Problem
When converting CSV data today, pyarrow users must either enumerate column names or construct a schema to force all fields to a given type. This makes simple workflows like "read everything as string" clumsy when the schema is not known ahead of time.
Proposed change
Expose a single default column type on arrow::csv::ConvertOptions and plumb it through the bindings so callers can say ConvertOptions(column_type=pa.string()). The option should apply to any columns not listed explicitly in column_types, including columns added via include_missing_columns.
Implementation status
A local branch adds ConvertOptions::column_type, wires it through the C++ reader, exposes it in pyarrow.csv, updates the docs, and adds unit tests covering the new behavior.
Component(s)
C++, Python
Next steps
Raise a PR with the implementation and tests once this ticket is accepted.
Problem
When converting CSV data today, pyarrow users must either enumerate column names or construct a schema to force all fields to a given type. This makes simple workflows like "read everything as string" clumsy when the schema is not known ahead of time.
Proposed change
Expose a single default column type on
arrow::csv::ConvertOptionsand plumb it through the bindings so callers can sayConvertOptions(column_type=pa.string()). The option should apply to any columns not listed explicitly incolumn_types, including columns added viainclude_missing_columns.Implementation status
A local branch adds
ConvertOptions::column_type, wires it through the C++ reader, exposes it inpyarrow.csv, updates the docs, and adds unit tests covering the new behavior.Component(s)
C++, Python
Next steps
Raise a PR with the implementation and tests once this ticket is accepted.