Setting routing to null in some Builder methods can result in `?routing=null` being part of the URL

[Stexxen]

Java API client version

9.3

Java version

25

Elasticsearch Version

9.3

Problem description

In 9.3 you changed the routing code to use a List<String> instead of String

elasticsearch-java/java-client/src/main/java/co/elastic/clients/elasticsearch/core/IndexRequest.java

Line 304 in b4b8c3e

private final List<String> routing;

Previously setting routing to null meant nothing as it was detected and not used. But in 9.3, setting routing to null creates a List of a single null element

elasticsearch-java/java-client/src/main/java/co/elastic/clients/elasticsearch/core/IndexRequest.java

Lines 771 to 773 in de9b5d4

	public final Builder<TDocument> routing(String value, String... values) {
	this.routing = _listAdd(this.routing, value, values);
	return this;

Then on a build call this section is executed

elasticsearch-java/java-client/src/main/java/co/elastic/clients/elasticsearch/core/IndexRequest.java

Lines 1033 to 1035 in b4b8c3e

	if (ApiTypeHelper.isDefined(request.routing)) {
	params.put("routing", request.routing.stream().map(v -> v).collect(Collectors.joining(",")));
	}

which results in the following being appended to url

?routing=null

And elasticsearch then ends up looking for or adding the document in potential the wrong place (depending on how many shades are being used for that index).

[Stexxen]

In further review this seems to effect multiple classes. Here are some I've spotted, but there may be others.

• co.elastic.clients.elasticsearch.core.GetRequest
• co.elastic.clients.elasticsearch.core.ExistsRequest
• co.elastic.clients.elasticsearch.core.mget.MultiGetOperation
• co.elastic.clients.elasticsearch.core.UpdateByQueryRequest

[l-trotta]

Hey @Stexxen, thanks for pointing this out. I'm taking a look, this is a result of elastic/elasticsearch-specification#5676 being merged to support multiple comma separated routing options, I'll fix serialization to avoid adding null.

[l-trotta]

I'm trying to understand if this is a rule that should be applied to every path/query parameter that serializes as a csv string, since this is a common behavior that isn't unique to "routing", I'll update this as soon as I find out.

[Stexxen]

Thanks @l-trotta. Although I'm inclined to look at it from the other direction. As in, Why would any URLs require nulls to be stringified as as a parameter?

But good luck and thanks for looking at this issue.