Introduction

AllegroGraph implements the SPARQL 1.1 standard, with certain conformance notes. Deviations from the standard are considered bugs.

SPARQL query execution in AllegroGraph is influenced by the configuration of the SPARQL dataset, setting query options, and by the choice of query engine (see SPARQL Query Engines).

The query engine will generate query execution warnings for suspicious query patterns.

AllegroGraph also provides support for SPIN, and reification of triples using triple IDs.

AllegroGraph can also be queried using Prolog.


SPARQL implementation notes

SPARQL CONSTRUCT query form combines the triples constructed from the bindings produced by the WHERE clause into a single RDF graph by set union (i.e. all the duplicates are removed). For example, given the data set

:s :p 1.  
:s :p 2.  
:s :p 3. 

the following query

CONSTRUCT { ?s ?p 4 }  
WHERE { ?s ?p ?o } 

produces the single triple

:s :p 4. 

instead of three triples

:s :p 4  
:s :p 4  
:s :p 4 

which is one of the possible interpretations of the specification here. Note that some of the implementations return the latter result.


Executing SPARQL queries

Queries can be run in various ways:

Limiting results of a query

Query results can be limited with a LIMIT clause, or by control boxes in AGWebView. And individual users can have the number of results limited (as well as export and duplicates export operations). See Limiting the results a user can see for details.

The size of the limit (for all users and repos) is set by the QueryResultsLimit configuration directive, whose default value is 1000.

For complex queries the limit is applied using the rule

<query-limit> + <query-offset> <= <query-results-limit> 

that is, if query asks for results beyond the results limit with its limit and offset modifiers, the new modifier values will be chosen to fit the above condition.

Query results caching

AllegroGraph supports caching of SPARQL query results. When results caching is enabled, the same query (with perhaps different offsets and limits) can be performed multiple times without having to do the actual query processing.

This behavior has to be enabled explicitly on per-query basis using the allowCachingResults query option.

Results cache size can be configured using the configuration directives QueryResultsCacheSize and QueryResultsCacheStorageSize.

See allowCachingResults query option for more details.


Comparing SPARQL and Prolog

To give a comparison of Prolog and SPARQL when it comes to querying an AllegroGraph repository:

The Prolog and the SPARQL query engines are separate components with different performance characteristics:

Please see the Prolog tutorial and Prolog select documentation.


First class triples

AllegroGraph currently permits you to make assertions about triples via the tripleId magic property or Lisp function triple-id.

A tripleId value is special numerical value that can be projected from a query, but which does not interact well with other SPARQL features like FILTER or ORDER BY. Please contact us if you have questions or requests in this area.

AllegroGraph provides complete (according to the current version of the draft spec) support for RDF-star and SPARQL-star.


SPARQL dataset

The SPARQL dataset refers to the the set of IRIs that make up the default graph and named graphs in which triples are looked up. In principle the FROM and FROM NAMED clauses in a query specify the dataset. AllegroGraph offers options to customize and override this.

Dataset loading

In the Lisp client the handling of FROM and FROM NAMED clauses can be overridden by specifying a :load-function argument to run-sparql, or by setting *dataset-load-function*. This enables dynamically loading triples into the repository during query execution.

Default dataset handling

For queries that do not provide a FROM or FROM NAMED clause the defaultDatasetBehavior query option controls which triples are in the default graph and named graphs of the dataset. In the Lisp client this can also be configured via the run-sparql :default-dataset-behavior argument, defaulting to *default-dataset-behavior*.


Query execution options

AllegroGraph provides control over SPARQL query execution via various options. These options can be specified per query by including a special PREFIX line of the form:

PREFIX franzOption_optionName: <franz:optionValue> 

It is also possible option values to be used for all queries in the Server Configuration.

You can specify query options on the Query page in AGWebView. See the WebView document, in particular the New Query Page section.

The following query execution options are currently available:

allowCachingResults
query-option

If yes, query results may be cached on disk.

Query will be executed ignoring the limit and offset modifiers and its results will be stored on disk. All subsequent calls of the same query, potentially with different limit and offset values, will read results from the cache, apply the new limit and offset and return them.

