CSV loader options. More...
#include <data_file_CSV.hpp>
Public Attributes | |
| std::string | class_column |
| which column carries the class label: "" = the last column; a decimal integer = that 0-based file column index; anything else = a header column name (which implies a header row) | |
| std::string | delimiter |
| field separator: "auto" (detect from the first line), "comma", "semicolon" or "tab" | |
| std::string | header |
| whether the first line is a header row: "auto" (detect), "yes" or "no" | |
| bool | missing_to_active |
| when true, missing FEATURE tokens are replaced by missing_to and the row is kept instead of skipped; a missing CLASS label still skips the row, since a row without a label cannot be assigned to any class (the fst4 tool sets this from --csv-missing-to) | |
| float | missing_to |
| the numeric code that replaces a missing feature value when missing_to_active. The mapping is purely a substitution: to have the code imputed away, configure the accessor's missing-value substitution with the SAME code; leave that unset to keep the code in the data as an explicit missingness indicator | |
| bool | missing_to_unpaired |
| set by the caller when no missing-value substitution is configured; the load summary then names the pairing option, so that mapping-without-imputing reads as a deliberate choice rather than an oversight | |
CSV loader options.