XML schema creation takes long time

Question

I have the following code:

public XsdValidator(Resource... xsds) {
    Preconditions.checkArgument(xsds != null);
    try {
      this.xsds = ImmutableList.copyOf(xsds);
      SchemaFactory schemaFactory = SchemaFactory.newInstance(W3C_XML_SCHEMA_NS_URI);
      LOGGER.debug("Schema factory created: {}",schemaFactory);
      StreamSource[] streamSources = streamSourcesOf(xsds);
      LOGGER.debug("StreamSource[] created: {}",streamSources);
      Schema schema = schemaFactory.newSchema(streamSources);
      LOGGER.debug("Schema created: {}",schema);
      validator = schema.newValidator();
      LOGGER.debug("Validator created: {}",validator);
    } catch ( Exception e ) {
      throw new IllegalArgumentException("Can't build XsdValidator",e);
    }
  }

It seems the line schemaFactory.newSchema(streamSources); takes a very long time (30 seconds) to execute against my XSD file.

After many tests on this XSD, it seems it's because I have:

  <xs:complexType name="entriesType">
    <xs:sequence>
      <xs:element type="prov:entryType" name="entry" minOccurs="0" maxOccurs="10000" />
    </xs:sequence>
  </xs:complexType>

The problem is maxOccurs="10000"

With maxOccurs="1" or maxOccurs="unbounded", it is very fast.

Can someone tell me what's the problem of using maxOccurs="10000" ?

Can you please try using maxOccurs="unbounded"? What do you get in your case? — Petru Gardea, Oct 23 '13 at 15:24
The result for unbounded is already in my question @PetruGardea : it's fast — Sebastien Lorber, Oct 25 '13 at 21:10
Somehow I've missed that... It means that it is trying to allocate memory (most likely). I would run a performance profiler, just to confirm. — Petru Gardea, Oct 25 '13 at 21:34
Hmm... you've asked "what's the problem of using"... hence the what. As far as why, I've used to run into [this problem](http://www-01.ibm.com/support/docview.wss?uid=swg21296034). I am sure that if you narrow search terms to match your particular setup, you'll find out more... — Petru Gardea, Oct 26 '13 at 01:52
Link to xerces bug : https://issues.apache.org/jira/browse/XERCESJ-773 — Bee, Apr 03 '15 at 18:14

score 4 · Accepted Answer · answered Oct 26 '13 at 13:54

4

Based on my personal experience, having particles bounded by what some may consider "unreasonably" high values is cause for performance problems (this link is from my browser's favourites).

The underlying cause seems to be memory allocation (to the effect indicated by the maxOccurs value).

Also, I recall a documentation item which was stating a threshold value beyond which, for all intents and purposes, the parser would actually treat the maxOccurs as unbounded, regardless of what the XSD says (I'll revisit this post if I find it).

answered Oct 26 '13 at 13:54

Petru Gardea

21,373
2
50
62

The link you provided is only a RAD (Eclipse) performance issue and not an application performance problem... right? @Petru Gardea – javaPlease42 Dec 03 '13 at 20:51
@javaPlease42, not really, if you go through it, you may notice that it is a Xerces problem; some people reported that it is still manifesting in 2.11 (I cannot confirm that). – Petru Gardea Dec 03 '13 at 22:51

XML schema creation takes long time

1 Answers1