This reverts commit 9da0bd76.

* MLv2: Fix matching of field literals to field integer IDs (#31487)

* MLv2: Fix matching of field literals to field integer IDs

* Better test

* Update src/metabase/lib/equality.cljc

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

---------

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

* Fix #31482 not showing join display name for busted breakout from drill thru

---------

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

Use MLv2 for metadata calculation in QP (part 1); make column name calculation consistent with QP (#29583)

* Simplified impl

* Add #31266 test

* Yay

* Very nice

* Test fixes

* Fix Kondo warnings

This patch allows Metabase to cast timestamp in milliseconds to datetime

Improve `type-of` calculation for arithmetic expressions; fix schema for temporal arithmetic (#30921)

* Fix #29946

* Sort namespaces

* Implement available-aggregation-operators

* Support display-info and getting columns from aggregation-operators

* Support adding aggregation operators

* Add :cum-sum to the type hierarchy

Fixes #30703.

No `metabase.lib.js` wrapping yet.

The current state is a collection of hacks to preserve some legacy
behavior (eg. normalization sets a blank `widget-type` to `"category"`)
and it doesn't work properly right now.

An example of a currently-broken case:
- `SELECT * FROM Orders WHERE {{tag}}`
- Set it to a Field Filter and select People > LONGITUDE
  - Or some other field for which type we don't have a widget.
- It shows "None" as the widget type, and the widget isn't shown at the
  top.
- Save the question and reopen it.
- Now the widget is visible (and a bit broken) as a `"category"` type.

This is because on reading from the appdb it gets normalized.

This PR uses `"none"` properly for cases where there are no widget
options for the field type (eg. longitude), and updates all the usage
sites to properly handle `"none"` rather than missing/undefined.

Existing saved questions with a blank `widget-type` will continue to
"work" as they currently do.

* Fix clj-kondo analysis for `defclause`

* Fix Kondo config

* Fix typo

Added embeddings to preprocess sql completions so that we downselect to an appropriate prompt size. This also gives us substantial performance improvements for model and table matching inference.

Use the canonical `metabase.mbql.js/normalize` instead, with a bit of
customization for the query builder's needs. In the long run this should
be fully unified; for now at least the logic is inside MLv2 now.

All the test cases from the existing code are still exercised between
`selectors.unit.spec` and `metabase.mbql.js_test`.

This reverts commit a817ed8a.

* shared.ns/import-fn(s) emit optimized defn forms

* Code simplification

From https://github.com/metabase/metabase/issues/29545

* Improved implementation

* Amazing code cleanup

* Cleanup

* More tests

* Query  D E S C R I P T I O N S

* WIP

* MLv2 Query description

* Test fixes

* More robust error handling

* MLv2 more shuffling things around

* Remove unneeded impl

* lib.expressions -> lib.expression

* Address PR feedback

* MLv2 metadata overhaul, second try

* Stricter linting for MLv2

* Address PR feedback

* Revert fancy fn schema for now

* Remove unused import

* Test fixes 

* Oops fix bad :require form

* MLv2 mega metadata overhaul 2 of infinity

* Test fixes

* Cljc humanization

* Misc MLv2 improvements from #28921

* Fix indentation

* expression.non-integer-real

* Remove debugging stuff

* More cleanup

* Add method for `:advanced`

* non-integer-real

* Test fixes 

* Test fixes: ::ref/field shouldn't fail if checked against a non-sequential object

* MLv2 `define-mbql-clause` and `type-of` calculation

* Continued [ci skip]

* MLv2 overhaul :flex:

* Address PR feedback

* Merge type-of overhaul

* Address PR feedback

* MLv2 `define-mbql-clause` and `type-of` calculation

* Continued [ci skip]

* MLv2 overhaul :flex:

* Mark `hx/` and `h1x/` namespaces and their vars as deprecated

* Fix snowflake failures

* Remove unneeded honey-sql-version bindings

* Improves consistency of substring in MBQL

Standard SQL SUBSTRING() uses a 1-based start index, with various drivers
handling a 0 start index in different ways. To make the behavior of substring
consistent across drivers, we will always convert a 0 start index in MBQL
to a 1 in SQL. The ->honeysql implementation of substring has been generalized
and drivers can now override just the name of the substring function, or the
entire expression if necessary. SQL Server now calls the parent implementation.

Resolves #12445

* Fixes substring from 0 in Mongo

* Adds arg to substring parent call

* Fix docs and helper text for substring()

* Adds schema to require positive integer index for substring

* WIP, adds frontend validation for substring index arg

* Adds frontend test for substring arg validation

* Updates substring test to expect exception on index 0

* Update src/metabase/driver/sql/query_processor.clj

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

* Use s/if instead of s/conditional

* Fix substring call in mysql substring spec

* Reverts driver and sql.qp changes for substring index

* Reverts changes to dispatch-by-clause-name-or-class

* Fixes typeinferencer test substring index

* Adds test for canonicalizing substring

* Fixes substring canonicalization with extra args

* Updates string extracts test to account for substring canonicalization

* Rewrites resolver validation application

* Adds test case for frontend expression validation errors

* Fixes frontend compiler spec for new substring validation

---------

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

* Fix shared.util.i18n

* Simplify stuff

* Support Honey SQL 2 for SQL QP

* Sort namespace

* Make Postgres/Redshift be Honey SQL 2-only

* Test fix 

* Fix busted

* Redshift test fixes 

* Sort namespaces

* Fix Postgres convert_timezone

* Fix linter config.