Here is an example of using query results caching for paging SPARQL query results:

Let us assume a complex SPARQL query that takes a long time to run and returns a lot of results

SELECT <variables> { <patterns> } 

The results of the query can be split into pages of size 1000 using the combination of limit and offset modifiers:

SELECT <variables> { <patterns> } LIMIT 1000 OFFSET 0  
SELECT <variables> { <patterns> } LIMIT 1000 OFFSET 1000  
SELECT <variables> { <patterns> } LIMIT 1000 OFFSET <N * 1000> 

but this is very inefficient because the execution of the query for a given page takes about the same amount of time as the whole query. With query results caching, the first page query will take the usual amount of time

PREFIX franzOption_allowCachingResults: <franz:yes>  
SELECT <variables> { <patterns> } LIMIT 1000 OFFSET 0 

and will cache all the results of the query, so all subsequent calls to

PREFIX franzOption_allowCachingResults: <franz:yes>  
SELECT <variables> { <patterns> } LIMIT <N> OFFSET <M> 

for any values of limit and offset will read the results from cache and will be significantly faster.

Please note that query results caching is not always possible. If it is not possible, the query will be executed as usual, but a query warning will be returned, explaning why caching is impossible.

Query results cache is currently stored on disk and both the maximum number of cache entries and the maximum disk space allowed can be configured using the respective configuration options. The possible values are:

  • yes - turn the option on
  • no - turn the option off

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_allowCachingResults: <franz:no> 

The default value is: no

cancelQueryOnWarnings
query-option

If true, then warnings found during query parsing, planning and execution will cause a query to fail immediately rather than continuing.

Warnings include things like unknown variables in a ORDER BY clause or FILTER expression, constants in the query that cannot be in the store and so on.The possible values are:

  • yes - turn the option on
  • no - turn the option off

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_cancelQueryOnWarnings: <franz:no> 

The default value is: no

chunkProcessingAllowed
query-option

Controls whether to use Chunk at a Time (CaaT) processing.

It can be:

  • possibly - use CaaT for unordered queries with small limits and use the single-set approach otherwise. Note that this works best when solutions are found in the first several chunks processed which means that the query can finish quickly. If a large portion of the search space must be scanned, then the single-set approach can be faster.

  • yes - always use CaaT when possible (some query clauses like EXISTS filters and SPIN magic properties do not yet support CaaT).

  • no - always use the single set approach and never use CaaT.

The default value is possibly which means that AllegroGraph is optimizing for speed rather than space. The no option is focused on speed at the possible cost of higher memory use whereas the yes option is more constrained in memory use at the cost of slower queries.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_chunkProcessingAllowed: <franz:possibly> 
chunkProcessingMemory
query-option

Specifies the maximum amount of memory used by a single chunk.

Controls the size (in bytes) of the chunks used by the CaaT executor. This option takes precedence over the deprecated chunkProcessingSize option.

The minimium allowed value is 200M.

See the chunkProcessingAllowed option for additional query control.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_chunkProcessingMemory: <franz:4294967296> 

The default value is: 4,294,967,296

chunkProcessingSize
query-option

(Deprecated) Specifies the chunk processing size in rows

Deprecated in favor of the chunkProcessingMemory option.

Control the size (in rows of answers) of the chunks used by the CaaT executor. The higher the number, the larger the chunks processed will be which is both more efficient and more memory intensive. A typical value is 400000 or 1000000.

See the chunkProcessingAllowed option for additional control.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_chunkProcessingSize: <franz:400000> 

The default value is: 400,000

clauseReorderer
query-option

The strategy used to reorder triple patterns in a query.

This option controls how the triple patterns in a single Basic Graph Pattern (BGP) are reordered.

The available strategies will depend on the query engine being used but will always include identity which tells the query planner to not reorder the triple patterns of the BGPs. Another common choice is statistical which uses the statistics of the triple-store to try to reorder clauses most efficiently.

Note that other query planning algebraic manipulations may cause BGPs in your query to be merged and that reordering does not extend to larger query structures (like UNION or OPTIONAL).

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_clauseReorderer: <franz:statistical> 

The clause reorderer defaults to statistical

defaultAttributes
query-option

Specify the default attributes to assign to any triple created by a SPARQL update command.

The attributes must be specified in URL encoded JSON format. So the example below is using the URL encoded form of which is the URL encoded form of {"rank": "High" }.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_defaultAttributes: <franz:%7B%22rank%22%3A%20%22High%22%20%7D> 

No attributes

defaultDatasetBehavior
query-option

In a SPARQL query the FROM clause specifies the default graph of the dataset, and the FROM NAMED clause specifies the named graphs of the dataset (see Specifying RDF Datasets in the SPARQL 1.1 standard).

