Column type guessing

[raffaem]

readxl::read_excel guesses column types based on the value of the first guess_max (an argument of the function) of its cells (reference)

This causes problems in importing when a numeric cell way down in my Excel file is silently converted into a boolean, without any warning of any sort.

This problem doesn't happen with openxlsx.

Can you specify in the documentation how openxlsx guesses column types?

My understanding is that Excel only provides cell types, not column types.

[mlell]

openxlsx looks at the whole table and not only on the first guess_max lines. In readxl you can set guess_max to a very high value (e.g. Inf) to avoid the problematic behaviour.

In openxlsx, the process is aided by the fact that excel stores every string only once in a central storage and every cell with that string has the address of the correct string. Similarly for logical values. Dates have another treatment that I do not know right now. A column that has no cells with type "string references" is numeric

openxlsx/src/read_workbook.cpp

Lines 665 to 672 in 1632db0

	// If it contains any strings it will be a character column
	IntegerVector char_cols_unique;
	if(all(is_na(string_inds))){
	char_cols_unique = -1;
	}else{
	IntegerVector columns_which_are_characters = cols[string_inds - 1];
	char_cols_unique = unique(columns_which_are_characters);