# Opposite of Cache Sweeper?

**URL:** <https://www.ruby-forum.com/t/opposite-of-cache-sweeper/142071>\
**Category:** Rails\
**Created:** [July 3, 2008, 6:31am UTC](https://www.ruby-forum.com/t/opposite-of-cache-sweeper/142071 "2008-07-03T06:31:27Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![iplat](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/i/ed655f/32.png) [@iplat](https://www.ruby-forum.com/u/iplat)\
**Post date:** [July 3, 2008, 6:31am UTC](https://www.ruby-forum.com/t/opposite-of-cache-sweeper/142071/1 "2008-07-03T06:31:27Z")

</div>

I’m looking for an acceptable way to cache images when they are  
inserted into the database. Essentially an observer for the model,  
however, i feel that this is not a concern of the model.

Currently i’m using the caches\_action method to cache these images.  
This only caches when the request is made for the image though.  
Processing these images/pdf’s with Rmagick takes some time so i’d  
like for the image to be cached before it is requested.

Any suggestions?

---

<div class="post-metadata">

**Author:** ![iplat](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/i/ed655f/32.png) [@iplat](https://www.ruby-forum.com/u/iplat)\
**Post date:** [July 3, 2008, 6:47am UTC](https://www.ruby-forum.com/t/opposite-of-cache-sweeper/142071/2 "2008-07-03T06:47:13Z")

</div>

If you want them on disk (not in the database) then why are you not just  
storing them in the file system and then referencing them in the db? Do  
all  
the work on the upload (rmagick / whatever).

But furthermore to you question:

I would argue that caching them to disk upon insertion is the job of the  
model (after\_save) but if that’s not clean enough for you then you  
should  
investigate active record observers, or some sort of batch operation  
that  
runs separately.

---

<div class="post-metadata">

**Author:** ![iplat](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/i/ed655f/32.png) [@iplat](https://www.ruby-forum.com/u/iplat)\
**Post date:** [July 3, 2008, 3:06pm UTC](https://www.ruby-forum.com/t/opposite-of-cache-sweeper/142071/3 "2008-07-03T15:06:33Z")

</div>

Thanks Brian, I chose not to store the images on disk because it  
limits the application to retrieve images from one specific server.  
Using MySQL to store images i could potentially replicate and  
distribute traffic amongst different servers. I wanted to leave my  
options open.

Thank you for your suggestion. However, from what i know about  
observers they are very close to the model. I could be wrong. Do you  
know a good way to invoke an action in a controller from a rake task?

---

<div class="post-metadata">

**Author:** ![iplat](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/i/ed655f/32.png) [@iplat](https://www.ruby-forum.com/u/iplat)\
**Post date:** [July 3, 2008, 8:27pm UTC](https://www.ruby-forum.com/t/opposite-of-cache-sweeper/142071/4 "2008-07-03T20:27:50Z")

</div>

Explain what you’re doing first.

If you want to cache the stuff to disk to avoid constant processing,  
then  
write them to disk and reference in the db. Scaling out to multiple  
machines is a non-issue - just mount the file system on multiple  
machines  
and all is good. Or investigate Amazon S3 for your storage. If you’ve  
used  
attachment\_fu, you can easily switch from filestore to database store to  
s3  
without much effort.

Why wouldn’t you want to do this close to the model? Why would you want  
a  
controller action to be invoked via a rake task?

My implementation would be this:

class Attachment \> ActiveRecord::Base

…

after\_create :cache\_to\_disk

def cache\_to\_disk  
# code that writes to some folder specified by a configuration  
value  
end

end

When things are saved, you cache to disk automatically.

When you need to force a rewrite of the files on your file system, use  
Rake

desc “cache files”  
task :write\_attachments =\> :environment do  
attachments = Attachment.find(:all)  
attachments.each{|a.cache\_to\_disk}  
end

Personally, I would not store files in the database at all. It seems  
like a  
good idea at first but files should be served by the file server. You  
can  
write code to replicate your filesystem across servers, you could use a  
shared folder or an nfs mount, or you could use S3 instead of doing  
something like this.

Just my .02, and feel free to argue, just hoping to help you find the  
best  
solution.

Good luck!