Do what you did with the iterator but in parallel in multiple threads,
if you can break up the query to produce independent distinct 'streams" that's
best, if not, do 1 pass using the query but wrap
the whole thing in xdmp:estimate()
Take that as a rough approximation and then spawn N Threads (say 10)
and for each request a distinct range -
say N = estimate()
T = # of threads
M = T/N
each thread : ID = 0.. T-1
( query ...) [ ID*M + 1 to ID*M+M ]
then finish it with a query of [ position() > M*T ]
From: [email protected]
[mailto:[email protected]] On Behalf Of Mark Shanks
Sent: Tuesday, October 11, 2016 4:27 PM
To: MarkLogic Developer Discussion <[email protected]>
Subject: Re: [MarkLogic Dev General] How to pull data out of marklogic quickly?
I previously tried not wrapping the output in the <data> element. Running the
following code:
for $x in cts:search(fn:doc(),cts:and-query((
cts:element-value-query(xs:QName('Department'), 'Sales'),
cts:element-range-query(xs:QName('Date'), '>', xs:date('2015-01-01')),
cts:element-range-query(xs:QName('Date'), '<', xs:date('2015-01-03')),
cts:not-query(cts:element-value-query(xs:QName('Date'), 'NULL')) )),
'unfiltered' , 0.0) )), 'unfiltered' , 0.0) return
fn:concat($x//Department,'|',$x//Total,'|',$x//Location' ')
It would return the required documents in the console. However, when I ran the
same code using the rest api and java using:
theCall.xquery(query);
out.println(theCall.evalAs(String.class));
It would print out only a single document. I then tried the iterator instead:
theCall.xquery(query);
EvalResultIterator result = theCall.eval();
while (result.hasNext()) {
out.println(result.next().getString());
}
This did retrieve all of the documents, but was benchmarked as slower -
presumably because you have so much back and forth between java and the server.
Is there another way to get the results into java that does not involve the
iterator but returns all of the documents?
________________________________
From:
[email protected]<mailto:[email protected]>
<[email protected]<mailto:[email protected]>>
on behalf of Geert Josten
<[email protected]<mailto:[email protected]>>
Sent: Tuesday, 11 October 2016 6:15:57 PM
To: MarkLogic Developer Discussion
Subject: Re: [MarkLogic Dev General] How to pull data out of marklogic quickly?
Hi Mark,
The best way to tackle this would be to parallelize output. Have 10 or more
worker threads consume parts of the total (how many might depend on your
cluster size, and the total amount of records you need to produce), and make
each write a CSV on its own.
The cts:search is a good starting point, but if you want to emit CSV anyhow,
then don't wrap the results of cts:search in a <data> element. Instead let each
doc found from cts:search return one or more line-strings, which you don't join
either. MarkLogic will insert line-ends between such strings automatically, and
this way it will allow for streaming.
Doing it right, one worker should be able to produce a 1 mln record csv file in
a few minute on an average laptop.
At this point, I would worry less about using $x//Department, but assuming $x
holds the document node, you could write $x/Record/Department. That would
indeed be a little quicker.
Not sure if Corb(2) can produce CSV, and if it would leverage parallelism in
the same way as I meant, but it could be worth taking a look at cluster-based
tools like Hadoop. Apache Camel might allow parallel processing too..
Cheers,
Geert
From:
<[email protected]<mailto:[email protected]>>
on behalf of Mark Shanks
<[email protected]<mailto:[email protected]>>
Reply-To: MarkLogic Developer Discussion
<[email protected]<mailto:[email protected]>>
Date: Tuesday, October 11, 2016 at 12:27 AM
To: MarkLogic Developer Discussion
<[email protected]<mailto:[email protected]>>
Subject: Re: [MarkLogic Dev General] How to pull data out of marklogic quickly?
MLCP isn't an option as it doesn't provide text-delimited output.
Text-delimited is a useful format as it allows the data to be pulled into
practically any other application and with little overhead, unlike xml/json.
Another problem with xml/json output other than compatibility is the file size.
When a text-delimited file can be over 30GB with the data we are working with,
the same data in xml or json becomes absolutely gigantic.
What you say about $x//Department makes sense. If the data is in Marklogic as:
<Record>
<Department>Sales</Department>
</Record>
What is the best way to get the Department value (i.e., fastest)?
________________________________
From:
[email protected]<mailto:[email protected]>
<[email protected]<mailto:[email protected]>>
on behalf of Sekhon, Navdeep
<[email protected]<mailto:[email protected]>>
Sent: Tuesday, 11 October 2016 6:22:44 AM
To: [email protected]<mailto:[email protected]>
Subject: Re: [MarkLogic Dev General] How to pull data out of marklogic quickly?
Have you looked into using MLCP? https://developer.marklogic.com/products/mlcp
You can provide your cts query as an option to mlcp, get the documents out of
ml and do your processing.
