如何为PHP中的多客户共享代码迁移到utf8

Question

in our company we use a proprietary CMS, this CMS is in use for almoust a thousand websites and is aged of approximatively 15 years. (it has evolved but is very features rich)

until now whe have always used iso-8859-1 as charset, but whe have the nescessity to use utf-8 for one project.

there are my questions :

1. do you think that this way to do is good (maintaining one SVN version, automatic conversion to the utf8 search/replace the problematic php functions and do some magic there).
2. Have you done this kind of evolution before, what is hasardous regarding to you?

---

TLDR infos :

• the core of our CMS is centralised (SVN) and deployed (rsynch) on a specific path on each of our servers, this path is in the include path of each websites.
• The databases are different for each projects (but same structure for the core tables).
• each website use a document_root holding website's specifics files (media, js, specific PHP code)

in this configuration, we cannot migrate every website on a single time (ie : because there is local code). So I want to make two version of our core : one in iso-8859-1 and the other in utf8. From now I think that I'm going to develop a script on our deployement system, this script, will create a copy utf8 encoded of our core before the rsynch-ing.

My concern is for example about all the "mb_ " stuff in PHP that won't be called, so i'll have to search/replace every php native fucntion to replace it by a custome one that will use the "mb_ " version if nescessary (furthermore, the overloding of those functions must be in the php.ini file, it cannot be defined in the .htaccess of a particulare website ( source ))

PS : sorry for my poor english, it's not my native language :(

Answer 1

So here is how I handled it :

• each server is either iso-8859-1 or utf-8, and have dedicated conf (ie : mbstring.func_overload)
• before each deployement a script create a copy of the iso version and convert it to utf8
• each server get either utf-8 or iso-8859-1 source code

for each individual CMS, a convertion tool exist, it convert filesystem, database charset/collation, configuration (ie : connect to mysql with utf8), ....

for non compatible functions such as utf8_encode, I search/replaced it with a function who is in charge to not call the function if the server is utf8