Optionally specify limit for number of entities in a record.

[blackwinter]

Just a proof-of-concept to gauge interest in such a feature; the immediate need has been rendered moot (due to the record's holdings being suppressed from publishing now).

This is a brute-force approach to dealing with OOM situations when Alma records have an excessive number of items (e.g. 99374518570506441: >12000 entities = ~10 GB heap for the Record instance). Loosely modeled after JAXP's totalEntitySizeLimit, except that we don't abort processing but instead skip the remaining entities.

Use Metafix instance setter setMaxEntityCount(int) or set system property org.metafacture.metafix.maxEntityCount=<int>.

Alternative options:

• Increase maximum heap size for JVM.
• Significantly reduce memory requirement for Record instances.

Possible optimizations (implemented in dc4f1af):

1. Drop the setter and make the limit static final; this way the JVM can (potentially) eliminate the check altogether when no limit has been set (less than zero).
2. Skip all remaining literals (instead of just the ones inside entities), thereby simplifying the check in literal().

[dr0i]

I think this looks good. It prevents possible flooding, as we already experienced. So I like this feature.

[fsteeg]

Looks good, but should probably include a test.

[blackwinter]

Regarding test: As currently implemented, the feature cannot be exercised under test conditions because the limit is already set (system property already evaluated) before the test gets the chance to influence it. Either change the implementation to make it testable (keeping performance implications in mind) or skip test coverage in this case (assuming that this feature is only used in extremely rare edge cases).

[blackwinter]

Adding a unit test proved to be quite messy when attempting to keep the instance member variable final. Added an integration test instead (a10b3d8). This has the added benefit that it actually exercises the retrieval of the system property.