Profiler les requêtes avec EXPLAIN et PROFILE

Before optimizing a query, you must understand how Neo4j executes it. The EXPLAIN and PROFILE keywords reveal the execution plan.

EXPLAIN shows the planned operators without actually executing the query. Use it to inspect plans safely, even for writes.

PROFILE runs the query and reports actual db hits and rows per operator. This is your main tuning tool.

Plans are trees of operators such as:

• NodeByLabelScan — scans all nodes with a label
• NodeIndexSeek — uses an index (fast)
• Expand — follows relationships
• Filter — applies a predicate

A NodeByLabelScan on a large label means Neo4j reads every node of that type. Replacing it with a NodeIndexSeek is the most common big win.

db hits estimate work done. Fewer is better. Compare db hits before and after a change to confirm improvement.

If PROFILE shows a label scan filtered by a property, add an index on that property and re-profile.

Each operator shows estimated and actual rows. A big gap between them means the planner's statistics are stale.

An Eager operator forces Neo4j to materialize all rows before continuing, often inflating memory. Restructuring the query can sometimes remove it.

A CartesianProduct warning means two disconnected patterns multiply together. Connect them with a relationship or a WHERE to avoid an explosion.

Profile, find the most expensive operator, address it (index, restructure, anchor the match), then profile again. Iterate until db hits are acceptable.

You learned to profile Cypher queries:

• EXPLAIN shows the plan; PROFILE adds real metrics
• Replace label scans with index seeks
• Minimize db hits
• Avoid cartesian products and unnecessary eager operators
• Iterate: profile, fix, re-profile