Cached Search

- See It Live

Running locally:

• bundle install

• rake db:migrate

• Make sure you have redis installed: brew install redis

• Make sure redis is running, redis-server

• foreman start (to run both the web/worker processes)

Caching Strategy:

• If a product_search exists matching query AND has a cached_at within 7 days

  • show that one

• If a product_search exists matching query BUT has a cached_at older than 7 days

  • this one is expired
  • delete all associated result records
  • fetch and associate a new result set
  • update the cached_at date to Date.current

• If NO product_search exists for a query

  • create a new one and populate the results

Pros

• For frequently requested query's we have a positive cache_hit every time (minus one every 7 days).
• Avoid a cache stampede, by avoiding a mass cache_refresh all at once.

Cons

• Infrequently requested query results may not have their cache busted often, or with any regularity (maybe this is okay?)
• We might be needlessly updating caches every 7 days if result sets do not change within that time interval.

Known Shortcomings

• Currently I call destroy_all during a re-saturation of results for a particular product_search
  • This produces a DELETE FROM query for each associated result record, which is an n+1 situation :(
• Did not implement pagination, however, based on the docs for the semantic3 gem, it appears as though one might pass through an offset parameter into the get_products() endpoint.