If a SPARQL query does not specifiy FROM or FROM NAMED then it is up to the implementation to choose a behaviour. AllegroGraph offers three possibilities. The table below indicates which triples are present in which part of the dataset for each possible behaviour:

                                        +--------------------------+  
             default dataset behaviour: |   all  |   rdf  | default|  
                                        + - - - -| - - - -| - - - -|  
dataset's default graph / named graphs: | DG  NG | DG  NG | DG  NG |  
                                        |        |        |        |  
       triple (s,p,o) in default graph: |  x     |  x     |  x     |  
       triple (s,p,o,g) in named graph: |  x  x  |     x  |        |   "x" means present  
                                        +--------+--------+--------+ 

The possible values are:

  • all - All triples are present in the dataset's default graph. Triples in a named graph are present in the dataset in that named graph. Triples in a named graph thus occur twice in the dataset: once in the default graph, and once in the named graph.
  • rdf - Triples in the default graph are present in the dataset's default graph. Triples in a named graph are present in the dataset in that named graph.
  • default - Triples in the default graph are present in the dataset's default graph. The dataset has no named graphs.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_defaultDatasetBehavior: <franz:all> 

The default value is: all

diskChunkRowCount
query-option

Specifies the number of solutions to keep in memory before writing temporary files.

This should be a number like 500000 or 100m. The larger the value, the more memory AllegroGraph will use during query processing. Smaller values can be more memory efficient but also can perform more slowly because the will be more I/O activity.

Note that this setting controls the memory used to hold completed solutions not the memory used to hold intermediate solutions. See the chunkProcessingMemory option for more details.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_diskChunkRowCount: <franz:500000> 

The default value is: 500,000

engine
query-option

The query engine to use.The possible values are:

  • mjqe - New merge-join based query engine for supported queries ~ (e.g. select); fallback to :sbqe for other queries.
  • sbqe - Old set-based query engine (with or without Chunk-at-a-Time) ~ for all queries

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_engine: <franz:sbqe> 

The default value is: sbqe

logBacktraceOnQueryFailure
query-option

When true, the query engine prints backtraces of any errors that happen during query execution.

Only applicable when queries are executed through HTTP endpoint.The possible values are:

  • yes - turn the option on
  • no - turn the option off

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_logBacktraceOnQueryFailure: <franz:no> 

The default value is: no

logLineLength
query-option

Controls the length of query log lines.

The logLineLength query option limits the maximum length of each line of the query log. This can make the log easier to read at the cost of removing some information. Use zero to print the entirety of every log message.

See the logQuery option for more details.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_logLineLength: <franz:512> 

The default value is: 512

logQuery
query-option

Controls whether or not query execution details are logged.

logQuery can be 'no', 'yes', or 'onFailure'.

The length of the log lines can be limited by using the logLineLength query option.

If logging is on, the query engine prints additional information to the AllegroGraph log file as it plans and executes a query. If logging is onFailure, then query log information is gathered but not emitted unless there is a query failure.

Logging on failure has a small cost especially when the amount of data logged is high (e.g., when chunkProcessingAllowed is turned on). We recommend setting the value to onFailure during development and then turning it to no for production.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_logQuery: <franz:no> 

The default value is: no

maximumSolutionsSize
query-option

Specifies an upper limit on the number of solutions that are allowed during query processing before a warning is logged.

Queries run best when the solution space is kept small. This warning is in an indication that a query is generating many intermediate results. This is a normal part of query processing but can indicate that a query should be optimized

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_maximumSolutionsSize: <franz:100k> 

The default is to warn when the intermediate solution space is larger than 100,000,000 solutions

maximumValuesCountForService
query-option

This option limits the number of VALUES that AllegroGraph will send to a SPARQL endpoint when executing a SERVICE clause. Sending partial results to the endpoint can help it answer the query more quickly but if the number of partial results is very large, the cost of data transfer can offset the help of supplying the data.

If the number of VALUES exceeds the limit, the the query will be sent to the endpoint with no VALUES supplied.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_maximumValuesCountForService: <franz:1048576> 

no limit

memoryExhaustionWarningPercentage
query-option

Specify how much system memory must be free for a query to continue.

If the query process is using more than this setting's percentage of total physical memory, then the query will be canceled. The default value is 90%.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_memoryExhaustionWarningPercentage: <franz:90.0> 

The default value is: 90.0

memoryLimit
query-option

Specifies the memory limit per query.

If a query tries to use more than this, it will be canceled.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_memoryLimit: <franz:8G> 

The default value will be 85% of the physical memory on the server

openaiApiKey
query-option

The OpenAI API key needed for the magic predicate and SPARQL functions in LLM namespace.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_openaiApiKey: <franz:sk-U01ABc2defGHIJKlmnOpQ3RstvVWxyZABcD4eFG5jiJKlmno> 

You must provide an openai API key to use LLM Magic Predicates.

presentationTimeZone
query-option

The timezone in which xsd:dateTimes and xsd:times are serialized.

