Late FQCN resolution using ::class

This is an interesting discussion, and since I use phpdoc extensively, it's an
area that I have been thinking about, but WHY do you have to make it so
difficult to follow by simply bundling more text at the top. THIS needs to be
quoted properly and THEN perhaps other people can follow the discussion. Try
reading the rest of this message when you don't know what came before !!!!!

Benjamin Eberlei wrote:

This is indeed not possible, because strings are not class context
independent, you can pass them to anywhere. A string just doesn't know what
namespace it belongs to and this does not make sense without providing more
context in client libraries (such as docblocks).

Also the use statement information is NOT available in the engine, because
its resolved way before in the compiling step (as far as i understood it).

What Doctrine Annotations does to solve this is parsing this information
ourselves, using token_get_all(), then caching this information on a per
file basis.

Ok I get that, thankyou for the explanation.

static::class is not an option. I'm trying to resolve class names defined
in docblocks, since phpdoc2 allows for entering type hints (classes) as
namespace/use relative. And I can tell there is no current way of resolving
class names in strings, to FQCN's, unless I'm missing something? (There is
no way to get a list of the currently used namespaces as far as I can tell

• would such a function be possible to add to the SPL, without any major
  rewriting?)

So, I'm simply wondering if this would require any major rewriting to
support? Otherwise I could look into it and try to write a patch... Because
I think this would be really useful for framework developers, php unit
testing and php doc for example.

-Jens

On Mon, 25 Feb 2013 14:00:04 +0400, Jens Riisom Schultz ibmurai@me.com
wrote:

Hi everybody,

I have read up on this, and done some testing.

First up, my findings with PHP5.5 alpha5:

<?php
namespace spacy;

class classy {
public static function fqcn() {
/* This works but is not useful enough: */
//return self::class;

            $me = 'classy';

            /* This just doesn't work, but I wish it did: */
            //return $me::class;

            /* This simply does not work as expected: */
            return eval("return $me::class;");
            /* Output: "classy" - Expected output: "spacy\classy" */
    }

}
?>

I'm trying to late resolve a class name contained in a variable to the
FQCN. I understand that this is hard (maybe even impossible) with the
current implementation, because class name resolution happens compile time,
but eval("return $me::class;") simply returns something that is weird.

I guess what I'm trying to ask is whether it would be impossible to
support late FQCN resolution in any way? It would be very useful for
frameworks to be able to do this.

• Jens Riisom Schultz
  --

Hi Jens,

Here's what happened in your code:
When you invoked fqcn(), you created var $me = "classy";
Then you tried to invoke this code in eval: "return classy::class;"
But when php evals code, it's like including another file. So it
executes the code without any namespace (it's in global namespace), and php
didn't discover class with name classy (there's only spacy\classy) yet, so
it tries to resolve all your "use" statements (but you didn't write any)
and then just gives you "classy", it didn't throw any error just because it
tried to delay autoloading of this class as far as possible, if would do
eval("return new $me;") then you would get fatal error.

--
Lester Caine - G8HFL

Contact - http://lsces.co.uk/wiki/?page=contact
L.S.Caine Electronic Services - http://lsces.co.uk
EnquirySolve - http://enquirysolve.com/
Model Engineers Digital Workshop - http://medw.co.uk
Rainbow Digital Media - http://rainbowdigitalmedia.co.uk

This is indeed not possible, because strings are not class context independent, you can pass them to anywhere. A string just doesn't know what namespace it belongs to and this does not make sense without providing more context in client libraries (such as docblocks).

Also the use statement information is NOT available in the engine, because its resolved way before in the compiling step (as far as i understood it).

Thankyou for clarifying this. It was really all I wanted to know.
I guess this discussion is dead (though it birthed a different discussion, which is also impossible to follow since it is laced with flames)
I'm pulling out of this one. Thankyou all.

What Doctrine Annotations does to solve this is parsing this information ourselves, using token_get_all(), then caching this information on a per file basis.

Ok I get that, thankyou for the explanation.

static::class is not an option. I'm trying to resolve class names defined in docblocks, since phpdoc2 allows for entering type hints (classes) as namespace/use relative. And I can tell there is no current way of resolving class names in strings, to FQCN's, unless I'm missing something? (There is no way to get a list of the currently used namespaces as far as I can tell - would such a function be possible to add to the SPL, without any major rewriting?)

So, I'm simply wondering if this would require any major rewriting to support? Otherwise I could look into it and try to write a patch... Because I think this would be really useful for framework developers, php unit testing and php doc for example.

-Jens

Hi everybody,

I have read up on this, and done some testing.

First up, my findings with PHP5.5 alpha5:

<?php
namespace spacy;

class classy {
public static function fqcn() {
/* This works but is not useful enough: */
//return self::class;

           $me = 'classy';

           /* This just doesn't work, but I wish it did: */
           //return $me::class;

           /* This simply does not work as expected: */
           return eval("return $me::class;");
           /* Output: "classy" - Expected output: "spacy\classy" */
   }

}
?>

I'm trying to late resolve a class name contained in a variable to the FQCN. I understand that this is hard (maybe even impossible) with the current implementation, because class name resolution happens compile time, but eval("return $me::class;") simply returns something that is weird.

I guess what I'm trying to ask is whether it would be impossible to support late FQCN resolution in any way? It would be very useful for frameworks to be able to do this.

• Jens Riisom Schultz
  --

Hi Jens,

Here's what happened in your code:
When you invoked fqcn(), you created var $me = "classy";
Then you tried to invoke this code in eval: "return classy::class;"
But when php evals code, it's like including another file. So it executes the code without any namespace (it's in global namespace), and php didn't discover class with name classy (there's only spacy\classy) yet, so it tries to resolve all your "use" statements (but you didn't write any) and then just gives you "classy", it didn't throw any error just because it tried to delay autoloading of this class as far as possible, if would do eval("return new $me;") then you would get fatal error.