Optional diagnostics produced by read_json_with_diagnostics.
More...
#include <json.hpp>
Public Attributes | |
| std::vector< std::string > | top_level_columns_with_schema_mismatch |
Top-level column names whose value tree contained at least one JSON node whose category (NC_STRUCT / NC_LIST / NC_VAL) did not match the requested schema type. Empty when the reader is invoked without a user-supplied schema or when no mismatches were observed. More... | |
Optional diagnostics produced by read_json_with_diagnostics.
Reserved for signals that don't belong on the result table_metadata because they are reader-specific (and adding them to the shared column_name_info would change the public ABI for that struct, breaking downstream packages that have its destructor inlined into their binaries).
Definition at line 935 of file io/json.hpp.
| std::vector<std::string> cudf::io::json_reader_diagnostics::top_level_columns_with_schema_mismatch |
Top-level column names whose value tree contained at least one JSON node whose category (NC_STRUCT / NC_LIST / NC_VAL) did not match the requested schema type. Empty when the reader is invoked without a user-supplied schema or when no mismatches were observed.
Consumers (e.g. spark-rapids-jni) can use this signal to implement their own per-column policy on schema mismatch (e.g. Spark's JacksonParser nulls the entire depth-1 ancestor field for such rows).
Definition at line 946 of file io/json.hpp.
1.9.1