Changes

This [[Module Library|XQuery Module]] provides functions for evaluating XQuery expressions in separate threadsorganizing scheduled, queued, running and cached jobs. Query execution Jobs can both be parallelized and postponed to be executed asynchronouslycommands, queries, client or HTTP requests.

All functions in this module are assigned to the <code><nowiki>http://basex.org/modules/asyncjobs</nowiki></code> namespace, which is statically bound to the {{Code|asyncjobs}} prefix. Errors will be bound to the same prefix.

=Parallelized ExecutionServices=

Parallel query execution is recommendable if you have various calls that require a lot of time, but cannot A job can be sped up registered as ''service'' by rewriting supplying the code. This is e. g. the case if external URLs are called. If you are parallelizing local data reads (such as accessing a database), your single-threaded query will usually be faster, because parallelized access {{Code|service}} option to disk data will often lead to randomized access patterns, which can hardly be optimized by your HD or SSD.{{Function|Jobs|jobs:eval}}:

<syntaxhighlight lang="xquery">(: register job as service; will be run every day at 1 am :)jobs:eval('db:drop("tmp")', (), map { 'id':'cleanup', 'start':'01:00:00', 'interval':'P1D', 'service': true() }), (: list registered services :)jobs:services(),(: result: <job base-uri=async"..." id="cleanup" interval="P1D" start="01:00:00">db:drop("tmp")</job> :) (: unregister job :)jobs:stop('cleanup', map { 'service': true() })</syntaxhighlight> '''Some more notes:''' * All job services will be scheduled for evaluation when the BaseX server or BaseX HTTP server is started.* If a job service is outdated (e.g. because a supplied end time has been exceeded), it will be removed from the jobs file at startup time.* The job definitions are stored in a {{Code|jobs.xml}} file in the database directory. It can also be edited manually. =Basic Functions= ==jobs:fork-joincurrent==

|{{Func|asyncjobs:fork-joincurrent|$functions as function(*)*|item()*}}<br/ >{{Func|async:fork-join|$functions as function(*)*, $options as map(xs:string, xs:string)|item()*}}<br/ >

|'''Summary'''|This function executes Returns the supplied (non-updating) functions in parallelid of the current job. The following {{Code|$options}} are available ==jobs:list==* <code>threads</code>{| width='100%'|-| width='120' | '''Signatures'''|{{Func|jobs: maximum number of parallel threads (defaultlist||xs: available number of cores)string*}}* <code>thread|-size</code>: number | '''Summary'''|Returns the ids of functions to be evaluated by each thread (default: <code>1</code>)all jobs that are currently registered. The list includes scheduled, queued, running, stopped, and finished jobs with cached results.

|<code>jobs:list()</code> returns the same job id as {{Function|Jobs|jobs:current}} if no other job is registered.|} * The following function sleeps in parallel; it will be finished in 1 second if your system has at least 2 cores==jobs:list-details==<pre class{| width='brush:xquery100%'>async:fork|-join( for $i in 1 to 2| width='120' | '''Signatures''' return function|{{Func|jobs:list-details||element(job) *}}<br/>{ prof{Func|jobs:list-details|$id as xs:sleepstring|element(1000job) *}}|-| '''Summary'''|Returns information on all jobs that are currently registered, or on a job with the specified {{Code|$id}} (or an empty sequence if this job is not found). The list includes scheduled, queued, running jobs, and cached jobs. A string representation of the job, or its URI, will be returned as value. The returned elements have additional attributes:* <code>id</precode>: job id* In <code>type</code>: type of the following job (command, query, up to two URLs will be requested in parallel:REST, RESTXQ, etc.)* <code>state<pre class='brush:xquery'/code>let $funcs := for $segment in 1 to 4 let $url := 'httpcurrent state of the job:<code>scheduled</code>, <code>queued</code>, <code>running</url.comcode>, <code>cached</code>* <code>user</path' || $segmentcode>: user who started the job return function() { http* <code>duration</code>:send-request(evaluation time (included if a job is running or if the result was cached), $url) }return async* <code>start</code>:fork-joinnext start of job ($funcs, map { 'threads': 2 }included if a job will be executed repeatedly)* <code>time</precode>: time when job was registered

|'''ErrorsExamples'''|{{Error|unexpected|#Errors}} an unexpected error occurred while <code>jobs:list-details()</code> returns information on the currently running job and possibly others:<syntaxhighlight lang="xml"><job id="job1" type="XQuery" state="running a query or function in a separate thread" user="admin" duration="PT0.001S"> XQUERY jobs:list-details()<br/job>{{Error|out-of-range|#Errors}} a supplied option is out of range.<br/syntaxhighlight>

