Add other types to give a more robust grouping proc

[gem-neo4j]

#116

[loveleif]

👍

[loveleif]

Regexing the property name, not very rubust, but I guess this is the way of the apoc.

[loveleif]

Compiling a regex pattern is a bit expensive though (I don't think they're cached in java) so we could consider replacing this with k.startsWith("avg_") && k.size() > "avg_".length. Or to compile the Pattern once in a field and reuse.

[loveleif]

Suggested change

	if (k.matches("^avg_.+") && v instanceof DurationValue) {
	Long count = ((Number) pc.getProperty(k + "_count", 0)).longValue();
	entry.setValue(divDurationValue((DurationValue) v, count));
	}
	if (k.matches("^avg_.+") && v instanceof DurationValue duration) {
	Long count = ((Number) pc.getProperty(k + "_count", 0)).longValue();
	entry.setValue(divDurationValue(duration, count));
	}

[loveleif]

Is it intentional to use .longValue() here? I thinking about Doubles for example, that will get rounded by that call.

[gem-neo4j]

APOC is the one that sets this value, so we know it is a long

[loveleif]

Suggested change

	if (value instanceof DurationValue) {
	DurationValue dv =
	(DurationValue) pc.getProperty(key, DurationValue.duration(0, 0, 0, 0));
	pc.setProperty(key, ((DurationValue) value).add(dv));
	if (value instanceof DurationValue duration) {
	DurationValue dv =
	(DurationValue) pc.getProperty(key, DurationValue.duration(0, 0, 0, 0));
	pc.setProperty(key, (duration.add(dv));

[loveleif]

Could we use org.neo4j.values.storable.Values.COMPARATOR here to not have to maintain this in apoc?

return switch (Values.COMPARATOR.ternaryCompare(Value.of(a), Value.of(b))) {
  case UNDEFINED -> ...;
  case EQUAL -> ...;
  case GREATER_THAN -> ...;
  case SMALLER_THAN -> ...;
}

[gem-neo4j]

I did not know this existed, looks interesting thanks! I'll play with it :D