In simple words, when you make a request in Jackett, the results are saved in memory (cache). The next request will return results form the cache improving response time and making fewer requests to the sites.
* We assume all indexers/sites are stateless, the same request return the same response. If you change the search term, categories or something in the query Jackett has to make a live request to the indexer.
* There are some situations when we don't want to use the cache:
** When we are testing the indexers => if query.IsTest results are not cached
** When the user updates the configuration of one indexer => We call CleanIndexerCache to remove cached results before testing the configuration
** When there is some error/exception in the indexer => The results are not cached so we can retry in the next request
* We want to limit the memory usage, so we try to remove elements from cache ASAP:
** Each indexer can have a maximum number of results in memory. If the limit is exceeded we remove old results
** Cached results expire after some time
* Users can configure the cache or even disable it
* When category ids in the indexer are "strings" we create a unique hash to make it compatible with Torznab
* This PR also fix several issues related to custom cats
* Core: Categories are stored in a real tree
* Sorting: First Torznab categories sorted by Id and then custom cats sorted by Name
* Filtering: Results with child category are not removed when searching by parent category. Details in #8049
* Jacket UI: Add parent category when at least one child category exists
* Torznab (caps): Remove non existent children categories. Remove duplicated categories. Details in #10006