For example, if presentationTimeZone is "-02:00", then "2013-10-01T15:21:23+03:00" is serialized as "2013-10-01T10:21:23-02:00". Zoneless xsd:datetimes and xsd:times are always presented without a timezone. This option has no effect on what is stored in the database. The allowed values are strings representing the timezone. The format of these strings is the same as in xsd:dateTimes. The special value "none" means that no conversion will take place.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_presentationTimeZone: <franz:-5:00> 

The default is set to none.

profileOutlineDepth
query-option

Controls the depth of the profile outline.

Only applicable when the outline format value is used for the profileOutputFormat option, ignored otherwise.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_profileOutlineDepth: <franz:100> 

The default value is: 100

profileOutputFormat
query-option

Controls the profile output format.

profileOutputFormat can be call-graph, flat or outline, which correspond to the prof:show-call-graph, prof:show-flat-profile and prof:show-outline-profile respectively.

See https://franz.com/support/documentation/current/doc/runtime-analyzer.htm for more information on call-graph, flat and outline options.

This option is only applicable for space and time profiling, and ignored if the value of profileQuery is perf. The possible values are:

  • call-graph - Use call graph profile output format
  • flat - Use flat profile output format
  • outline - Use outline profile output format (see profileOutlineDepth for depth configuration)

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_profileOutputFormat: <franz:flat> 

The default value is: flat

profilePerfEvents
query-option

List of perf events to use when profiling a query with perf tool.

Value should be a list of comma-separated words denoting perf events (see perf list for a list of pre-defined events). If the list is empty, perf will be run without --events argument, falling back to the default set of events.

See https://perf.wiki.kernel.org/index.php/Main_Page for more information on perf.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_profilePerfEvents: <franz:cache-misses,cache-references> 

Default value is an empty list (i.e. no --events argument).

profileQuery
query-option

Controls whether or not the profile information is collected during query execution.

profileQuery can be no, space, time or perf.

If profiling is enabled, the query profile will be written to the same place where the query log is written. If the value is perf but the tool is not available, a warning will be written to the query log and no information will be collected. The possible values are:

  • no - Do not collect profile information
  • space - Collect memory usage profile information
  • time - Collect CPU usage profile information
  • perf - Collect profile information using perf tool

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_profileQuery: <franz:no> 

The default value is: no

queryTimeout
query-option

Specifies a query timeout value in seconds.

Note that the timeout is not an interrupt; AllegroGraph checks for query timeout relatively infrequently so that a query can run for many seconds longer than the specified timeout. This is especially true for operations involving reasoning or non-triple-pattern based queries like free-text indexing or SNA path planning operators.

Setting the timeout to zero is the same as having no timeout.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_queryTimeout: <franz:30> 

The default is to have no query timeout. I.e., queries will run until complete.

reorderDuringExecution
query-option

Controls whether or not AllegroGraph interleaves query execution and triple-pattern selection.

If no, then AllegroGraph will perform all reordering during query planning. If yes, then AllegroGraph will defer reordering until query execution time. In many cases, the additional information available at execution time can enhance query performance.

Note that interleaving reordering is not always a win because performing all ordering at query planning time allows for the query engine to introduce joins which can sometimes enhance query performance.

This option may not be supported on all query engines. If specified yes for an engine where it is not supported, the option is silently ignored.

See the clauseReorderer option for additional information. The possible values are:

  • yes - turn the option on
  • no - turn the option off

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_reorderDuringExecution: <franz:no> 

The default value is: no

serpApiKey
query-option

The SERP API key needed for the SERP magic predicate in SPARQL.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_serpApiKey: <franz:ac34ac5984ac0094b15627d34b5859c76d17038d791c26e38f161e7938f167e9> 

You must provide an openai serpapi key to use SERP API. See llm.html

serviceTimeout
query-option

The number of seconds to wait before a remote query times out.

This will also have an effect on SPARQL Federated query (i.e., using the SERVICE clause).

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_serviceTimeout: <franz:120> 

The default value is: 120

slowQueryLogThreshold
query-option

Specifies query duration threshold in milliseconds that triggers slow query logging.

If a query's runtime exceeds the threshold, it will be logged either to the file specified by the slowQueryLogFile configuration setting or to agraph.log.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_slowQueryLogThreshold: <franz:1000> 

The default is to not log slow queries.

solrQueryLimit
query-option

Specifies the maximum number of results to return from a given SOLR query.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_solrQueryLimit: <franz:100> 

The default value is: 100

temporaryFilesystemSpaceLimit
query-option

Specifies the maximum amount of temporary file space that may be used by a query.

