Remove final constructor for Type #6705

ruudk · 2025-01-13T13:12:02Z

Q	A
Type	improvement
Fixed issues	#6704

Summary

This allows people to define a constructor when creating custom Doctrine\DBAL\Types\Type implementations.

They can be registered by passing an instance to Type::addType.

When a class string is passed, and the Type cannot be instantiated, an helpful error message will be thrown.

greg0ire · 2025-01-13T14:07:17Z

Note: this is another take on #5528

morozov · 2025-01-13T17:36:18Z

I've never used custom types as besides schema management, they are primarily designed for the ORM. I'll defer to Alexander.

derrabus · 2025-01-13T20:52:55Z

If we do that, we turn Type::addType() into a potential footgun. Maybe it's time to deprecate the static methods on the Type class in favor of the TypeRegistry?

ruudk · 2025-01-14T10:53:16Z

If we do that, we turn Type::addType() into a potential footgun.

How will it be a potential footgun? The current desired behavior is guarded, so I don't see how.

Maybe it's time to deprecate the static methods on the Type class in favor of the TypeRegistry?

I agree, but that can be done independently from this PR, no?

derrabus · 2025-01-14T12:21:32Z

If we do that, we turn Type::addType() into a potential footgun.

How will it be a potential footgun? The current desired behavior is guarded, so I don't see how.

Yes, you've done all you could to give the developer proper feedback. Yet, we're about to create a class of types that cannot be registered with Type::addType() anymore, an issue which is only discoverable at runtime.

Maybe it's time to deprecate the static methods on the Type class in favor of the TypeRegistry?

I agree, but that can be done independently from this PR, no?

Sure, but still we should talk about it. We somehow ended up with an only partially implemented feature and I want us to work towards a proper solution this time instead of adding a small hack on top to make it somewhat work.

ruudk · 2025-01-16T08:29:59Z

@derrabus @greg0ire You're right. Let's deprecate Type::addType immediately. Updated the PR.

greg0ire

It seems the tests I wrote in my initial PR were not sufficient to prove it's possible to do what's in the documentation. Maybe they should be improved?

greg0ire · 2025-01-16T08:48:55Z

UPGRADE.md

@@ -8,6 +8,10 @@ awareness about deprecated code.

 # Upgrade to 4.3

+## Deprecated Type::addType
+
+Use Type::getTypeRegistry()->register() instead.


Suggested change

Use Type::getTypeRegistry()->register() instead.

Use `Type::getTypeRegistry()->register()` instead.

greg0ire · 2025-01-16T08:59:26Z

The docs should be updated, they mention addType.

ruudk · 2025-02-07T14:24:04Z

@greg0ire I made some changes. Could you approve the workflow so we can see the test results?

GromNaN

Added some suggestion to keep the API consistent.

Injecting the TypeRegistry everywhere might be a long-term project that involves a lot of changes in other Doctrine packages.

GromNaN · 2025-02-06T16:29:50Z

src/Types/Type.php

@@ -141,10 +138,16 @@ public static function lookupName(self $type): string
     * @param class-string<Type> $className The class name of the custom type.
     *
     * @throws Exception
+     *
+     * @deprecated Use Type::getTypeRegistry()->register() instead.
     */
    public static function addType(string $name, string $className): void


We could keep this method if the arg type were widened.

Suggested change

public static function addType(string $name, string $className): void

public static function addType(string $name, string|Type $type): void

Either we keep this API, or we deprecate all the static methods of this class (getType, hasType, lookupName, overrideType, getTypesMap), which would have a large impact on the Doctrine related packages.

If you don't want to change the signature of this method, you can add setType(string $name, Type $type, bool $override = false).

Note that the DoctrineBundle would use the optional "override" in ConnectionFactory.

I like it, I pushed the changes. Much simpler now!

GromNaN · 2025-02-06T16:31:46Z

tests/Functional/Schema/CustomIntrospectionTest.php

@@ -29,7 +29,7 @@ public static function setUpBeforeClass(): void
            self::markTestSkipped('Skip on Oracle');
        }

-        Type::addType(MoneyType::NAME, MoneyType::class);
+        Type::getTypeRegistry()->register(MoneyType::NAME, new MoneyType);


The API could be as simple as this:

Suggested change

Type::getTypeRegistry()->register(MoneyType::NAME, new MoneyType);

Type::addType(MoneyType::NAME, new MoneyType());

See doctrine#6704

ruudk · 2025-02-11T07:48:38Z

@GromNaN @greg0ire Updated the PR. I think this is a simpler approach. WDYT?

