#include <ml/ml_data/metadata.hpp>

Public Member Functions
bool	has_target () const

bool	is_indexed (size_t column_index) const

bool	is_indexed (const std::string &column_name) const

const std::shared_ptr< ml_data_internal::column_indexer > &	indexer (size_t column_index) const

const std::shared_ptr< ml_data_internal::column_indexer > &	indexer (const std::string &column_name) const

bool	target_is_indexed () const

const std::shared_ptr< ml_data_internal::column_statistics > &	statistics (size_t column_index) const

const std::shared_ptr< ml_data_internal::column_statistics > &	statistics (const std::string &column_name) const

size_t	num_columns (bool include_untranslated_columns=true) const

size_t	num_untranslated_columns () const

bool	has_translated_columns () const

bool	has_untranslated_columns () const

const std::string &	column_name (size_t column_index) const

std::vector< std::string >	column_names () const

size_t	column_index (const std::string &column_name, bool max_on_error=false) const

bool	contains_column (const std::string &column_name) const

const std::string &	target_column_name () const

size_t	column_size (size_t column_index) const

const flex_nd_vec::index_range_type &	nd_column_shape (size_t column_index) const

const flex_nd_vec::index_range_type &	nd_column_shape (const std::string &column_name) const

size_t	target_column_size () const

size_t	index_size (size_t column_index) const

size_t	index_size (const std::string &column_name) const

size_t	global_index_offset (size_t column_index) const

size_t	global_index_offset (const std::string &column_name) const

size_t	target_index_size () const

size_t	num_dimensions () const

bool	is_categorical (size_t column_index) const

bool	is_categorical (const std::string &column_name) const

bool	target_is_categorical () const

bool	is_untranslated_column (size_t column_index) const

bool	is_untranslated_column (const std::string &column_name) const

ml_column_mode	column_mode (size_t column_index) const

ml_column_mode	column_mode (const std::string &column_name) const

ml_column_mode	target_column_mode () const

flex_type_enum	column_type (size_t column_index) const

flex_type_enum	column_type (const std::string &column_name) const

flex_type_enum	target_column_type () const

size_t	get_version () const

std::string	feature_name (size_t column_idx, size_t index, bool quote_string_values=false) const

std::vector< std::string >	feature_names (bool unpack_categorical_columns=true) const

void	save (turi::oarchive &oarc) const

void	load (turi::iarchive &iarc)

void	set_training_index_sizes_to_current_column_sizes ()

ml_data_internal::column_metadata_ptr	get_column_metadata (size_t column_index) const

Detailed Description

ml_metadata provides all the column-wise statistics and column translation information for ml_data.

Definition at line 23 of file metadata.hpp.

Member Function Documentation

◆ column_index()

size_t turi::ml_metadata::column_index	(	const std::string &	_column_name,
		bool	max_on_error = `false`
	)		const

inline

Returns the index of the column matching column_name, or throws an error if it does not exist.

Parameters

column_name	The name of the column.
max_on_error	If true, then size_t(-1) is returned if the column is not present.

Returns the index of the column matching column_name, or throws an error if it does not exist.

Definition at line 27 of file metadata_impl.hpp.

◆ column_mode() [1/2]

ml_column_mode turi::ml_metadata::column_mode ( size_t column_index ) const

inline

Returns the mode of the column. See ml_data_column_modes.hpp for details on the column modes.

Parameters

column_index The index of the column.

Definition at line 232 of file metadata_impl.hpp.

◆ column_mode() [2/2]

ml_column_mode turi::ml_metadata::column_mode ( const std::string & column_name ) const

inline

Returns the mode of the column. See ml_data_column_modes.hpp for details on the column modes.

This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.

Parameters

column_name The name of the column.

Definition at line 243 of file metadata_impl.hpp.

◆ column_name()

const std::string & turi::ml_metadata::column_name ( size_t column_index ) const

inline

Returns the name of the column at column_index.

Parameters

column_index The index of the column.

Returns the size of the columns in the metadata that were present at train time.

Definition at line 125 of file metadata_impl.hpp.

◆ column_names()

std::vector<std::string> turi::ml_metadata::column_names ( ) const

Returns all column names as a vector.

◆ column_size()

size_t turi::ml_metadata::column_size ( size_t column_index ) const

inline

Returns the current index size of the columns in the metadata.

Parameters

column_index The index of the column.

Returns the current index size of the columns in the metadata.

Definition at line 152 of file metadata_impl.hpp.

◆ column_type() [1/2]

flex_type_enum turi::ml_metadata::column_type ( size_t column_index ) const

inline

Returns the size of the columns in the metadata that were present at train time.

Parameters

column_index The index of the column.

Definition at line 260 of file metadata_impl.hpp.