If a query tries to use more file space than this, it will be canceled.

Queries write intermediate results to the filesystem when they will not fit in memory. With a huge query it is possible for such temporary files to fill the filesystem. In order to prevent this, the temporaryFilesystemSpaceLimit query option may be set.

The minimum allowable value for this setting is 2-gigabytes.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_temporaryFilesystemSpaceLimit: <franz:1008692224> 

The default value is to use the minimum of 8-gigabytes and one quarter of the available filesystem space at the time the query begins.

trustEncodedDatatypesForRangeQueries
query-option

If yes, then range queries will not scan typed literal triples.

This means that only encoded triples will be considered. The only reason to set this option to no is if your triple-store contains typed literals that are not encoded (i.e., that are in the string-table) which could happen if you disabled AllegroGraph's datatype mapping.The possible values are:

  • yes - turn the option on
  • no - turn the option off

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_trustEncodedDatatypesForRangeQueries: <franz:yes> 

The default value is: yes

trustPredicateTypeMappingsForRangeQueries
query-option

If yes, then predicate type mappings will be used for range queries.

This means that any triples whose encoded data-type does not match their predicate mapping will be ignored. This could happen only if a predicate mapping was added or changed after triples had been added.The possible values are:

  • yes - turn the option on
  • no - turn the option off

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_trustPredicateTypeMappingsForRangeQueries: <franz:yes> 

The default value is: yes

unifyOwlSameAsDuplicates
query-option

If true, when reasoning is used, all occurences of resources in results that are declared to be owl:sameAs other resources will be replaced with a canonical member of its owl:sameAs equivalence class and deduplicated. For example, given the following data

@prefix owl: <http://www.w3.org/2002/07/owl#>.  
@prefix : <http://example.org/>.  
 
:Batman :livesIn :Gotham;  
        owl:sameAs :Bruce_Wayne. 

the following query would normally returns 2 different results

PREFIX : <http://example.org/>  
SELECT * { ?s :livesIn :Gotham }  
 
----------------  
| s            |  
================  
| :Bruce_Wayne |  
| :Batman      |  
---------------- 

but with this query option, it will return a single one

PREFIX franzOption_unifyOwlSameAsDuplicates: <franz:yes>  
PREFIX : <http://example.org/>  
SELECT * { ?s :livesIn :Gotham }  
 
-----------  
| s       |  
===========  
| :Batman |  
----------- 

Please note that the canonical instance is chosen using the shortest string representation criterion. In the example above, :Batman is the shortest-string instance of the given owl:sameAs class, so it is chosen as the canonical instance. If another instance with shorter string representation is added to the triple store, for example:

:Bruce owl:sameAs :Bruce_Wayne. 

the result will change because :Bruce is shorter than :Batman:

PREFIX franzOption_unifyOwlSameAsDuplicates: <franz:yes>  
PREFIX : <http://example.org/>  
SELECT * { ?s :livesIn :Gotham }  
 
----------  
| s      |  
==========  
| :Bruce |  
---------- 

The possible values are:

  • yes - turn the option on
  • no - turn the option off

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_unifyOwlSameAsDuplicates: <franz:no> 

The default value is: no

usePredicateConstrainedUpiTypeInformation
query-option

Use subject and object UPI type-codes to improve constraint inference

If yes, then the query engine will gather information about the subjects and objects associated with particular predicates. This can be used in constraint analysis and query transformations. As an example, suppose we have a query like:

?one ex:date ?date1 .  
?two ex:date ?date2 .  
filter( ?date1 > ?date2 ) 

If there is no predicate type-mapping, then the query engine can not make any assumptions about the range comparison. If there is a predicate type-mapping and trustPredicateTypeMappingsForRangeQueries is true, then the engine can know that the filter can be treated as a date comparison. If usePredicateConstrainedUpiTypeInformation is yes, then the query engine will check the triple-store to determine which UPI type-codes the subjects and objects associated with ex:date can take on. If the objects of ex:date only have, e.g., UPI type-code +rdf-date+, then the filter will be handled more efficiently.

The type-code information is cached but if the store is changing rapidly, then the cache will often be invalid and this computation will slightly add to the cost of queries.The possible values are:

  • yes - turn the option on
  • no - turn the option off

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_usePredicateConstrainedUpiTypeInformation: <franz:yes> 

The default value is: yes

usePredicateConstrainedXsdTypeInformation
query-option

Use typed-literal XSD types to improve constraint inference.

