Recover the column names from a fixed-width Sea-Bird ASCII header

The .asc header is fixed-width, and in most CalCOFI files adjacent names touch: Sbeox0ML/L and Sbeox0Mm/Kg arrive as Sbeox0ML/LSbeox0Mm/Kg. Splitting on whitespace therefore produces the wrong number of columns and silently mis-assigns every column after the collision — which in a QC tool is worse than not reading the file at all.

Usage

sbe_split_header(header, data_rows)

Arguments

header

the header line

data_rows

a few data lines to take column edges from

Value

character vector of column names

Details

Both the numbers and the header names are RIGHT-ALIGNED in their column, so the aligned edge is each field's stop position. Those are taken from the data rows (where fields are always separated) and the header is sliced at them.

It ERRORS rather than guessing when the result is not self-consistent — an empty name, a name containing a space, or a count that does not match the data. Measured on the CalCOFI archive this reads ~86% of .asc files cleanly; the rest report the problem and ask for the .cnv, whose header is unambiguous.