AppSearchSession

See also:

• GlobalSearchSession

close

public void close ()

Closes the AppSearchSession to persist all schema and document updates, additions, and deletes to disk.

getByDocumentId

public void getByDocumentId (GetByDocumentIdRequest request, 
                Executor executor, 
                BatchResultCallback<String, GenericDocument> callback)

Gets GenericDocument objects by document IDs in a namespace from the AppSearchSession database.

getNamespaces

public void getNamespaces (Executor executor, 
                Consumer<AppSearchResult<Set<String>>> callback)

Retrieves the set of all namespaces in the current database with at least one document.

getSchema

public void getSchema (Executor executor, 
                Consumer<AppSearchResult<GetSchemaResponse>> callback)

Retrieves the schema most recently successfully provided to setSchema(SetSchemaRequest, Executor, Executor, Consumer).

getStorageInfo

public void getStorageInfo (Executor executor, 
                Consumer<AppSearchResult<StorageInfo>> callback)

Gets the storage info for this AppSearchSession database.

This may take time proportional to the number of documents and may be inefficient to call repeatedly.

put

public void put (PutDocumentsRequest request, 
                Executor executor, 
                BatchResultCallback<String, Void> callback)

Indexes documents into the AppSearchSession database.

Each GenericDocument object must have a schemaType field set to an AppSearchSchema type that has been previously registered by calling the setSchema(SetSchemaRequest, Executor, Executor, Consumer) method.

remove

public void remove (RemoveByDocumentIdRequest request, 
                Executor executor, 
                BatchResultCallback<String, Void> callback)

Removes GenericDocument objects by document IDs in a namespace from the AppSearchSession database.

Removed documents will no longer be surfaced by search(String, SearchSpec) or getByDocumentId(GetByDocumentIdRequest, Executor, BatchResultCallback) calls.

Once the database crosses the document count or byte usage threshold, removed documents will be deleted from disk.

remove

public void remove (String queryExpression, 
                SearchSpec searchSpec, 
                Executor executor, 
                Consumer<AppSearchResult<Void>> callback)

Removes GenericDocuments from the index by Query. Documents will be removed if they match the queryExpression in given namespaces and schemaTypes which is set via SearchSpec.Builder.addFilterNamespaces(String) and SearchSpec.Builder.addFilterSchemas.

An empty queryExpression matches all documents.

An empty set of namespaces or schemaTypes matches all namespaces or schemaTypes in the current database.

reportUsage

public void reportUsage (ReportUsageRequest request, 
                Executor executor, 
                Consumer<AppSearchResult<Void>> callback)

Reports usage of a particular document by namespace and ID.

A usage report represents an event in which a user interacted with or viewed a document.

For each call to reportUsage(ReportUsageRequest, Executor, Consumer), AppSearch updates usage count and usage recency metrics for that particular document. These metrics are used for ordering search(String, SearchSpec) results by the SearchSpec.RANKING_STRATEGY_USAGE_COUNT and SearchSpec.RANKING_STRATEGY_USAGE_LAST_USED_TIMESTAMP ranking strategies.

Reporting usage of a document is optional.

search

public SearchResults search (String queryExpression, 
                SearchSpec searchSpec)

Retrieves documents from the open AppSearchSession that match a given query string and type of search provided.

Query strings can be empty, contain one term with no operators, or contain multiple terms and operators.

For query strings that are empty, all documents that match the SearchSpec will be returned.

For query strings with a single term and no operators, documents that match the provided query string and SearchSpec will be returned.

The following operators are supported:

• AND (implicit)

  AND is an operator that matches documents that contain all provided terms.

  NOTE: A space between terms is treated as an "AND" operator. Explicitly including "AND" in a query string will treat "AND" as a term, returning documents that also contain "AND".

  Example: "apple AND banana" matches documents that contain the terms "apple", "and", "banana".

  Example: "apple banana" matches documents that contain both "apple" and "banana".

  Example: "apple banana cherry" matches documents that contain "apple", "banana", and "cherry".

• OR

  OR is an operator that matches documents that contain any provided term.

  Example: "apple OR banana" matches documents that contain either "apple" or "banana".

  Example: "apple OR banana OR cherry" matches documents that contain any of "apple", "banana", or "cherry".

