Separate existing converters from the DataFrame core so consumers only acquire the format-specific dependencies they need.
Proposed packages:
Microsoft.Data.Analysis: DataFrame, columns, storage and existing data-wrangling operations. Retains IDataView implementation and existing ML.NET interoperability.Microsoft.Data.Analysis.Converters: Existing CSV and ADO.NET converters, using core and framework APIs without database-provider dependencies.Microsoft.Data.Analysis.Arrow: Existing Arrow conversion and the Apache.Arrow dependency.
Additional converters can be implemented as independent packages using the public DataFrame API, without modifying the core library. Converters requiring external libraries should have separate packages rather than introducing those dependencies into core or the basic-converters package.
Separate existing converters from the DataFrame core so consumers only acquire the format-specific dependencies they need.
Proposed packages:
Microsoft.Data.Analysis: DataFrame, columns, storage and existing data-wrangling operations. RetainsIDataViewimplementation and existing ML.NET interoperability.Microsoft.Data.Analysis.Converters: Existing CSV and ADO.NET converters, using core and framework APIs without database-provider dependencies.Microsoft.Data.Analysis.Arrow: Existing Arrow conversion and theApache.Arrowdependency.Additional converters can be implemented as independent packages using the public DataFrame API, without modifying the core library. Converters requiring external libraries should have separate packages rather than introducing those dependencies into core or the basic-converters package.