--emit neo4j: all JField nodes share id <fqn>#field#null, collapsing every field of a class into one node #156

Summary

--emit neo4j assigns every :JField node the id <ownerFqn>#field#null, so all fields of a class collapse into a single :JField node (the unique-constraint MERGE keeps only the last). Field data is almost entirely lost.

Environment

• codeanalyzer-java 2.4.0

Impact

Measured on the daytrader8 sample (level 2): all 122 :JField nodes in the database have an id ending in #field#null, and every one has name = null. A class with 17 fields (e.g. XMLObject, RunStatsDataBean) ends up with a single :JField node. Any consumer reading fields from the graph (e.g. the CLDK SDK's read-only Java backend) gets one field per class instead of all of them.

MATCH (f:JField) WHERE f.id ENDS WITH '#field#null' RETURN count(f);   // 122  (== total JField count)
MATCH (:JType {id:'...XMLObject'})-[:J_HAS_FIELD]->(f:JField) RETURN f.id;
//  -> '...XMLObject#field#null'   (a single node; ref analysis.json has 17 field_declarations)

Root cause

neo4j/GraphProjector.java, projectField:

String id = ownerFqn + "#field#" + f.getName();

A Java field declaration is keyed by its variables list, not a single name — Field.getName() is null here (the model has no scalar field name; the projected node even sets "name", f.getName() ⇒ null). So every field's id is <fqn>#field#null, and the j_field_id uniqueness constraint collapses them on MERGE.

Suggested fix

Key the field id by something unique per declaration — e.g. the first variable name, the joined variables, or a positional index:

String key = (f.getVariables() != null && !f.getVariables().isEmpty())
        ? String.join("+", f.getVariables())
        : String.valueOf(index);
String id = ownerFqn + "#field#" + key;

(and project the real field name/variables so they round-trip).