[11.x] Make floating-point types consistent

[hafezdivandari]

Fixes #3151, #7298, #9089, #9103, #13258, #18776, #21032, #21827, #42308 Related to laravel/ideas#1527 and #47080

This PR makes float and double column types consistent across all databases and also removes unsignedDecimal, unsignedDouble, and unsignedFloat column types.

Why?

Almost all databases have 4-bytes single-precision and 8-bytes double-precision storage types when declaring floating-point types (approximate numeric data values). Current syntax is inconsistent on Laravel and there is no way to declare precision value.

• As of MySQL 8.0.17, the nonstandard FLOAT(M,D) and DOUBLE(M,D) syntax is deprecated and you should expect support for it to be removed in a future version of MySQL. Source
• As of MySQL 8.0.17, the UNSIGNED attribute is deprecated for columns of type FLOAT, DOUBLE, and DECIMAL (and any synonyms); you should expect support for it to be removed in a future version of MySQL. Source

Changes

• Changes $blueprint->float($column, $total = 8, $places = 2, $unsigned = false) to $blueprint->float($column, $precision = 53)
• Changes $blueprint->double($column, $total = 8, $places = 2, $unsigned = false) to $blueprint->double($column)
• Changes $blueprint->decimal($column, $total = 8, $places = 2, $unsigned = false) to $blueprint->decimal($column, $total = 8, $places = 2)
• Removes $blueprint->unsignedFloat()
• Removes $blueprint->unsignedDouble()
• Removes $blueprint->unsignedDecimal()

Before this PR

Syntax	MySQL	PostgreSQL	SQL Server	SQLite
$table->float('foo')	double(8, 2)	double precision	float	float
$table->float('foo', 9)	double(9, 2)	double precision	float	float
$table->float('foo', 9, 3)	double(9, 3)	double precision	float	float
$table->float('foo', 9, null)	double	double precision	float	float
$table->float('foo', null, null)	double	double precision	float	float
$table->double('foo')	double(8, 2)	double precision	float	float
$table->double('foo', 9)	double(9, 2)	double precision	float	float
$table->double('foo', 9, 3)	double(9, 3)	double precision	float	float
$table->double('foo', 9, null)	double	double precision	float	float
$table->double('foo', null, null)	double	double precision	float	float

After this PR

Syntax	MySQL	PostgreSQL	SQL Server	SQLite
$table->float('foo')	float(53)	float(53)	float(53)	float
$table->float('foo', 24)	float(24)	float(24)	float(24)	float
$table->float('foo', null)	float	float	float	float
$table->double('foo')	double	double precision	double precision	double

Comparison of Floating-Point Types on Different DBs

Syntax	MySQL	PostgreSQL	SQL Server	SQLite
FLOAT(M, D)	Deprecated	N/A	N/A	Double (REAL)
DOUBLE(M, D)	Deprecated	N/A	N/A	Double (REAL)
FLOAT(1 to 24)	Single (FLOAT)	Single (real)	Single (float(24))	Double (REAL)
FLOAT(25 to 53)	Double (DOUBLE)	Double (double precision)	Double (float(53))	Double (REAL)
FLOAT	Single (FLOAT)	Double (double precision)	Double (float(53))	Double (REAL)
DOUBLE	Double (DOUBLE)	N/A	N/A	Double (REAL)
DOUBLE PRECISION	Double (DOUBLE)	Double (double precision)	Double (float(53))	Double (REAL)
REAL	Double (DOUBLE)	Single (real)	Single (float(24))	Double (REAL)

Upgrade guide

The double and float column types of database migrations have been rewritten to make these types consistent across all databases:

• The double column type now creates a DOUBLE equivalent column without total digits and places (digits after decimal point), which is the standard syntax in SQL. Therefore, you may remove the arguments for $total and $places:

