As the new GDELT GEN4 crawling infrastructure continues to come online, we are completing the transition of our internal JSON formatting libraries used to create our output datasets from our legacy ordered formatting to native JSON construction. Historically, GDELT's JSON datasets relied on standardized external JSON libraries to construct the output values, but we then assembled them into strictly ordered output records in order to ensure that variables always appeared in the same order.
This eased parsing for some major legacy tooling that made assumptions about field ordering, but introduced significant overhead to maintain strict ordering. Now that nearly all major analytic packages natively support JSON import, as we transition GDELT's datasets over to our new GEN4 crawling and processing infrastructure, we will be completing the transition to native JSON output construction, meaning that moving forward the internal ordering of fields within a JSON record will be nondeterministic, in accordance with the JSON format specification.