If you're thinking the `TableRow` here is related to SQL, it isn't. This is a 
BigQuery datatype. The JSON parsing is being configured to use a global 
instance of Jackson in `ParseTableRowJson`, later in this file. The `TableRow` 
type is really just `Map<String,Object>`, so Jackson will pick a type for 
numbers based on global config which must not be the same on all runners. (See 
https://fasterxml.github.io/jackson-databind/javadoc/2.8/index.html?com/fasterxml/jackson/databind/DeserializationFeature.html
 for some of these config options.) One way to fix this would be to provide the 
actual types for json decoding rather then using `Map<String,Object>`.

[ Full content available at: https://github.com/apache/beam/pull/6311 ]
This message was relayed via gitbox.apache.org for [email protected]

Reply via email to