=Asynchronous Execution=jobs:finished==

Asynchronous query execution is recommendable {| width='100%'|-| width='120' | '''Signatures'''|{{Func|jobs:finished|$id as xs:string|xs:boolean}}|-| '''Summary'''|Indicates if a client does notthe evaluation of an already running job with the specified {{Code|$id}} has finished. As the ids of finished jobs will usually be discarded, or cannotunless caching is enabled, wait until a request is fully processed. This is e. g. the case with web browsers, which function will usually cancel a request after a specific timeoutalso return <code>true</code> for unknown jobs. In such cases* <code>false</code> indicates that the job id is scheduled, you can use asynchronous execution to trigger another server-side processqueued, which or currently running.* <code>true</code> will start be returned if the time-consuming processjob has either finished, and fetch or if the result later on as soon as it id is availableunknown (because the ids of all finished jobs will not be cached).|}

==asyncjobs:evalservices==

|{{Func|asyncjobs:evalservices|$query as xs:string|xs:string}}<br />{{Func|async:eval|$query as xs:string, $bindings as mapelement(*job)|xs:string}}<br />{{Func|async:eval|$query as xs:string, $bindings as map(*), $options as map(xs:string, xs:string)|xs:string}}<br />

|Prepares the supplied {{Code|$query}} string for asynchronous execution and returns Returns a query id. The query will be queued list of all jobs that have been persistently registered as described in the article on [[Transaction Management]], and the result will be cached in main-memory until it is fetched via [[#async:resultServices|async:resultServices]], or until {{Option|ASYNCTIMEOUT}} is exceeded. Queries may be updating.<br/>Variables and context items can be declared via {{Code|$bindings}} (see [[XQuery Module#xquery:eval|xquery:eval]] for more details). The {{Code|$options}} parameter contains evaluation options:* {{Code|cache}}: indicates if the query result will be cached or ignored (default: <code>true</code>). If the query result will not be cached, the query id will immediately be discarded after query execution, too.* {{Code|base-uri}}: set [https://www.w3.org/TR/xquery-31/#dt-static-base-uri base-uri property] for the query. This URI will be used when resolving relative URIs by functions such as {{Code|fn:doc}} (default: ''empty string'').

| '''ExamplesErrors'''|* {{CodeError|async:eval("1+3")}} returns a query id, e.g. {{Codeservices|Query-abc#Errors}}Registered services cannot be parsed. The result can be retrieved via a second query in the same BaseX context: {{Code|async:result("Query-abc")}}<br />* The following [[RESTXQ]] function will return the id of the query thread, which evaluates the query that has been specified in the body of a POST request:<pre class='brush:xquery'>declare %rest:POST("{$query}") %rest:path('/eval') function local:eval($query) { async:eval($query)};</pre>

=Execution=async There are cases in which a client does not, or cannot, wait until a request is fully processed. The client may be a browser, which sends an HTTP request to the server in order to start another time-consuming query job. The functions in this section allow you to register a new query job from a running query. Jobs can be executed immediately (i.e., as soon as the [[Transaction Management#Concurrency Control|Concurrency Control]] allows it) or scheduled for repeated execution. Each registered job gets a job id, and the id can be used to retrieve a query result, stop a job, or wait for its termination. ==jobs:resulteval== {{Mark|Updated with Version 9.5}}: Integers added as valid start and end times.