  - $table->double('amount', $total = 8, $places = 2, $unsigned = false);
  // `DOUBLE(8, 2)` on MySQL, `DOUBLE PRECISION` on PostgreSQL, and `FLOAT` on SQLite and SQL Server
  + $table->double('amount');
  // `DOUBLE` on MySQL and SQLite, `DOUBLE PRECISION` on PostgreSQL and SQL Server

• The float column type now creates a FLOAT equivalent column without total digits and places (digits after decimal point), but with an optional $precision specification to determine storage size as a 4-byte single-precision column or an 8-byte double-precision column. Therefore, you may remove the arguments for $total and $places and specify the optional $precision to your desired value and according to your database's documentation. The default $precision is set to 53 that would result in 8-byte double-precision column on all database drivers:

  - $table->float('amount', $total = 8, $places = 2, $unsigned = false);
  // `DOUBLE(8, 2)` on MySQL, `DOUBLE PRECISION` on PostgreSQL, and `FLOAT` on SQLite and SQL Server
  + $table->float('amount', $precision = 53);
  // `FLOAT(53)` on MySQL, PostgreSQL, and SQL Server. `FLOAT` on SQLite

• The $unsigned parameter of the $table->decimal(), $table->double(), and $table->float() methods has been removed. Actually, this were only supported on MySQL. However, You may force deprecated UNSIGNED attribute using ->unsigned() modifier:

  - $table->decimal('foo', $total = 8, $places = 2, $unsigned = true);
  - $table->double('bar', $total = 8, $places = 2, $unsigned = true);
  - $table->float('baz', $total = 8, $places = 2, $unsigned = true);
  + $table->decimal('foo', $total = 8, $places = 2)->unsigned();
  + $table->double('bar')->unsigned();
  + $table->float('baz', $precision = 53)->unsigned();

• The unsignedDecimal, unsignedDouble, and unsignedFloat column types have been removed. Therefore, you may use equivalent column types and forcing deprecated UNSIGNED attribute using ->unsigned() modifier. Actually, this were only supported on MySQL:

  - $table->unsignedDecimal('foo', $total = 8, $places = 2);
  - $table->unsignedDouble('bar', $total = 8, $places = 2);
  - $table->unsignedFloat('baz', $total = 8, $places = 2)
  + $table->decimal('foo', $total = 8, $places = 2)->unsigned();
  + $table->double('bar')->unsigned();
  + $table->float('baz', $precision = 53)->unsigned();

[taylorotwell]

Please provide the documentation that would need to be included in a step by step upgrade for Laravel 11 if this PR was merged.

[taylorotwell]

Mark as ready for review when done please.

[bluedreamer]

Why was the unsigned removed from decimal()? Who made the decision to merge this. I use this and am having issues upgrading to 11. 1 database out of them all deprecating a feature doesn't mean it should be removed. Mariadb still has it and mysql hasn't got rid of it yet. What a stupid decision to allow the removal of the unsigned attribute on decimal()

[hafezdivandari]

@bluedreamer just use decimal()->unsigned() as explained on the upgrade guide.

[bluedreamer]

But why was a breaking public API change allowed for what amounts to a cosmetic reason? Why wasn't the param deprecated in documentation only? Wasted my time.

[devajmeireles]

Regardless of anything, the PRs that @hafezdivandari delivers are always of a high standard of description, incredible!

[Vimiso]

Laravel: ^11.9
MySQL: 8.4.2

Heads up, your documentation specifies before your PR $table->float('foo') will create a double, and after a float.

This is not the case.

$table->float('foo'), with no precision argument, will continue to create a double.

I have found having to set the precision explicitly to null, eg. $table->float('foo', precision: null) then indeed sets the data type to float.

[hafezdivandari]

@Vimiso I've explained that the default precisions is 53:

$table->float('foo') => FLOAT(53) => double-precision on all DBs
$table->float('foo', 24) => FLOAT(24) => single-precision on all DBs except SQLite
$table->float('foo', null) => FLOAT => double-precision on all DBs except MySQL and MariaDB

[Vimiso]

Ah got you, my bad!