Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:69970
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: pass (pb1.pair.com: domain gmail.com designates 209.85.128.47 as permitted sender)
MIME-Version: 1.0
Sender: jakub.php@gmail.com
Date: Wed, 30 Oct 2013 21:37:02 +0000
Message-ID: <CAEKnhAH3rF22zJ=9=XGR77Rqn0sGjW3vuX-5KiRhU5aX7g-Zsg@mail.gmail.com>
To: "nikic@php.net" <nikic@php.net>, PHP internals list <internals@lists.php.net>
Content-Type: multipart/alternative; boundary=047d7b33d17448b8da04e9fc2214
Subject: GMP serialization
From: bukka@php.net (Jakub Zelenka)

--047d7b33d17448b8da04e9fc2214
Content-Type: text/plain; charset=ISO-8859-1

Hi,

There is a new patch that is changing serialization in GMP:

http://git.php.net/?p=php-src.git;a=commitdiff;h=4218e89f8df4ca3897e3aad595e0c2c9cf4c3aca

It fixes the bug but it also introduces BC break for serialized string
(custom serialization replaced the std object serialization). The problem
is that there can be users (one big user is WordPress) that saves the
serialized into DB which can lead to the difficulties when updating PHP
version.

I think that the new serialized API that I proposed some time ago (
https://wiki.php.net/rfc/internal_serialize_api ) could solve the problem.
I actually just wrote a patch that fixes the same bug and uses a new
serialize API:

https://github.com/bukka/php-src/commit/d3c04d001f3a86409d45a0f268ea4945fa53489b

I also created few benchmarks:
== BEFORE THE FIX ==
O:3:"GMP":1:{s:3:"num";s:2:"42";}
SERIALIZATION: time for 100000 iterations: 0.230430
UNSERIALIZATION: time for 100000 iterations: 0.469807

== AFTER THE FIX ==
C:3:"GMP":15:{s:2:"42";a:0:{}}
SERIALIZATION: time for 100000 iterations: 0.319425
UNSERIALIZATION: time for 100000 iterations: 0.354855

== AFTER MY PATCH (new serialization API)  ==
O:3:"GMP":1:{s:3:"num";s:2:"42";}
SERIALIZATION: time for 100000 iterations: 0.195529
UNSERIALIZATION: time for 100000 iterations: 0.241552

As you can see after using a new serialization API, the performance has
improved and mainly the serialized string is the same as the one before the
fix.

If anyone is interested: I also have an initial patch for DateTime (still
need to do quite a few things so it's incomplete but it works fine for
DateTime):
https://github.com/bukka/php-src/compare/date_using_serialize_api#files_bucket

I think that the API patch is ready
https://github.com/bukka/php-src/compare/serialize_internal_api . There are
few extension that could profit from it so I think that it could be a
useful addition...

Any thoughts?

Thanks for reading this long and boring email... :)

Regards

Jakub

--047d7b33d17448b8da04e9fc2214--