Similar to usePredicateConstrainedUpiTypeInformation but involves a scan of all typed-literals (which can be expensive). This is currently not cached!The possible values are:

  • yes - turn the option on
  • no - turn the option off

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_usePredicateConstrainedXsdTypeInformation: <franz:no> 

The default value is: no

userAttributes
query-option

Specify the user attributes to use while evaluating the query.

The attributes must be specified in URL encoded JSON format. So the example below is using the URL encoded form of {"access-level": "medium", "department": "hr"}.

Example of specifying the query option value via a PREFIX:

PREFIX franzOption_userAttributes: <franz:%7B%22access-level%22%3A%20%22medium%22%2C%20%22department%22%3A%20%22hr%22%7D> 

empty


Query execution warnings

SPARQL is relatively lax when it comes to accepting and evaluating queries. For example, this query is valid SPARQL but it is probably not what was intended, as ?var is not actually bound by the triple patterns in the rest of the query:

select ?var {  
  ?s ?p ?o .  
} 

A more typical example could be caused by a typo. For example, changing the case of a variable:

select ?subjectOne ?class {  
 ?subjectone rdf:type ?Class .  
} 

Neither ?subjectOne nor ?class will be bound since the query uses ?subjectone and ?Class. When AllegroGraph plans and executes a query, it will detect problems like the above and generate query warnings. The cancelQueryOnWarnings query option can be used to stop execution at the first warning.

The following is a list of the currently defined warnings:

warn-bgp-cross-product
query-warning

This warning indicates that one or more basic graph patterns (BGPs) in the query create a cross product. I.e., there are patterns in the query that have disjoint sets of variables which will cause the SPARQL engine to find all possible matches between the sets which can lead to very large solution sets. For example:

SELECT * {  
  ?a ?b ?c .  
  ?d ?e ?f .  
} 

Since the first triple-pattern and the second triple-pattern share no variables, the above query will find all possible combinations of each pair of triples in the underlying repository.

warn-dataset-graph-not-found
query-warning

This warning is signaled when a query specifies a DATASET and one or more graphs in its FROM or FROM NAMED portions are not in the repository.

For example, if 'http://example#22` is not in the repository, then these queries will signal the warning:

SELECT *  
FROM <http://example#1>  
FROM <http://example#22> {  
  ?s ?p ?o .  
}  
 
SELECT *  
FROM NAMED <http://example#22> {  
  ?s ?p ?o .  
} 
warn-differing-limits
query-warning
This warning is signaled when the query string contains a LIMIT and a limit is also specified in the call to SPARQL (e.g., from the parameters to run-sparql or in the HTTP request), and these two values do not match. When this happens, the smaller of the two values is used.
warn-differing-offsets
query-warning
This warning is signaled when the query string contains an OFFSET and an offset is also specified in the call to SPARQL (e.g., from the parameters to run-sparql or in the HTTP request). In this case, the sum of the two offsets will be used.
warn-empty-bind-clause
query-warning

This warning is signaled when a SPARQL query uses BIND in such a way that no solutions are available to be processed. For example, the BIND in this query appears before anything else so there is nothing for it to do and ?X will always be unbound:

SELECT * {  
  BIND(?object as ?X)  
  ?s ?p ?object .  
} 
warn-graph-not-in-dataset
query-warning

This warning is signaled when a query uses a DATASET and the graph of a GRAPH clause is not in its FROM NAMED portion. For example

SELECT *  
FROM NAMED <http://graph1>  
FROM NAMED <http://graph2> {  
  GRAPH <http://example#22> {  
    ?s ?p ?o  
  }  
} 
warn-implicit-sample-introduced
query-warning
warn-impossible-atomic-constraint
query-warning
This warning occurs when a FILTER expression can be determined to be impossible at plan time. For example, FILTER ( false ) can never succeed.
warn-impossible-freetext-query
query-warning
This warning is signaled for queries that use AllegroGraph's freetext search Magic Properties when it can be determined that such a query cannot succeed. For example, if the search expression contains only stop words (which are not indexed), then the expression cannot match.
warn-impossible-in-constraint
query-warning
This warning is signaled when an IN expression has an empty list. For example FILTER( ?x IN ( ) ).
warn-impossible-language-literal-constraint
query-warning

This warning indicates that a FILTER expression puts invalid constraints on the language of a variable. For example:

SELECT * { ?s ?p ?o . FILTER( LANG(?o) = 'es' && LANG(?o) = 'en') } 

cannot succeed because the LANG(?o) cannot be both 'es' and 'en'.

warn-impossible-membership-constraint
query-warning

This warning is signaled when a query has a filter with multiple alternatives, and it is known at plan time that none of the alternatives will succeed.

