Package ca.uhn.fhir.jpa.search.cache

Interface ISearchCacheSvc

All Known Implementing Classes:

DatabaseSearchCacheSvcImpl

public interface ISearchCacheSvc

Method Summary

Modifier and Type	Method	Description
Optional<Search>	fetchByUuid(String theUuid, RequestPartitionId theRequestPartitionId)	Fetch a search using its UUID.
Optional<Search>	findCandidatesForReuse(String theResourceType, String theQueryString, Instant theCreatedAfter, RequestPartitionId theRequestPartitionId)	Look for any existing searches matching the given resource type and query string.
void	pollForStaleSearchesAndDeleteThem(RequestPartitionId theRequestPartitionId, Instant theDeadline)	This method will be called periodically to delete stale searches.
Search	save(Search theSearch, RequestPartitionId theRequestPartitionId)	Places a new search of some sort in the cache, or updates an existing search.
Optional<Search>	tryToMarkSearchAsInProgress(Search theSearch, RequestPartitionId theRequestPartitionId)	TODO: this is perhaps an inappropriate responsibility for this service

Method Details

save

Search save(Search theSearch,
 RequestPartitionId theRequestPartitionId)

Places a new search of some sort in the cache, or updates an existing search. The search passed in is guaranteed to have a UUID so that is a good candidate for consistent identification.

Parameters:

theSearch - The search to store

Returns:

Returns a copy of the search as it was saved. Callers should use the returned Search object for any further processing.

fetchByUuid

Optional<Search> fetchByUuid(String theUuid,
 RequestPartitionId theRequestPartitionId)

Fetch a search using its UUID. The search should be fully loaded when it is returned (i.e. includes are fetched, so that access to its fields will not cause database errors if the current tranaction scope ends.

Parameters:

theUuid - The search UUID

Returns:

The search if it exists

tryToMarkSearchAsInProgress

Optional<Search> tryToMarkSearchAsInProgress(Search theSearch,
 RequestPartitionId theRequestPartitionId)

TODO: this is perhaps an inappropriate responsibility for this service

This method marks a search as in progress, but should only allow exactly one call to do so across the cluster. This is done so that if two client threads request the next page at the exact same time (which is unlikely but not impossible) only one will actually proceed to load the next results and the other will just wait for them to arrive.

Parameters:

theSearch - The search to mark

Returns:

This method should return an empty optional if the search was not marked (meaning that another thread succeeded in marking it). If the search doesn't exist or some other error occurred, an exception will be thrown instead of Optional.empty()

findCandidatesForReuse

Optional<Search> findCandidatesForReuse(String theResourceType,
 String theQueryString,
 Instant theCreatedAfter,
 RequestPartitionId theRequestPartitionId)

Look for any existing searches matching the given resource type and query string.

This method is allowed to perform a "best effort" return, so it can return searches that don't match the query string exactly, or which have a created timestamp before theCreatedAfter date. The caller is responsible for removing any inappropriate Searches and picking the most relevant one.

Parameters:

theResourceType - The resource type of the search. Results MUST match this type

theQueryString - The query string. Results SHOULD match this type

theCreatedAfter - Results SHOULD not include any searches created before this cutoff timestamp

theRequestPartitionId - Search should examine only the requested partitions. Cache MUST not return results matching the given partition IDs

Returns:

A collection of candidate searches

pollForStaleSearchesAndDeleteThem

void pollForStaleSearchesAndDeleteThem(RequestPartitionId theRequestPartitionId,
 Instant theDeadline)

This method will be called periodically to delete stale searches. Implementations are not required to do anything if they have some other mechanism for expiring stale results other than manually looking for them and deleting them.