Batching SQL statements with Java

Regular expressions are a powerful tool in every Java developer’s toolkit. They allow us to validate input, parse strings, and perform complex text transformations with just a few lines of code. However, this power comes with a hidden performance cost if not used correctly.

The key to unlocking efficient regex in Java lies in understanding the java.util.regex.Pattern class. In this post, we’ll explore the best practices for using Pattern, how to avoid common performance pitfalls, and why you should be wary of the “convenient” regex methods on the String class.

Using agnostic ways

JDBC API

PreparedStatement.addBatch() + executeBatch() Gestion manuelle des lots, erreurs partielles, getUpdateCounts()

JPA

Flush manuel en boucle (persist() + flush() + clear() toutes les N entités) Limites : pas de addBatch direct ; dépend du provider pour l’ordre/optimisation

Using builders/DSL

JOOQ

dslContext.batch(insert…).execute() Avantage : type-safe, génère du vrai batch JDBC

MyBatis

SqlSessionFactory.openSession(ExecutorType.BATCH) <foreach> pour lots dynamiques

Using an ORM

Hibernate

Propriétés : hibernate.jdbc.batch_size, hibernate.order_inserts/updates @BatchSize sur entités/collections session.flush()/clear() obligatoire

Blaze persistence

Pièges et bonnes pratiques

The choice

Any abstraction over the driver brings its own overhead.

Quand clear() est vital (mémoire, ID generation) Erreurs de batch (tout le lot échoue si une ligne KO, sauf continueOnError) Comparaison : JDBC > DSL > ORM pour perfs brutes ; ORM pour simplicité avec entités

---

References

• Demystifying Java Object Sizes: Compact Headers, Compressed Oops, and Beyond by Peter Lawrey
• String.matches(String regex)
• RegExUtils.java

Demo

A showcase of the concepts illustrated in this post is available here: regex-performance-benchmark