Alternatives can occur in various forms:

{ ?s (<ex:foo> | <ex:bar>) ?o }      -- matches <ex:foo> or <ex:bar>  
FILTER(?var in (<ex:foo>, <ex:bar>)) -- matches <ex:foo> or <ex:bar>  
FILTER(str(?var) = 'ex:foo')         -- matches 'ex:foo' or <ex:foo> 
warn-impossible-range-constraint
query-warning
This warning is signaled when a range constraint cannot succeed because the set of possible values is empty. For example, this FILTER can never succeed: FILTER( ?o > 3 && ?o < 0 ).
warn-impossible-str-literal-constraint
query-warning

This warning indicates that a FILTER expression cannot succeed because it is comparing a string with a non-string. E.g.,

SELECT * { ?s ?p ?o . FILTER( STR(?o) = 45 ) } 
warn-invalid-triples-generated
query-warning

This warning is signaled at query time when a CONSTRUCT or UPDATE template generates invalid RDF triples. For example, this query will emit no triples because ?s takes on only literal bindings and these are not valid in the subject position:

CONSTRUCT {  
  ?s a example:Car .  
} WHERE {  
  VALUES ?s { 'mazda 3' 'ford pinto' 'bmw 300i' }  
} 
warn-invalid-triples-in-template
query-warning
This warning is signaled when a CONSTRUCT or UPDATE template has triple patterns in it that are not valid RDF. For example, a triple pattern may not have a literal in the subject position.
warn-join-cross-product
query-warning

This warning indicates that the query algebra contains one or more cross products. I.e., there are portions of the algebra that have disjoint sets of variables which will cause the SPARQL engine to find all possible matches between the sets which can lead to very large solution sets. For example:

SELECT * {  
  ?a a ?type .  
  VALUES ?FOO { <ex://a> <ex://b> }  
} 

Since the triple-pattern and the VALUES clause share no variables, the above query will find all possible combinations from the two sets.

warn-limit-in-ask
query-warning
This warning is signaled when an ASK query specifies a LIMIT different from 1.
warn-literal-graph-clause
query-warning

This warning is signaled if a GRAPH clause specifies a literal for the GRAPH. For example:

SELECT * {  
  GRAPH ?g {  
    ?s ?p ?o .  
  }  
  BIND( 'literal' as ?g )  
} 
warn-literal-variable-required
query-warning
This warning is signaled when a FILTER expression cannot succeed because a variable in it is known to be a resource or blank node and the expression requires it to be a literal. For example ?s ?p ?o . FILTER( ?s = 34 ) must fail because ?s is bound to the subject of a triple and subjects cannot be literals.
warn-offset-in-ask
query-warning
This warning is signaled when an ASK query specifies an OFFSET.
warn-order-by-in-ask
query-warning
This warning is signaled when an ASK query specifies ORDER BY.
warn-query-results-limited
query-warning
Warn the user that the LIMIT and OFFSET are restricted for export prevention.
warn-range-constraint-cannot-be-satisfied
query-warning
This warning is signaled when a range constraint cannot succeed because AllegroGraph has determined that the repository does not contain any matching triples.
warn-range-filter-predicate-mapping-mismatch
query-warning

This warning is signaled when a predicate has a type mapping which does not match the datatype of the values used in a range FILTER. For example, if <http://example#age> has a mapping to xsd:byte, then this query will signal the warning:

SELECT * {  
  ?s <http://example#age> ?age .  
  FILTER( ?age > '2001-10-15'^^xsd:date )  
} 

because the FILTER cannot succeed.

warn-repeated-projected-variables-ignored
query-warning
warn-results-caching-impossible
query-warning
This warning occurs when a SPARQL query requests results caching, but it cannot be used for some reason (specified).
warn-service-variable-is-unbound
query-warning
This warning indicates that the query contains a SERVICE clause with a variable HOST and the HOST variable is not bound by the rest of the query.
warn-sparql-type-errors
query-warning
This warning is signaled when SPARQL type errors are encountered during query evaluation. Examples of type errors include trying to add a number to a date, trying to compare a string with a time, etc.
warn-unknown-constants
query-warning

This warning is signaled when a query contains a constant value (e.g. IRI or literal value) that is not present in the repository. For example if is not used as predicate:

SELECT * { ?s <ex:pred> ?o }  -- gives: "No such predicate <ex:pred>" 
warn-unknown-variables
query-warning

This warning is signaled when it can be determined that a variable used in an expression is not bound anywhere in the query. Examples include:

SELECT * { ?s ?p ?o . } ORDER BY ?missing  
 
SELECT * { ?s ?p ?o . FILTER( ?missing > 5 )  
 
CONSTRUCT { ?missing ?p ?o } WHERE { ?s ?p ?o } 

and so on.

warn-unused-bind-variable
query-warning
This warning is signaled when a variable in a BIND expression is not used elsewhere in the query.


