Inconsistent exception types thrown while parsing XContent

[olcbean]

In most cases, if the expected START_OBJECT is not detected, an IllegalArgumentException is thrown :

elasticsearch/core/src/main/java/org/elasticsearch/common/xcontent/XContentParserUtils.java

Lines 75 to 80 in b88dbe8

	public static void ensureExpectedToken(Token expected, Token actual, Supplier<XContentLocation> location) {
	if (actual != expected) {
	String message = "Failed to parse object: expecting token of type [%s] but found [%s]";
	throw new ParsingException(location.get(), String.format(Locale.ROOT, message, expected, actual));
	}
	}

However ObjectParser throws IllegalStateException :

elasticsearch/core/src/main/java/org/elasticsearch/common/xcontent/ObjectParser.java

Lines 149 to 151 in b88dbe8

	if (token != XContentParser.Token.START_OBJECT) {
	throw new IllegalStateException("[" + name + "] Expected START_OBJECT but was: " + token);
	}

When parsing mget two different exception types may be thrown :
ParsingException

elasticsearch/core/src/main/java/org/elasticsearch/action/get/MultiGetRequest.java

Lines 323 to 330 in b88dbe8

	if ((token = parser.nextToken()) != XContentParser.Token.START_OBJECT) {
	final String message = String.format(
	Locale.ROOT,
	"unexpected token [%s], expected [%s]",
	token,
	XContentParser.Token.START_OBJECT);
	throw new ParsingException(parser.getTokenLocation(), message);
	}

IllegalArgumentException

elasticsearch/core/src/main/java/org/elasticsearch/action/get/MultiGetRequest.java

Lines 364 to 366 in b88dbe8

	if (token != XContentParser.Token.START_OBJECT) {
	throw new IllegalArgumentException("docs array element should include an object");
	}

while RepositoryData consistently throws ElasticsearchParseException :

elasticsearch/core/src/main/java/org/elasticsearch/repositories/RepositoryData.java

Line 487 in b88dbe8

throw new ElasticsearchParseException("start object expected");

---

I think that it would be much cleaner if only one exception type is used (at least from now on).

For example what exception type should be thrown when a stricter validation of query field names is introduced (#26013). A different one per request (based on the exception thrown till now in the request)? Or an exception of type X?

Or maybe some of the exception types should be changed? For example can IllegalStateException be replaced by IllegalArgumentException?

[javanna]

I think that ParsingException is the way to go , as it also includes the position where the problem was found in the original input. IllegalArgumentException sounds ok but it doesn't have the position info which is very useful. IllegalStateException seems wrong to me in this context.

[olcbean]

Thanks @javanna! So we should stick to the ParsingException in general.
Actually one more question : shouldn't the ParsingException be actually an IllegalArgumentException?

[javanna]

Yes let's stick to ParsingException

shouldn't the ParsingException be actually an IllegalArgumentException

it could be, I guess it is convenient to subclass ElasticsearchException so that a few common methods around serialization (transport and REST) can be inherited. I do not see an advantage in changing this at this point.

[liketic]

Hi @olcbean @javanna I found this issue is marked as adoptme. Can I take this? 👍

[olcbean]

Hey @liketic, I have already pushed a PR for ObjectParser. Sorry about that.

[olcbean]

Closed by #27302