• Exclusion (-)

  Exclusion (-) is an operator that matches documents that do not contain the provided term.

  Example: "-apple" matches documents that do not contain "apple".

• Grouped Terms

  For queries that require multiple operators and terms, terms can be grouped into subqueries. Subqueries are contained within an open "(" and close ")" parenthesis.

  Example: "(donut OR bagel) (coffee OR tea)" matches documents that contain either "donut" or "bagel" and either "coffee" or "tea".

• Property Restricts

  For queries that require a term to match a specific AppSearchSchema property of a document, a ":" must be included between the property name and the term.

  Example: "subject:important" matches documents that contain the term "important" in the "subject" property.

The above description covers the query operators that are supported on all versions of AppSearch. Additional operators and their required features are described below.

LIST_FILTER_QUERY_LANGUAGE: This feature covers the expansion of the query language to conform to the definition of the list filters language (https://aip.dev/160). This includes:

• addition of explicit 'AND' and 'NOT' operators
• property restricts are allowed with groupings (ex. "prop:(a OR b)")
• addition of custom functions to control matching

The newly added custom functions covered by this feature are:

• createList(String...)
• search(String, List<String>)
• propertyDefined(String)

createList takes a variable number of strings and returns a list of strings. It is for use with search.

search takes a query string that will be parsed according to the supported query language and an optional list of strings that specify the properties to be restricted to. This exists as a convenience for multiple property restricts. So, for example, the query `(subject:foo OR body:foo) (subject:bar OR body:bar)` could be rewritten as `search("foo bar", createList("subject", "body"))`.

propertyDefined takes a string specifying the property of interest and matches all documents of any type that defines the specified property (ex. `propertyDefined("sender.name")`). Note that propertyDefined will match so long as the document's type defines the specified property. Unlike the "hasProperty" function below, this function does NOT require that the document actually hold any values for this property.

NUMERIC_SEARCH: This feature covers numeric search expressions. In the query language, the values of properties that have AppSearchSchema.LongPropertyConfig.INDEXING_TYPE_RANGE set can be matched with a numeric search expression (the property, a supported comparator and an integer value). Supported comparators are <, <=, ==, >= and >.

Ex. `price < 10` will match all documents that has a numeric value in its price property that is less than 10.

VERBATIM_SEARCH: This feature covers the verbatim string operator (quotation marks).

Ex. `"foo/bar" OR baz` will ensure that 'foo/bar' is treated as a single 'verbatim' token.

LIST_FILTER_HAS_PROPERTY_FUNCTION: This feature covers the "hasProperty" function in query expressions, which takes a string specifying the property of interest and matches all documents that hold values for this property. Not to be confused with the "propertyDefined" function, which checks whether a document's schema has defined the property, instead of whether a document itself has this property.

Ex. `foo hasProperty("sender.name")` will return all documents that have the term "foo" AND have values in the property "sender.name". Consider two documents, documentA and documentB, of the same schema with an optional property "sender.name". If documentA sets "foo" in this property but documentB does not, then `hasProperty("sender.name")` will only match documentA. However, `propertyDefined("sender.name")` will match both documentA and documentB, regardless of whether a value is actually set.

SCHEMA_EMBEDDING_PROPERTY_CONFIG: This feature covers the "semanticSearch" and "getEmbeddingParameter" functions in query expressions, which are used for semantic search.

Usage: semanticSearch(getEmbeddingParameter({embedding_index}), {low}, {high}, {metric})

• semanticSearch matches all documents that have at least one embedding vector with a matching model signature (see EmbeddingVector.getModelSignature()) and a similarity score within the range specified based on the provided metric.
• getEmbeddingParameter({embedding_index}) retrieves the embedding search passed in SearchSpec.Builder.addEmbeddingParameters based on the index specified, which starts from 0.
• "low" and "high" are floating point numbers that specify the similarity score range. If omitted, they default to negative and positive infinity, respectively.
• "metric" is a string value that specifies how embedding similarities should be calculated. If omitted, it defaults to the metric specified in SearchSpec.Builder.setDefaultEmbeddingSearchMetricType(int). Possible values:
  • "COSINE"
  • "DOT_PRODUCT"
  • "EUCLIDEAN"