Standard SPARQL functions

AllegroGraph supports the standard functions specified by SPARQL 1.1, several XPath constructor functions, XPath mathematical functions, and a number of custom functions designed to help with using AllegroGraph's extensions such as Geospatial and Social Network Analysis.

XPath Constructor Functions

AllegroGraph supports the standard SPARQL casting operations. For details, refer to the SPARQL reference for more details.

Functions on Dates and Times

Hash Functions

Functions on Numerics

SPARQL Operators

These functions are described in detail in the Operator Mappings section of the W3C SPARQL reference.

Functions on RDF terms

Functional Forms

Functions on Strings

Supported XPath functions

AllegroGraph also supports several XPath functions:

Functions on Dates and Times

XPath Mathematical Functions

For additional details on the XPath mathematical functions, see https://www.w3.org/TR/xpath-functions-3/.

Miscellaneous functions

Functions on Numerics

Functions on Strings

AllegroGraph extensions

AllegroGraph supports several functions above and beyond those defined by the SPARQL standard. The additional functions are named by URI (or prefixed name) and can appear anywhere in a SPARQL expression (e.g., in BIND, FILTER or ORDER BY).

nD Geospatial

These functions are useful in working with AllegroGraph nD-geospatial facilities:

2D Geospatial

These functions are useful in working with the older AllegroGraph 2D Geospatial facilities:

2D Geospatial (deprecated)

These deprecated 2D Geospatial functions are still supported but you should change your queries to use the above functions instead.

Social Networking (SNA) Related functions

These functions can help when using AllegroGraph's SNA extensions.

Large Language Models (LLM) LLMagic related functions

These functions are useful in queries using LLM. Please see the LLM examples file for examples of these functions being used. Search, e.g., for llm:response in that file. There are also LLM SPARQL magic properties, see Large Language Models for more information

Miscellaneous other functions

These functions help connect SPARQL to various other AllegroGraph features.


Defining SPARQL extension functions

SPARQL allows for query engines to associate extension functions with URIs, and call them from within queries.

You can define your own URI functions through defurifun, or associate existing functions with a URI through associate-function-with-uri. defurifun does some manipulation of the arguments, so you should use it whenever possible.

associate-function-with-uri function  uri  &key  cache-now-p  db
function
Assert a mapping between uri, which is a string or a valid part, and the provided function, which is a symbol or a function. If cache-now-p, and function is a symbol, its function binding is stored instead of the symbol itself.
print-function-uri-mappings &key  stream  db
function
Print all mappings between URIs and functions to stream *standard-output* by default).
defurifun name  uri  args  &body  body
macro
Define a new function, name, and associate it with uri as with associate-function-with-uri. args is not evaluated, exactly as with defun.

Here's an example: a function that will do an HTTP HEAD request against the provided URL, returning the HTTP status code as an integer literal, or 0 if there's a problem.

(The built-in functions are quite robust, so a Lisp integer will be treated as an RDF literal with data type xsd:integer.)

(defurifun ex-head-request !<http://example.com/fn/head> (uri)  
  (or  
    (when uri  
      (ignore-errors  
        (format t "~&Performing HTTP HEAD request on <~A>...~%"  
                  (upi->value uri))  
        (second  
          (multiple-value-list  
            (net.aserve.client:do-http-request (upi->value uri)  
                                               :method :head)))))  
    0)) 

You can use this function in a query exactly as you would a built-in function.

Using this data as an example:

<http://ex.com/a> <http://ex.com/foo> "200"^^<http://www.w3.org/2001/XMLSchema#integer> . 

we can run a query like so:

sparql(54): (run-sparql  
"  
PREFIX f: <http://example.com/fn/>  
SELECT ?x {  
  ?x <http://ex.com/foo> ?y .  
  FILTER ( ?y = f:head("http://franz.com\") )  
}"  
  :results-format :count) 

which produces this output:

Performing HTTP HEAD request on <http://franz.com>...  
1  
:select  
(?x) 

… we know, then, that franz.com is returning a 200 status code.

Note that these filter functions can be called an arbitrary number of times during the execution of a query. It's not a good idea to actually perform expensive operations like HTTP requests in your queries.


Footnotes