Improve performance of saving entities by using insert(…) if possible [DATAMONGO-1054]

[spring-projects-issues]

Titi Wangsa opened DATAMONGO-1054 and commented

Mongo Template has this method:

void insert(Collection<? extends Object> batchToSave, Class<?> entityClass);

that does a bulk insert

Basic repository interfaces that extend from MongoRepository does not have it. They have:

<S extends T> List<S> save(Iterable<S> entites);

this is considerably much slower.

actual:
saving using
MongoRepository.save(Iterable<S> entites)
for 6,000+ items took a few minutes when write concern is set to journaled

expected
ability to save it in matter of milliseconds
by using
MongoOperation.insert(Collection<? extends Object> batchToSave, Class<?> entityClass);

workaround:
I created a CustomImpl that for my repository class that calls mongoOperation. this is much faster. I am hoping that it .insert would be a built-in feature on the next release

---

Attachments:

• test.tar.gz (2.84 kB)

Issue Links:

• DATAMONGO-903 Add insert method to MongoRepository interface
  ("supersedes")

Referenced from: pull request #253, and commits 2e74c19, 1b7e077, e58fa28, 7ebaf93, 714446b, a212b75, fa2c609

Backported to: 1.6.2 (Evans SR2)

[spring-projects-issues]

Oliver Drotbohm commented

Sounds like a legit request. Any chance you provide a tiny test case that shows the performance improvement?

[spring-projects-issues]

Titi Wangsa commented

sample project to show the perfomance impact of MongoRepository.save vs MongoTemplate.insert.

here is the output of: mvn clean test
Sep 13, 2014 10:06:44 PM test.test.ShorelineRepositoryTest testSave
INFO: shorelineRepository.save took 17492ms
Sep 13, 2014 10:06:44 PM test.test.ShorelineRepositoryTest testInsert
INFO: mongoTemplate.insert took 40ms

mongo --version
MongoDB shell version: 2.6.4

java -version
java version "1.8.0_20"
Java(TM) SE Runtime Environment (build 1.8.0_20-b26)
Java HotSpot(TM) 64-Bit Server VM (build 25.20-b23, mixed mode)

uname -a
Darwin titis-mbp 13.3.0 Darwin Kernel Version 13.3.0: Tue Jun 3 21:27:35 PDT 2014; root:xnu-2422.110.17~1/RELEASE_X86_64 x86_64

[spring-projects-issues]

Thomas Darimont commented

Please revise

[spring-projects-issues]

Oliver Drotbohm commented

We're now optimizing calls to the save(…) by checking whether the given entities are considered new (determined by inspecting the identifier). In case they're all new we trigger the insert(…) method instead of save(…). For modeling scenarios with manually assigned identifiers Thomas introduced dedicated insert(…) methods on MongoRepository so that you can force inserts to be used. We generally recommend sticking to save(…) or explicitly declaring the insert(…) methods on your repository to avoid exposing a store specific repository to clients