Does Django's render_to_response cache unrendered templates or load from disk every time?

Fetching a template from disk is significantly slower than pulling it out of somethin开发者_运维知识库g like memcached, so loading them from disk each time is wasteful.

Does Django cache unrendered templates in memory or in the CACHE_BACKEND or do I have to implement that myself?

Django comes with a cached template loader as well.

I can't seem to find anyplace where the filesystem loader might cache the templates. It sure does look like it loads them every time.

This probably isn't really a problem, though. For one thing, the filesystem loader does exactly what it says, it reads the templates from disk every time they are used. It will always pick up changes to the on disk template as soon as the changes are made.

If you don't want that behavior, you might want to use a different loader. In fact, that's pretty easy to do; just subclass from django.template.loader.filesystem.Loader, overload load_template, and enjoy.

I'm pretty suspicious of caching, though. If I were to implement such a solution, I'd make sure that cache evictions happen often. Like every 10 seconds.

You have to enable cache yourself. By default, every time you render a template it is read from disk, parsed and rendered.

There are many ways to implement cache in Django. All of them described in the Django's cache framework help page.

Depending on what you're trying to achieve you can cache entire requests or only template blocks. Read the page above for details.

I suspect that if the template is read every time, it will probably stay in the buffer cache of the kernel most of the time. You shouldn't invest any time in such optimization as long as you don't have a prove, that this is really something that drains don performance.