Django autocomplete fields, better to pre-populate?

Question

I'm looking to integrate jqueryui-autocomplete with a django project. My starting point for a solution was https://code.djangoproject.com/wiki/AutoCompleteSolutions .

The solutions i've come across use *__icontains or *__startswith , although these searches can of course be cached (looking at http://djangosnippets.org/snippets/233/ ). I wondered if pre-populating the page with a relatively small list of terms (10,000?) in an array would actually save on database hits & be a better solution. For example, outputting the terms via django in one go:

return [str(e.keyword) for e in CM_Dict.objects.all()]

and then using the resulting list as source: in a really basic jquery-ui-autocomplete config.

Is there a good way to benchmark stuff like this? Does anyone use django+autocomplete in production?

Edit: I should add that the autocomplete is only for logged in users but the data isn't sensitive in any way.

Answer 1

This sounds like an early optimization to me. Would it save on DB hits? Maybe though you have to hit the DB to pre-populate anyway so your initial page load will suffer. You could cache the pre-population data to avoid the DB hit but you could do that on the server-side view as well. That would seem to integrate better with the existing Django autocomplete solutions.

I've used a couple of the auto-complete solutions in production. I'm also the author of one of them: django-selectable . Benchmarking this type of change requires knowing how expensive each query is and how often it's used. Parsing server logs for the counts/load times would be a start. Overall I would be surprised if your auto-complete ajax views are the bottle-neck in your application.