Tracing inconsistency

[eapache]

Describe the bug

The order of tracing events is inconsistent depending on how you execute a query.

If you instantiate a GraphQL::Query with a query string and schema, then call #result on it, then the lex and parse trace events fire outside of execute_multiplex.

If instead you call Schema.execute or Schema.multiplex, then the lex and parse trace events fire inside of execute_multiplex.

This is because Query calls Multiplex.run_queries inside of with_prepared_ast:

graphql-ruby/lib/graphql/query.rb

Lines 198 to 200 in 5c3a8ec

	with_prepared_ast {
	Execution::Multiplex.run_queries(@schema, [self], context: @context)
	}

Where-as Schema.multiplex calls Multiplex.run_all directly which fires the execute_multiplex tracer before it does anything that triggers lexing and parsing.

@rmosolgo maybe this isn't technically a bug but it's surprising enough it should at least be called out in the documentation.

Versions

1.12.5, efcedef, probably master

cc @chrisbutcher

[eapache]

It might be as simple as not calling with_prepared_ast in Query.result?

[rmosolgo]

Thanks for reporting this -- I agree it would be better if it was consistent!

It might be as simple as not calling with_prepared_ast in Query.result?

I believe it -- since execution will eventually call some other method that causes the query to prepare its AST.