Clients may use this option to prevent unnecessary indexing of long tokens. Depending on * the use case, indexing all of 'Supercalifragilisticexpialidocious' may be unnecessary - a * user is unlikely to type that entire query. So only indexing the first n bytes may still * provide the desired behavior without wasting resources. */ int getMaxTokenLength(); /** * The size (measured in bytes) at which Icing's internal indices should be merged. Icing * buffers changes together before merging them into a more compact format. When the buffer * exceeds index_merge_size during a Put operation, the buffer is merged into the larger, more * compact index. * *
This more compact index is more efficient to search over as the index grows larger and has * smaller system health impact. * *
Setting a low index_merge_size increases the frequency of merges - increasing * indexing-time latency and flash wear. Setting a high index_merge_size leads to larger * resource usage and higher query latency. */ int getIndexMergeSize(); /** * Whether to use namespace id or namespace name to build up fingerprint for * document_key_mapper_ and corpus_mapper_ in document store. */ boolean getDocumentStoreNamespaceIdFingerprint(); /** * The threshold of the percentage of invalid documents at which to rebuild index during * optimize. * *
We rebuild index if and only if |invalid_documents| / |all_documents| >= threshold. * *
Rebuilding the index could be faster than optimizing the index if we have removed most of * the documents. Based on benchmarks, 85%~95% seems to be a good threshold for most cases. */ float getOptimizeRebuildIndexThreshold(); /** * The level of gzip compression for documents in the Icing document store. * *
NO_COMPRESSION = 0, BEST_SPEED = 1, BEST_COMPRESSION = 9 */ int getCompressionLevel(); /** * Whether to allow circular references between schema types for the schema definition. * *
Even when set to true, circular references are still not allowed in the following cases: * 1. All edges of a cycle have index_nested_properties=true 2. One of the types in the cycle * has a joinable property, or depends on a type with a joinable property. */ boolean getAllowCircularSchemaDefinitions(); /** * Flag for {@link com.google.android.icing.proto.SearchSpecProto}. * *
Whether to use the read-only implementation of IcingSearchEngine::Search. * *
The read-only version enables multiple queries to be performed concurrently as it only * acquires the read lock at IcingSearchEngine's level. Finer-grained locks are implemented * around code paths that write changes to Icing during Search. */ boolean getUseReadOnlySearch(); /** * Flag for {@link com.google.android.icing.proto.IcingSearchEngineOptions}. * *
Whether or not to pre-map the potential memory region used by the PersistentHashMap. This * will avoid the need to re-map the mmapping used by PersistentHashMap whenever the underlying * storage grows. */ boolean getUsePreMappingWithFileBackedVector(); /** * Flag for {@link com.google.android.icing.proto.IcingSearchEngineOptions}. * *
Whether or not to use the PersistentHashMap in the QualifiedIdTypeJoinableIndex. If false, * we will use the old IcingDynamicTrie to store key value pairs. */ boolean getUsePersistentHashMap(); /** * Flag for {@link com.google.android.icing.proto.ResultSpecProto}. * *
The maximum byte size to allow in a single page. This limit is only loosely binding. * AppSearch will add results to the page until either 1) AppSearch has retrieved {@link * SearchSpec#getResultCountPerPage()} results or 2) total size of the page exceeds this value. * Therefore, AppSearch will always retrieve at least a single result, even if that result * exceeds this limit. */ int getMaxPageBytesLimit(); /** * Flag for {@link com.google.android.icing.proto.IcingSearchEngineOptions}. * *
Threshold for integer index bucket split. Integer index stores hits in several buckets, * and splits if # of hits in a single bucket exceed the threshold. Splitting bucket accelerates * numeric search exact query, but potentially downgrades the performance of range query. * *
This flag is for rolling out new threshold 65536. If identifying any issues, then change * it back to 341 (the previous bucket split threshold, capacity of full max-sized posting * list). */ int getIntegerIndexBucketSplitThreshold(); /** * Flag for {@link com.google.android.icing.proto.IcingSearchEngineOptions}. * *
Whether Icing should sort and merge its lite index HitBuffer unsorted tail at indexing * time. * *
If set to true, the HitBuffer will be sorted at indexing time after exceeding the sort * threshold. If false, the HifBuffer will be sorted at querying time, before the first query * after inserting new elements into the HitBuffer. */ boolean getLiteIndexSortAtIndexing(); /** * Flag for {@link com.google.android.icing.proto.IcingSearchEngineOptions}. * *
Size (in bytes) at which Icing's lite index should sort and merge the HitBuffer's unsorted * tail into the sorted head for sorting at indexing time. Size specified here is unsorted tail * section. * *
Setting a lower sort size reduces querying latency at the expense of indexing latency. */ int getLiteIndexSortSize(); /** * Flag for {@link com.google.android.icing.proto.IcingSearchEngineOptions}. * *
Whether to use the new qualified Id join index. */ boolean getUseNewQualifiedIdJoinIndex(); /** * Flag for {@link com.google.android.icing.proto.IcingSearchEngineOptions}. * *
Whether to build the metadata hits used for property existence check, which is required to * support the hasProperty function in advanced query. */ boolean getBuildPropertyExistenceMetadataHits(); }