|{{Func|asyncjobs:resulteval|$id query as xs:anyAtomicItem|xs:string}}<br />{{Func|jobs:eval|$query as xs:anyAtomicItem, $bindings as map(*)?|xs:string}}<br />{{Func|jobs:eval|item$query as xs:anyAtomicItem, $bindings as map(*)?, $options as map(*)?|xs:string}}<br />

|Returns Schedules the result evaluation of an asynchronously executed the supplied {{Code|$query}} and returns a query id. The query with will be queued, and the specified result will optionally be cached. Queries can be updating. The query can be a URI or a string, and variables and context items can be declared via {{Code|$bindings}} (see [[XQuery Module#xquery:eval|xquery:eval]] for more details). The following {{Code|$idoptions}} can be supplied:* {{Code|cache}}: indicates if the query result will be cached or ignored (default: <code>false</code>):* Results * The result will be cached in main-memory until it is fetched via [[#jobs:result|jobs:result]], or until {{Option|CACHETIMEOUT}} is exceeded.** If the query raises an error, it will be cached and returned instead.* {{Code|start}}: a dayTimeDuration, time, dateTime or integer can only be retrieved oncespecified to delay the execution of the query:** If a dayTimeDuration is specified, the query will be queued after the specified duration has passed. Examples for valid values are: <code>P1D</code> (1 day), <code>PT5M</code> (5 minutes), <code>PT0.1S</code> (100 ms). An error will be raised if a negative value is specified.** If a dateTime is specified, the query will be executed at this date. Examples for valid values are: <code>2018-12-31T23:59:59</code> (New Year's Eve 2018, close to midnight). An error will be raised if the specified time lies in the past.** If a time is specified, the query will be executed at this time of the day. Examples for valid times are: <code>02:00:00</code> (2am local time), <code>12:00:00Z</code> (noon, UTC). After retrievalIf the time lies in the past, the cached result query will be discardedexecuted the next day.* * An integer will be interpreted as minutes. If the specified number exceeds the minutes of the current hour, the query will be executed one hour later.* {{Code|interval}}: a dayTimeDuration string can be specified to execute the query periodically. An error is raised an errorif the specified interval is less than one second (<code>PT1S</code>). If the next scheduled call is due, and if a query with the same id is still running, it will be skipped.* {{Code|end}}: scheduling can be stopped after a given time or duration. The string format is the same as for {{Code|start}}. An error is raised if the resulting end time is smaller than the start time.* {{Code|base-uri}}: sets the [https://www.w3.org/TR/xquery-31/#dt-static-base-uri base-uri property] for the query. This URI will be raised insteadused when resolving relative URIs, such as with {{Code|fn:doc}}.* {{Code|id}}: sets a custom job id. The id must not start with the standard <code>job</code> prefix, and it can only be assigned if no job with the same name exists.* {{Code|service}}: additionally registers the job as [[#Services|service]]. Registered services must have no variable bindings.* {{Code|log}}: writes the specified string to the [[Logging|database logs]]. Two log entries are stored, one at the beginning and another one after the execution of the job.

|{{Error|overflow|#Errors}} Query execution is rejected, because too many jobs are queued or being executed. {{Option|CACHETIMEOUT}} can be decreased if the default setting is-runningtoo restrictive.<br/>{{Error|range|#Errors}} the query A specified time or duration is still runningout of range.<br/>{{Error|unknownid|#Errors}} the supplied query The specified id is unknown: The query result may invalid or has already have been retrieved, or query execution may have been stoppedassigned.<br/>{{Error|options|#Errors}} The specified options are conflicting.

* Cache query result. The returned id can be used to pick up the result with [[#jobs:result|jobs:result]]:<syntaxhighlight lang="xquery">jobs:eval("1+3", (), map { 'cache': true() })</syntaxhighlight>* A happy birthday mail will be sent at the given date:<syntaxhighlight lang="xquery">jobs:eval("import module namespace mail='mail'; mail:send('Happy birthday!')", (), map { 'start': '2018-09-01T06:00:00' })}}</syntaxhighlight>* The following [[RESTXQ]] functions can be called to execute a query at 2 am every day. An id will be returned by the first function will either return , which can be used to stop the scheduler via the result of a previously started query or an errorsecond function:<pre classsyntaxhighlight lang="xquery">declare %rest:POST("{$query}") %rest:path('/start-scheduling') function local:start($query) { jobs:eval($query, (), map { 'start'brush:xquery'>02:00:00', 'interval': 'P1D' })};declare %rest:path('/resultstop-scheduling/{$id}') function local:resultstop($id) { asyncjobs:resultstop($id)

