. The default however, is 'true', so the check had to change
to reflect the frontend.

Column settings have to be formatted so that the js code running in the backend looks in the right places. For
example, `:number-format` must become `:number_format`. I move the function that does that change up so that it can be
rolled up into `settings-from-column` to reduce the number of call sites for `update-col-for-js`.

* Adjust number format fn to consider :decimal_digits of currencies

* Simpler date-separator fix

* Prevent non-temporal labels from being passed to datetime formatter

* Table Temporal format columns are improved

The column headers will now be properly looked up, the key format was incorrect before, causing the `:column_title`
lookup to return nil, which ignores when a user changes the column name.

Table columns with temporal formatting settings come in with `:unit` `:default`, so that case was added to the
datetime formatter. Additionally, a helper function was added to pull the correct column viz settings to pass to the datetime.

* Change some table rendering to look for '_' keys instead of '-'

And also allow the field-id to take the form `[:field-id 10 nil]` OR `[:field-id 10]`

* Added test utils for static-viz rendering. Docstrings for details.

A big set of functions to enable cleaner static-viz rendering tests.

Next push will be thorough table tests, and some other static viz cleanup

* Add some viz-scenarios to help with table testing

* Lint

* Wrong map here

* Currently unused bindings

* Writing tests with render-test utils now, and catching new bugs

* Use column's field ref to maybe get the column settings

* Adjustments so that column-settings are properly grabbed

* Number formatter no longer assumes decimal places

The frontend will only display decimal places when strictly necessary OR when the column formatting dictates the
number of decimal places to show.

The number formatter on the backend made different assumptions about decimal places, so precision may be lost OR
needlessly displayed.

The formatter now checks the actual value passed in and adjusts the format string according to the decimal places
present in the value, not just via column-settings, though the column settings will still take priority.

* Shame on me. inline def is not needed here :/

* Adjust number format tests to assert that now we match FE behavior

* Really trying to get formatting buttoned up.

* Matching frontend's default format behavior for numbers

* Some tests failed because they assumed two decimal places always

* Values less than one still round after 2 sig-figs

* Column formatting on tables follows expected default behavior

* Getting public-settings included in number and datetime formatters

* Test for custom formatting public settings

* Removed unused require.

* address some PR feedback

* Try to clarify the logic in the number formatter

* Add some clarifying comments

* Use the existing viz tools to help create field-refs properly

* Cleaned up column-name impl. to use normed field-refs. Much nicer!

* Remove some unnecessary tests, and add a negative scientific num

* Simplify table columns test

* Simplify and add more format settings to column-formatting test

* Unused require

* Another one

* Fix format of column refs in test

* Remove unused thread

Previously, some automatic behavior was causing problems in serdes.
(The worst example is generating an `entity_id` on insert while
deserializing an entity we don't own - if deserialized again it would be
duplicated.)

A whole cascade of design choices fell out of this problem:
- `insert!` would generate `entity_id`s, so use `simple-insert!`
- But `simple-insert!` doesn't convert eg. MBQL `:definition` maps back
  into JSON strings
- We got the maps as Clojure data and not JSON strings because `select`
  and `simple-select` both run `post-select` and parse the JSON.
- So we ended up with a raw query on the select side and `simple-insert!`
  on the storage side.

This change unwinds that whole stack, and instead uses a dynamic
var to suppress the few pieces of `pre-insert` and `pre-update` logic
that causes problems.

The end result is much cleaner, and much more consistent with the rest
of Metabase's backend logic.

* Ensure static visualizations with multiple cards are rendered

[Fixes #21559]

* [static viz] Render multiple scalars as a bar chart

* add static viz build watch command

* reload graalvm context each time in dev mode

* run static viz with yarn:dev

* address comments

* use watch webpack env var

Co-authored-by: Noah Moss <noahbmoss@gmail.com>

* Install `deploysentinel` Cypress analytics

* Start recording Cypress runs on `master` using `deploysentinel`

* Initialize Metabase from a file

* Code cleanup

* Support optional template tags; more dox

* Finish documentation for ns str

* Force the first User to be an admin

* Tests for DB creation

* Add test to ensure that you can set User password with an {{env ...}} template

* Remove data app permission code (#26016)

* Revert "Use "apps" collection namespace for app collections (#25963)"

This reverts commit a1894202.

* Revert "Implement EE/Pro app permissions (#25764)"

This reverts commit cf9b1bf5.

* Revert "Implement global app permissions for the "All Users" group (#25679)"

This reverts commit 63e950f2.

* Revert "Require data permissions for executing actions (#25784)"

This reverts commit 732bf6b1.

* Revert "Prevent modifying app permissions via the collection endpoint (#25684)"

This reverts commit 5d3f8279.

* Revert "Implement DB specific execution permissions (#25629)"

This reverts commit 141a13ce.

* Revert "Introduce /execution/ permission (#25529)"

This reverts commit 57cbb921.

* Keep the app permission revision table

* [Apps] Remove emitters (#25999)

* [Apps] Remove emitters

Emitters were created when we thought that we would be adding actions to
multiple places in metabase. Instead we've coalesced to actions being a
core feature of data-apps and are executed from within them. Attaching
actions to models, as a form of re-use rather than a form of execution,
means that actions can still be executed directly from a model, if we
want to allow that in the future.

* Fix lint errors

* Move perform-action! tests out of the api.

Since we want to keep the implicit bulk functionality but remove those
endpoints move the tests to the actions namespace.

* Fix linter

* Change driver test to use function instead of endpoint

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

* fix detection of backend files when generation i18n edn

* appease whitespace linter

* tweak comment

* fix cypress test

* appease linter

* Use eduction instead of threading

* Update bin/i18n/test/i18n/create_artifacts/backend_test.clj

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

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