Just a placeholder for now.

Some ideas...
* Bypass theming functions
* Use separate script with lower bootstrap level at query time
* Remove some arguments from queries, like bf (for instance, "recip(ms(NOW,ds_created),3.16e-11,0.5,0.5)^700.0" might be added, and is not really necessary).
* Implement a suggestion component in the config, like mentioned here: http://solr.pl/en/2010/11/15/solr-and-autocomplete-part-2/

Of course, each needs a proper benchmark/comparison to see if any improvement is possible.

Comments

janusman’s picture

janusman CreditAttribution: janusman commented
Issue summary: View changes
janusman’s picture

janusman CreditAttribution: janusman commented

@NickVh also suggests looking at http://docs.lucidworks.com/display/solr/Suggester

janusman’s picture

janusman CreditAttribution: janusman commented

Another possible strategy that came up could be:
* wait until a small number of characters has been typed (say, 3 characters minimum).
* only then, send the request to the backend, and return the largest practical dataset possible, optimized by returning only "popular" or "best" hits.
* process that dataset on the client side (JS) to only show relevant hits, for the current and further typed characters.

Pros:
* Less requests to the backend
* Cacheable
* Quicker?
Cons:
* Transfer time?
* Memory load on browser?