GromNaN

The approach seems correct at first.

Ideally, someone should take care of injecting the TypeRegistry everywhere the Type::getType() is used.

src/Types/Exception/TypeArgumentCountError.php

src/Types/Type.php

ruudk · 2025-02-11T08:28:39Z

@GromNaN thanks, applied your feedback!

ruudk mentioned this pull request Jan 13, 2025

Documentation about Custom Mapping Types with custom constructor is invalid #6704

Open

greg0ire requested review from morozov and derrabus January 13, 2025 14:06

ruudk force-pushed the ruudk/2025/1/remove-final-type-constructor branch from e6b5c83 to 355887c Compare January 16, 2025 08:29

greg0ire reviewed Jan 16, 2025

View reviewed changes

derrabus added Improvement Deprecation labels Jan 16, 2025

andrew-demb mentioned this pull request Feb 6, 2025

Autoconfigure DBAL types doctrine/DoctrineBundle#1867

Open

ruudk force-pushed the ruudk/2025/1/remove-final-type-constructor branch 2 times, most recently from 2c1e7df to b3b1dac Compare February 7, 2025 14:22

GromNaN reviewed Feb 7, 2025

View reviewed changes

Remove final constructor for Type

ca6bc55

See doctrine#6704

ruudk force-pushed the ruudk/2025/1/remove-final-type-constructor branch from b3b1dac to e466fd8 Compare February 11, 2025 07:45

Allow adding type instance

f9917a1

ruudk force-pushed the ruudk/2025/1/remove-final-type-constructor branch from e466fd8 to f9917a1 Compare February 11, 2025 07:47

ruudk requested review from GromNaN and greg0ire February 11, 2025 07:48

GromNaN suggested changes Feb 11, 2025

View reviewed changes

src/Types/Exception/TypeArgumentCountError.php Outdated Show resolved Hide resolved

src/Types/Type.php Outdated Show resolved Hide resolved

src/Types/Type.php Outdated Show resolved Hide resolved

Apply feedback

4d5934e

ruudk requested a review from GromNaN February 11, 2025 08:28

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Remove final constructor for Type #6705

Remove final constructor for Type #6705

ruudk commented Jan 13, 2025 •

edited

Loading

greg0ire commented Jan 13, 2025

morozov commented Jan 13, 2025

derrabus commented Jan 13, 2025

ruudk commented Jan 14, 2025

derrabus commented Jan 14, 2025

ruudk commented Jan 16, 2025 •

edited

Loading

greg0ire left a comment

greg0ire Jan 16, 2025

greg0ire commented Jan 16, 2025

ruudk commented Feb 7, 2025

GromNaN left a comment

GromNaN Feb 6, 2025

ruudk Feb 11, 2025

GromNaN Feb 6, 2025

ruudk commented Feb 11, 2025

GromNaN left a comment

ruudk commented Feb 11, 2025

	Use Type::getTypeRegistry()->register() instead.
	Use `Type::getTypeRegistry()->register()` instead.

	public static function addType(string $name, string $className): void
	public static function addType(string $name, string\|Type $type): void

	Type::getTypeRegistry()->register(MoneyType::NAME, new MoneyType);
	Type::addType(MoneyType::NAME, new MoneyType());

Remove final constructor for Type #6705

Are you sure you want to change the base?

Remove final constructor for Type #6705

Conversation

ruudk commented Jan 13, 2025 • edited Loading

Summary

greg0ire commented Jan 13, 2025

morozov commented Jan 13, 2025

derrabus commented Jan 13, 2025

ruudk commented Jan 14, 2025

derrabus commented Jan 14, 2025

ruudk commented Jan 16, 2025 • edited Loading

greg0ire left a comment

Choose a reason for hiding this comment

greg0ire Jan 16, 2025

Choose a reason for hiding this comment

greg0ire commented Jan 16, 2025

ruudk commented Feb 7, 2025

GromNaN left a comment

Choose a reason for hiding this comment

GromNaN Feb 6, 2025

Choose a reason for hiding this comment

ruudk Feb 11, 2025

Choose a reason for hiding this comment

GromNaN Feb 6, 2025

Choose a reason for hiding this comment

ruudk commented Feb 11, 2025

GromNaN left a comment

Choose a reason for hiding this comment

ruudk commented Feb 11, 2025

ruudk commented Jan 13, 2025 •

edited

Loading

ruudk commented Jan 16, 2025 •

edited

Loading