ppforest2::io::csv Namespace Reference

Classes
struct	FeatureSet
	Result of parsing a feature-only CSV (no response column). More...

Functions
stats::DataPacket	read (std::string const &filename)
	Read a CSV file into a DataPacket.
stats::DataPacket	read (std::string const &filename, types::Mode mode)
	Read a CSV file, parsing the response column in a known mode.
FeatureSet	read_features_from_string (std::string const &content)
	Parse a feature-only CSV from an in-memory string.
stats::DataPacket	read_sorted (std::string const &filename)
	Read a CSV file and sort rows ascending by the response column.
stats::DataPacket	read_sorted (std::string const &filename, types::Mode mode)
	Read a CSV file sorted by response, parsing the response in a known mode.
types::OutcomeVector	remap_labels (stats::DataPacket const &data, types::Names const &target_groups)
	Re-encode data.y label codes into a target label space.
void	write (stats::DataPacket const &data, std::string const &filename)
	Write a DataPacket to a CSV file (features followed by label, no header).

Function Documentation

read() [1/2]

stats::DataPacket ppforest2::io::csv::read

(

std::string const &

filename

)

Read a CSV file into a DataPacket.

Assumes the last column is the response variable (group label as string) and all preceding columns are features. Categorical feature columns are automatically detected and integer-encoded. String labels are mapped to contiguous integer codes starting at 0.

Parameters

filename

Path to the CSV file.

Returns

A DataPacket containing the feature matrix and response vector.

Exceptions

UserError

on any failure: missing file, empty file, rows whose column count differs from the header, or a column mixing numeric and non-numeric cells (stray "NA", empty, or non-finite values).

read() [2/2]

stats::DataPacket ppforest2::io::csv::read	(	std::string const &	filename,
		types::Mode	mode )

Read a CSV file, parsing the response column in a known mode.

Same as read, but skips the written-form mode detection: mode states how to parse the response column. Classification maps labels to integer codes in first-appearance order (populating group_names); regression parses a continuous float response and rejects non-numeric values. Rows keep their file order — use this when predictions must line up with the input rows (e.g. predict), with the mode taken from the model.

Exceptions

UserError

on any failure (missing file, parse error, malformed shape, non-numeric regression response).

read_features_from_string()

FeatureSet ppforest2::io::csv::read_features_from_string

(

std::string const &

content

)

Parse a feature-only CSV from an in-memory string.

The first row is the header (feature names); every subsequent row is one observation. There is no response column. Used by the serve subcommand to parse POST /predict request bodies — categorical encoding runs per-call, so callers must encode categoricals consistently with the training data.

Exceptions

UserError

on empty body, missing header, no data rows, or malformed shape.

read_sorted() [1/2]

stats::DataPacket ppforest2::io::csv::read_sorted

(

std::string const &

filename

)

Read a CSV file and sort rows ascending by the response column.

Mode is detected from the y column's written form:

• If any value carries fractional or scientific notation (., e, E), y is parsed as a continuous float response (regression shape) and group_names is empty.
• Otherwise, y is mapped to integer codes in first-appearance order and group_names carries the original label strings (classification shape). This keeps integer-coded label CSVs like Wine and Glass on the classification path.

Rows are sorted ascending by the encoded y, which gives the training routines what they need: classification — contiguous groups; regression — y-ordered rows for ByCutpoint::init's median split.

Exceptions

UserError

on any failure (missing file, parse error, malformed shape) — CSV reading failures are user-facing by nature.

read_sorted() [2/2]

stats::DataPacket ppforest2::io::csv::read_sorted	(	std::string const &	filename,
		types::Mode	mode )

Read a CSV file sorted by response, parsing the response in a known mode.

Same as the one-argument read_sorted, but skips the written-form mode detection: mode states how to parse the response column. Use this when the mode is already decided (e.g. train --mode regression) so that integer-written regression responses stay numeric instead of being label-encoded.

Exceptions

UserError

on any failure, including a non-numeric response in regression mode.

remap_labels()

types::OutcomeVector ppforest2::io::csv::remap_labels	(	stats::DataPacket const &	data,
		types::Names const &	target_groups )

Re-encode data.y label codes into a target label space.

data.y carries integer codes indexing data.group_names — the order labels first appeared in the file. Model predictions and metrics use the model's own label order (meta.groups), which can differ when a prediction file lists classes in a different order than the training file. Returns a copy of data.y re-encoded so that code i means target_groups[i].

Exceptions

UserError

if the file contains a label that is not present in target_groups.

write()

void ppforest2::io::csv::write	(	stats::DataPacket const &	data,
		std::string const &	filename )

Write a DataPacket to a CSV file (features followed by label, no header).

Parameters

data	The DataPacket to write.
filename	Output file path.