[PHP-DEV] Add a static initialization method to `BcMath\Number`

Hi internals!

In this issue, a suggestion was made that it would be convenient if `BcMath\Number` could be initialized in a callable-style form (i.e. initialized via a static method):

Indeed, as pointed out in the issue, similar static methods exist in classes such as `DateTime`.

I am in favor of introducing such a static method. It would probably look something like this:

// $scale is optional
$num = BcMath\Number::createFrom($val, $scale);

It would effectively serve as an alias for the constructor.

Incidentally, since this does not introduce any BC break, I do not believe an RFC is necessary.

I would like to hear everyone’s thoughts on this.

Regards,

Saki

On Sat, May 23, 2026, at 7:50 AM, Saki Takamachi wrote:

Hi internals!

In this issue, a suggestion was made that it would be convenient if
`BcMath\Number` could be initialized in a callable-style form (i.e.
initialized via a static method):
BcMath\Number::createFrom · Issue #22128 · php/php-src · GitHub

Indeed, as pointed out in the issue, similar static methods exist in
classes such as `DateTime`.

I am in favor of introducing such a static method. It would probably
look something like this:

// $scale is optional
$num = BcMath\Number::createFrom($val, $scale);

It would effectively serve as an alias for the constructor.

Incidentally, since this does not introduce any BC break, I do not
believe an RFC is necessary.

I would like to hear everyone’s thoughts on this.

Regards,

Saki

I'm generally a fan of static factory methods like this, so +1 from me. It would also allow it to be partialed by the new PFA syntax, too.

--Larry Garfield

On Saturday, 23 May 2026 at 14:53, Saki Takamachi <saki@sakiot.com> wrote:

Hi internals!

In this issue, a suggestion was made that it would be convenient if `BcMath\Number` could be initialized in a callable-style form (i.e. initialized via a static method):
BcMath\Number::createFrom · Issue #22128 · php/php-src · GitHub

Indeed, as pointed out in the issue, similar static methods exist in classes such as `DateTime`.

I am in favor of introducing such a static method. It would probably look something like this:

// $scale is optional
$num = BcMath\Number::createFrom($val, $scale);

It would effectively serve as an alias for the constructor.

Incidentally, since this does not introduce any BC break, I do not believe an RFC is necessary.

I would like to hear everyone’s thoughts on this.

I'm against this.

We should instead find a solution to allow FCCs/PFAs to work on constructors and new rather than adding static factories for every possible class in existence.

I restarted work on a PR to clean up the semantics of constructors that, I believe, would make it feasible to allow FCCs/PFAs on new MyClass. [1]

Best regards,

Gina P. Banyard

[1] core: set zend_pass_function as default constructor by Girgias · Pull Request #19797 · php/php-src · GitHub

Hi

Am 2026-05-25 11:07, schrieb Gina P. Banyard:

I'm against this.

We should instead find a solution to allow FCCs/PFAs to work on constructors and new rather than adding static factories for every possible class in existence.

I agree with that and had also commented as such on the PR before seeing this thread.

Best regards
Tim Düsterhus

Hey Saki,

···

Important though: ::createFrom() is probably a bad name, and ::fromValueAndScale() or such would allow clearer/future expansion via other constructors.

Marco Pivetta

https://mastodon.social/@ocramius

https://ocramius.github.io/

On Mon, May 25, 2026, at 4:07 AM, Gina P. Banyard wrote:

We should instead find a solution to allow FCCs/PFAs to work on
constructors and new rather than adding static factories for every
possible class in existence.

I restarted work on a PR to clean up the semantics of constructors
that, I believe, would make it feasible to allow FCCs/PFAs on new
MyClass. [1]

The original PFA RFC included constructor support, but it was complicated. It was dropped from the new one that passed in order to simplify things, but I am open to re-adding it if a good approach is found. So it should certainly be possible, just a bit hard.

--Larry Garfield

On 29-05-2026 5:59 PM, Larry Garfield wrote:

On Mon, May 25, 2026, at 4:07 AM, Gina P. Banyard wrote:

We should instead find a solution to allow FCCs/PFAs to work on
constructors and new rather than adding static factories for every
possible class in existence.

I restarted work on a PR to clean up the semantics of constructors
that, I believe, would make it feasible to allow FCCs/PFAs on new
MyClass. [1]

The original PFA RFC included constructor support, but it was complicated. It was dropped from the new one that passed in order to simplify things, but I am open to re-adding it if a good approach is found. So it should certainly be possible, just a bit hard.

--Larry Garfield

An idea that has always played through my head for this, was to add a syntax-alternative to "new MyClass()" in the form: "MyClass::__construct()". (In other words, make the constructor statically callable).

But seeing it was dropped from the initial RFC, it's probably not be as trivial as it seems.

It could even be argued that the constructor should have been a static method in the first place.

I only found out today that you can call "->__construct()" on an existing object. I don't really see a use case for that, to be honest, and with readonly properties and property-promotion it's a whole snake-pit all by itself.

Regards,
Bernard