Great article. One thing I've found with the InferSchemaWithAvro and working with CSVs is that it only seems to infer the type based on the first line after the header. The "Number Of Records To Analyze" property specifically states that it's only to be used with JSON content type. I'm having some difficulty right now thinking on a way to work around this. I've got a CSV column with "12345" on the first line and "12345.45" on the second line and it's inferring the type as LONG when I want it to be DOUBLE. EDIT: It also doesn't work completely as expected with negative values. (STRING instead of numeric) ... View more