</presyntaxhighlight>* The following query demonstrates how Query execution is scheduled for every second, and for 10 seconds in total. As the results of an asynchronously executed query can itself will take 1.5 seconds, it will only be returned in a single query. Please remember that this is not the common way how these functions are used in practiceexecuted every second time:<pre classsyntaxhighlight lang="xquery">jobs:eval("prof:sleep(1500)", (), map { 'interval'brush:xquery'PT1S', 'end': 'PT10S' })</syntaxhighlight>let $* The query in the specified file will be evaluated once:<syntaxhighlight lang= async"xquery">jobs:eval(xs:anyURI('(1 to 10000000)[cleanup. = 1]xq'))return (</syntaxhighlight> hof* The following expression, if stored in a file, will be evaluated every 5 seconds:until( function($result) { async<syntaxhighlight lang="xquery">jobs:finishedeval($query) }, function static-base-uri($curr) { prof:sleep(10) }, () )map { }, asyncmap { 'start':result($query)'PT5S' }

</presyntaxhighlight>

==asyncjobs:finishedresult==

|{{Func|asyncjobs:finishedresult|$id as xs:string|xs:booleanitem()*}}

|Indicates if Returns the evaluation cached result of a query job with the specified query job {{Code|$id}} has finished:* Results can only be retrieved once. After retrieval, the cached result will be dropped. * If <code>false</code> is returnedthe original job has raised an error, the query is still running. An cached error will be raised if the query result was not cached or has already been retrievedinstead.

|{{Error|running|#Errors}} the job is still running.<br/>{{Error|unknown|#Errors}} the supplied query id is unknown: The query id is unknown, or the result may has already have been retrieved, .<br/>|-| '''Examples'''|* The following [[RESTXQ]] function will either return the result of a previously started job or raise an error:<syntaxhighlight lang="xquery">declare %rest:path('/result/{$id}') function local:result($id) { jobs:result($id)};</syntaxhighlight>* The following query demonstrates how the results of an executed query execution may have been stoppedcan be returned within the same query (see below why you should avoid this pattern in practice):<syntaxhighlight lang="xquery">let $query := jobs:eval('(1 to 10000000)[.= 1]', map { }, map { 'cache': true() })return ( jobs:wait($query), jobs:result($query))<br/syntaxhighlight>Queries of this kind can cause deadlocks! If the original query and the new query perform updates on the same database, the second query will only be run after the first one has been executed, and the first query will wait for the second query forever. You should resort to [[XQuery Module#xquery:fork-join|xquery:fork-join]] if you want to have full control on parallel query execution.

==asyncjobs:stop==

|{{Func|asyncjobs:stop|$id as xs:string|empty-sequence()}}

|Cancels Triggers the execution cancelation of a query job with the specified query {{Code|$id}}, or drops the cached result of a query result if , or cancels a scheduled job. Unknown ids are ignored. All jobs are gracefully stopped; it is up to the process to decide when it has already been executedis safe to shut down. The following {{Code|$options}} can be supplied:* {{Code|service}}: additionally removes the job from the [[#Services|job services]] list.

| '''ErrorsExamples'''|{{Error|unknown|#Errors}} the supplied query id is unknown<code>jobs:list()[. != jobs:current()] ! jobs: The query result may already have been retrieved, or query execution may have been stoppedstop(.)<br/code>stops and discards all jobs except for the current one.

==asyncjobs:idswait==

|{{Func|asyncjobs:ids|wait|$id as xs:string*|empty-sequence()}}

|Returns Waits for the ids completion of all queries that are either being executed asynchronouslya job with the specified {{Code|$id}}:* The function will terminate immediately if the job id is unknown. This is the case if a future job has not been queued yet, or that have if the id has already been discarded after job evaluation.* If the function is called with the id of a queued job, or repeatedly executed job, it may stall and the results of which have been cachednever terminate.

| '''ExamplesErrors'''|* <code>async:ids() ! async:stop({{Error|self|#Errors}} The current job is addressed.)<br/code> stops and invalidates all asynchronous queries and results.

|{{Code|unexpectedoptions}}| An unexpected error occurred while running a query or function in a separate threadThe specified options are conflicting.

|{{Code|outid}}| The specified id is invalid or has already been assigned.|-of|{{Code|overflow}}| Too many queries or query results are queued.|-|{{Code|range}}| The supplied option A specified time or duration is out of range.|-|{{Code|running}}| A query is still running.|-|{{Code|self}}| The current job cannot be addressed.|-|{{Code|service}}| Registered services cannot be parsed, added or removed.