We have been using spark.read with inferSchema = True to validate AutoLoader schema inferencing. But I have a suspicion that they do these differently from each other and may not always yield the identical results.

Has anyone ever answered this question? Does anyone know of documentation that can speak to whether there is a difference between them?