How to implement semaphores in PHP without PHP Semaphore?

0 Views

Question:

How can I implement shared memory variable in PHP without the semaphore package (http://php.net/manual/en/function.shm-get-var.php) ?

Read More

Context

  • I have a simple web application (actually a plugin for WordPress)
  • this gets a url
  • this then checks the database if that url already exists
  • if not then it goes out and does some operations
  • and then writes the record in the database with the url as unique entry

What happens in reality is that 4,5,6 … sessions at the same time request the url and I get up to 9 duplicate entries in the database of the url.. (possibly 9 because the processing time and database write of the first entry takes just enough time to let 9 other requests fall through). After that all requests read the correct entry that the record already exists so that is good.

Since it is a WordPress plugin there will be many users on all kind of shared hosting platforms with variable compiles/settings of PHP.

So I’m looking for a more generic solution. I cant use database or text file writes since these will be too slow. while i write to the db the next session will already have passed.

fyi: the database code: http://plugins.svn.wordpress.org/wp-favicons/trunk/includes/server/plugins/metadata_favicon/inc/class-favicon.php

update

Using a unique key on a new md5 hash of the uri together with try catches around it seems to work.

I found 1 duplicate entry with 

SELECT uri, COUNT( uri ) AS NumOccurrences
FROM edl40_21_wpfavicons_1
GROUP BY uri
HAVING (
COUNT( uri ) >1
)
LIMIT 0 , 30

So I thought it did not work but this was because they were: 

http://en.wikipedia.org/wiki/Book_of_the_dead
http://en.wikipedia.org/wiki/Book_of_the_Dead

(capitals grin)

Post Views: 0

Related posts

Leave a Reply

2 comments

  1. This could be achieved with MySQL.

    You could do it explicitly by locking the table from read access. This will prevent any read access from the entire table though, so may not be preferable. http://dev.mysql.com/doc/refman/5.5/en/lock-tables.html

    Otherwise if the field in the table is classified as unique, then when the next session tries to write the same URL to the table they will get an error, you can catch that error and continue as there’s no need to do anything if the entry is already there. The only time wasted is the possibility of two or more sessions creating the same URL, the result is still one record, as the database won’t add the same unique URL again.

    As discussed in comments, because the length of a URL could be very long, and fixed length unique hash can help overcome that issue.

  2. There are other shared memory modules in PHP (shmop or APC for example), but I think what you are saying is that there is an issue relying on non-standard/not pre-installed libraries.

    My suggestion is that before you go and do “other operations” you need to make an entry in the database, perhaps with a status of “compiling” (or something) so you know it is still not available. This way you don’t run into issues with getting multiple entries. I would also be sure you are using transactions when they are available so your commits are atomic.

    Then, when you “other operations” are done, update the database entry to “available” and do whatever else it is you need to do.