* Fix one last test failure

* Redshift test fix

* Test fixes and some extra validation etc.

* Update common.clj

* Update common.clj

* Update common.clj

* Enable Kondo `:warn-on-reflection` linter

* Enable warnings for metabase.api.setting-test

* Add warn-on-reflection to metabase.util.jvm

Some JVM-specific functions are now in `metabase.util.jvm`, but they
are re-exported (in CLJ) by `metabase.util`.

This has the same interface as `clojure.tools.logging` but supports CLJS
as well. It was formerly known as `metabase.shared.util.log`.

Replaced, as needed, instances of str/lower|upper-case with the -en version. Also updated rules in our kondo config.

- Use currency-specific decimals count (eg. 0 for JPY, 8 for BTC)
- Use plural currency names in Java, matching JS
    - previously: singular proper name "7.23 US Dollar"
    - now: plural prose "7.23 US dollars"
- "Small numbers" hack to treat min/max fraction digits as significant
  digits now applies to percentages under 1%, as well as the original
  |n| < 1

This adds a cross-platform date/time formatting library in CLJC, with
(nearly) identical output in both JVM and JS environments.

The only known difference is week numbers.

CLJS (through Moment.js) has ordinal numbers, and renders the week
number as "34th". Neither the JVM nor Clojure has a handy library for
this (I'm sure it's out there somewhere, but it seems like a silly
dep to add) so it renders week numbers as simply "34".

Both platforms have date formatters that use pattern strings (eg.
`"YYYY-MM-dd"`) to turn date/time objects into strings. There's a strong
resemblance in how these work and what letters stand for what parts of
the date and time, but they are far from identical.

Rather than try to hackily convert one set of strings to another with
regexes or other manipulations, this PR defines a set of names for
fragments of dates (eg. `:year`, `:day-of-week-full`, `:hour-24-dd`)
and includes functions to transform a list of these keys into a
platform-specific format function.

This is portable and transparent, and can be written in Clojure or
JS code:

```clj
[:year "-" :month-dd "-" :day-of-month-dd]   ; 2022-04-08
[:month-full " " :day-of-month-d ", " :year] ; April 8, 2022
[:hour-12-dd ":" :minute-dd " " :am-pm]      ; 7:52 AM
```

```js
[":year", "-", ":month-dd", "-", ":day-of-month-dd"]   ; 2022-04-08
[":month-full", " ", ":day-of-month-d", ", ", ":year"] ; April 8, 2022
[":hour-12-dd", ":", ":minute-dd", " ", ":am-pm"]      ; 7:52 AM
```

Note that the original code allowed an unrecognized `:date-style` string
to be used directly as the formatting string. With the move to
formatting data structures that no longer works. Instead there is a
fixed map of format strings to the above data structures, that contains
all the currently used `:date-style` inputs.

If some caller needs a new format someday, we can either: (a) add the
style to the map; or (b) pass the data structure form directly to the
`:date-format` option, which if provided is used as the format.

It's tempting to go to all the call sites and replace these `:date-style`
strings with the new format structures. However, the strings are still
embedded in the `:visualization_settings` in user appDBs, so we still
need to recognize them.

Most of the existing functions in the TS library have the same API.

The set of allowed `:date-style` and `:time-style` values is turned into
a type using `keyof`, so that we get precise type-checking of these
values rather than simply `string`.

One function has been dropped from the API: `getDateFormatFromStyle`.
There was no practical way to implement it using the new format data
structures. The only caller was the date format column settings, and it
has been rewritten to use the formatted string for its sample date as
its key instead.

* restore aliases before annotating

* cleanup

* fix tests

* Don't add escaped->original if aliases have not changed

No need to walk and replace the aliases if they are identical. And in
that case, no need to keep a mapping of identical to identical. Not
super important but saves some time and complexity, and keeps other
tests passing since the presence of [:info :alias/escaped->original] in
the query caused them to trivially fail.

* oracle has a smaller limit and _expected_ mangled

previous testing behavior was "what happened" and not what should
happen. we fixed the bug but the "expect garbage" behavior was still
present

* Relax :alias/escaped->original schema

oracle tests use keywords for the alias

```clojure
{:alias/escaped->original
 {:test-data-venues--via-a763718f "test_data_venues__via__venue_id",
  :test-data-users--via--user-id "test_data_users__via__user_id"}}}
```

No idea why that is keyworded

* relax `:alias/escaped->original` schema

see previous commit

* add default values keys for dashboard.parmaeters

* fix tests and add tests

* oxford comma for parameter values with more than 2 items

* refactor logic to use a better translated string and fix tests

* standardize ns formatting in all src/ namespaces

* small fixes

* test formatting

* enterprise files

* fix whitespace linter error

* fix kondo error

* bin files

* reformat shared files

* fix errors

* fix another error

* mostly fix import to use parenthesis and have it at a new line

* fix wrong spacing

Co-authored-by: Ngoc Khuat <qn.khuat@gmail.com>

comparision -> comparison
[ci skip]

This adds a cross-platform number formatting library in CLJC, which has
(nearly) identical output in both JVM and JS environments.

The only known differences are:
- JVM Clojure has `BigDecimal` support, so isn't limited to a `double`'s
  range.
- Exact spelling/capitalization of the full names of currencies differs.
    - JVM: "7.34 US Dollar"
    - JS: "7.34 US dollars"

Some future work is required to fully land this. In particular, the
`currency_in_header` logic should be moved to the table formatters, not
implemented on the JS side of this change.