Also, this $x//Department is an expensive operation. You should instead give
the exact xpath.
Regards,
ns/.
-----Original Message-----
From:
[email protected]<mailto:[email protected]>
[mailto:[email protected]] On Behalf Of
[email protected]<mailto:[email protected]>
Sent: Monday, October 10, 2016 3:00 PM
To: [email protected]<mailto:[email protected]>
Subject: General Digest, Vol 148, Issue 13
Send General mailing list submissions to
[email protected]<mailto:[email protected]>
To subscribe or unsubscribe via the World Wide Web, visit
http://developer.marklogic.com/mailman/listinfo/general
or, via email, send a message with subject or body 'help' to
[email protected]<mailto:[email protected]>
You can reach the person managing the list at
[email protected]<mailto:[email protected]>
When replying, please edit your Subject line so it is more specific than "Re:
Contents of General digest..."
Today's Topics:
1. How to pull data out of marklogic quickly? (Mark Shanks)
----------------------------------------------------------------------
Message: 1
Date: Mon, 10 Oct 2016 18:43:52 +0000
From: Mark Shanks <[email protected]<mailto:[email protected]>>
Subject: [MarkLogic Dev General] How to pull data out of marklogic
quickly?
To: "[email protected]<mailto:[email protected]>"
<[email protected]<mailto:[email protected]>>
Message-ID:
<ps1pr03mb1820d1a6779e41e6fff4b16fe6...@ps1pr03mb1820.apcprd03.prod.outlook.com<mailto:ps1pr03mb1820d1a6779e41e6fff4b16fe6...@ps1pr03mb1820.apcprd03.prod.outlook.com>>
Content-Type: text/plain; charset="iso-8859-1"
Hi,
We have a need to pull large amounts of data out of marklogic as quickly as
possible. I found that doing xquery searches like query-by-example were very
slow. Using the cts functions led to a big speed increase. However, it isn't
clear whether my current approach is the optimum, or whether there are other
better alternatives. Unfortunately, while there is a lot of documentation
describing many different ways of doing things in marklogic, there seems to be
very little documentation describing what are the best or most efficient
approaches (e.g., what if your goal is not only to run a query successfully,
but to maximize its performance?). At present, I'm using the java api to pull
documents. I'm using the theCall.xquery(query) function in Java to run custom
xquery through the rest api. The xquery is as follows:
<data>
for $x in cts:search(fn:doc(),cts:and-query((
cts:element-value-query(xs:QName('Department'), 'Sales'),
cts:element-range-query(xs:QName('Date'), '>', xs:date('2015-01-01')),
cts:element-range-query(xs:QName('Date'), '<', xs:date('2015-01-03')),
cts:not-query(cts:element-value-query(xs:QName('Date'), 'NULL')) )),
'unfiltered' , 0.0) )), 'unfiltered' , 0.0) return
fn:concat($x//Department,'|',$x//Total,'|',$x//Location' ')}
</data>
There are indexes on Date and Department. The xquery wraps all of the documents
in the <data> tags and sends the results to the java program. It then strips
the <data> tags and prints the results to text file.
I have found that you can run multiple threads in the java that request
different "chunks" of the data by using the criterions of [1 to 1000000],
[1000001 to 2000000], etc.
This approach is much faster than our original approach - 12 hours with 8
threads, rather than 75 hours using query-by-example. However, it is not clear
if this is the fastest way, or there are further optimizations or better
approaches. For instance, when pulling the actual elements from the documents,
I found that having them indexed made no different to performance. Is there a
way of pulling from the indexes to improve performance? Is there a way to
specify the elements you want in the cts:search that will improve performance?
Is there a more efficient way to restrict the search range? Is there
documentation describing the most efficient approaches to querying marklogic?
Thanks.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://developer.marklogic.com/pipermail/general/attachments/20161010/d2e00150/attachment-0001.html
------------------------------
_______________________________________________
General mailing list
[email protected]<mailto:[email protected]>
Manage your subscription at:
http://developer.marklogic.com/mailman/listinfo/general
End of General Digest, Vol 148, Issue 13
****************************************
This message and any attachments are intended only for the use of the addressee
and may contain information that is privileged and confidential. If the reader
of the message is not the intended recipient or an authorized representative of
the intended recipient, you are hereby notified that any dissemination of this
communication is strictly prohibited. If you have received this communication
in error, please notify us immediately by e-mail and delete the message and any
attachments from your system.
_______________________________________________
General mailing list
[email protected]<mailto:[email protected]>
Manage your subscription at:
http://developer.marklogic.com/mailman/listinfo/general
_______________________________________________
General mailing list
[email protected]
Manage your subscription at:
http://developer.marklogic.com/mailman/listinfo/general