Examples:

• Basic: semanticSearch(getEmbeddingParameter(0), 0.5, 1, "COSINE")
• With a property restriction: property1:semanticSearch(getEmbeddingParameter(0), 0.5, 1)
• Hybrid: foo OR semanticSearch(getEmbeddingParameter(0), 0.5, 1)
• Complex: (foo OR semanticSearch(getEmbeddingParameter(0), 0.5, 1)) AND bar

SEARCH_SPEC_SEARCH_STRING_PARAMETERS: This feature covers the "getSearchStringParameter" function in query expressions, which substitutes the string provided at the same index in SearchSpec.Builder.addSearchStringParameters into the query as plain text. This string is then segmented, normalized and stripped of punctuation-only segments. The remaining tokens are then AND'd together. This function is useful for callers who wish to provide user input, but want to ensure that that user input does not invoke any query operators.

Usage: getSearchStringParameter({search_parameter_strings_index})

Ex. `foo OR getSearchStringParameter(0)` with SearchSpec.getSearchStringParameters returning {"bar OR baz."}. The string "bar OR baz." will be segmented into "bar", "OR", "baz", ".". Punctuation is removed and the segments are normalized to "bar", "or", "baz". This query will be equivalent to `foo OR (bar AND or AND baz)`.

Additional search specifications, such as filtering by AppSearchSchema type or adding projection, can be set by calling the corresponding SearchSpec.Builder setter.

This method is lightweight. The heavy work will be done in SearchResults.getNextPage(Executor, Consumer).

searchSuggestion

public void searchSuggestion (String suggestionQueryExpression, 
                SearchSuggestionSpec searchSuggestionSpec, 
                Executor executor, 
                Consumer<AppSearchResult<List<SearchSuggestionResult>>> callback)

Retrieves suggested Strings that could be used as queryExpression in search(java.lang.String, android.app.appsearch.SearchSpec) API.

The suggestionQueryExpression can contain one term with no operators, or contain multiple terms and operators. Operators will be considered as a normal term. Please see the operator examples below. The suggestionQueryExpression must end with a valid term, the suggestions are generated based on the last term. If the input suggestionQueryExpression doesn't have a valid token, AppSearch will return an empty result list. Please see the invalid examples below.

Example: if there are following documents with content stored in AppSearch.

• document1: "term1"
• document2: "term1 term2"
• document3: "term1 term2 term3"
• document4: "org"

Search suggestions with the single term suggestionQueryExpression "t", the suggested results are:

• "term1" - Use it to be queryExpression in search(String, SearchSpec) could get 3 SearchResults, which contains document 1, 2 and 3.
• "term2" - Use it to be queryExpression in search(String, SearchSpec) could get 2 SearchResults, which contains document 2 and 3.
• "term3" - Use it to be queryExpression in search(String, SearchSpec) could get 1 SearchResult, which contains document 3.

Search suggestions with the multiple term suggestionQueryExpression "org t", the suggested result will be "org term1" - The last token is completed by the suggested String.

Operators in search(String, SearchSpec) are supported.

NOTE: Exclusion and Grouped Terms in the last term is not supported.

example: "apple -f": This Api will throw an AppSearchException with AppSearchResult.RESULT_INVALID_ARGUMENT.

example: "apple (f)": This Api will return an empty results.

Invalid example: All these input suggestionQueryExpression don't have a valid last token, AppSearch will return an empty result list.

• "" - Empty suggestionQueryExpression.
• "(f)" - Ending in a closed brackets.
• "f:" - Ending in an operator.
• "f " - Ending in trailing space.

setSchema

public void setSchema (SetSchemaRequest request, 
                Executor workExecutor, 
                Executor callbackExecutor, 
                Consumer<AppSearchResult<SetSchemaResponse>> callback)

Sets the schema that represents the organizational structure of data within the AppSearch database.

Upon creating an AppSearchSession, setSchema(SetSchemaRequest, Executor, Executor, Consumer) should be called. If the schema needs to be updated, or it has not been previously set, then the provided schema will be saved and persisted to disk. Otherwise, setSchema(SetSchemaRequest, Executor, Executor, Consumer) is handled efficiently as a no-op call.