Column mapping

Automatic discovery derives names from text. Explicit sources instead make coordinates part of the mask, useful for neighboring tables, comments and ambiguous headers.

Explicit bindings

TableSource.columns binds a nonblank output/filter alias to an uppercase Excel column A through XFD. Aliases must be unique; several aliases can refer to one physical column. Header text is not interpreted. data_start_row is required, data_end_row is inclusive and optional, and skip_rows lies inside those bounds.

Use source for one region or sources for several. They are mutually exclusive. Each source may have a unique name and a sheet restriction. An omitted/null source sheet applies to every top-level selected worksheet. A source's rules replaces the entire shared rule set, including defaults; it is not a partial merge. Overlapping explicit regions intentionally emit a record for each source.

Header-bound sources

Alternatively omit columns and supply header_start_row. Optional first_column/last_column bound the header window. header.rows determines its height; the whole header must precede data_start_row. This mode retains real header names and normalization while using known row coordinates.

Values outside explicit scope are not converted. XML structure, coordinates and resource limits are still checked throughout the worksheet, even after the end bound. A corrupt tail remains an error.

CSV uses the same column letters for record positions and logical record numbers for source rows. See CSV. Complex layouts contains a runnable example; filter models lists exact constraints.