◆ column_type() [2/2]

flex_type_enum turi::ml_metadata::column_type ( const std::string & column_name ) const

inline

Returns the size of the columns in the metadata that were present at train time.

This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.

Parameters

column_name The name of the column.

Definition at line 271 of file metadata_impl.hpp.

◆ contains_column()

bool turi::ml_metadata::contains_column ( const std::string & column_name ) const

inline

Returns true if the metadata contains the given column.

Parameters

column_name The name of the column.

Definition at line 48 of file metadata_impl.hpp.

◆ feature_name()

std::string turi::ml_metadata::feature_name	(	size_t	column_idx,
		size_t	index,
		bool	quote_string_values = `false`
	)		const

Returns the feature name of a specific feature present in the metadata.

Numeric columns are represented by the column name.

Categorical / Categorical List / Dictionary columns are represented by "name[category]".

Vectors are represented by "vector[index]", where index is numerical.

ND vectors are represented by "nd_vector[idx1,idx2]" etc.

Returns: Names of features

◆ feature_names()

std::vector<std::string> turi::ml_metadata::feature_names ( bool unpack_categorical_columns = true ) const

Returns a list of all the feature names present in the metadata.

Numeric columns are represented by the column name.

Categorical / Categorical List / Dictionary columns are represented by "name[category]".

Vectors are represented by "vector[index]", where index is numerical.

ND vectors are represented by "nd_vector[idx1,idx2]" etc.

Returns: Names of features

◆ get_column_metadata()

ml_data_internal::column_metadata_ptr turi::ml_metadata::get_column_metadata ( size_t column_index ) const

inline

Returns a pointer to the internal column metadata of column column_index. Useful for dealing with the column_index

Definition at line 18 of file metadata_impl.hpp.

◆ get_version()

size_t turi::ml_metadata::get_version ( ) const

inline

Serialization version.

Definition at line 323 of file metadata.hpp.

◆ global_index_offset() [1/2]

size_t turi::ml_metadata::global_index_offset ( size_t column_index ) const

inline

Returns the global index offset of the columns in the metadata that were present at train time. This is fixed at setup time; global indices for the column c_idx are in the interval [global_index_offset(c_idx), global_index_offset(c_idx) + index_size(c_idx) - 1]

Parameters

column_index The index of the column.

Returns the global index offset of the columns in the metadata that were present at train time. This is fixed at setup time; global indices for the column c_idx are in the interval [global_index_offset(c_idx), global_index_offset(c_idx) + index_size(c_idx) - 1]

Definition at line 210 of file metadata_impl.hpp.

◆ global_index_offset() [2/2]

size_t turi::ml_metadata::global_index_offset ( const std::string & column_name ) const

inline

Returns the global index offset of the columns in the metadata that were present at train time. This is fixed at setup time; global indices for the column c_idx are in the interval [global_index_offset(c_idx), global_index_offset(c_idx) + index_size(c_idx) - 1]

This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.

Parameters

column_name The name of the column.

Definition at line 223 of file metadata_impl.hpp.

◆ has_target()

bool turi::ml_metadata::has_target ( ) const

inline

Returns true if there is a target column present and false otherwise.

Definition at line 53 of file metadata_impl.hpp.

◆ has_translated_columns()

bool turi::ml_metadata::has_translated_columns ( ) const

inline

Returns true if there are translated columns present, and false otherwise.

Definition at line 102 of file metadata_impl.hpp.

◆ has_untranslated_columns()

bool turi::ml_metadata::has_untranslated_columns ( ) const

inline

Returns true if there are untranslated columns present, and false otherwise.

Definition at line 109 of file metadata_impl.hpp.

◆ index_size() [1/2]

size_t turi::ml_metadata::index_size ( size_t column_index ) const

inline

Returns the index size of the column in the metadata that were present at train time. Index size differs from column size in that column_size may grow on test, but index_size is constant.

Parameters

column_index The index of the column.

Returns the index size of the columns in the metadata that were present at train time. Index size differs from column size in that column_size may grow on test, but index_size is constant.

Definition at line 180 of file metadata_impl.hpp.

◆ index_size() [2/2]

size_t turi::ml_metadata::index_size ( const std::string & column_name ) const

inline

Returns the index size of the column in the metadata that were present at train time. Index size differs from column size in that column_size may grow on test, but index_size is constant.

This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.

Parameters

column_name The name of the column.

Definition at line 192 of file metadata_impl.hpp.

◆ indexer() [1/2]

const std::shared_ptr< ml_data_internal::column_indexer > & turi::ml_metadata::indexer ( size_t column_index ) const

inline

Returns the indexer for a particular column.

Parameters

column_index The index of the column.

Definition at line 61 of file metadata_impl.hpp.

◆ indexer() [2/2]

const std::shared_ptr< ml_data_internal::column_indexer > & turi::ml_metadata::indexer ( const std::string & column_name ) const

inline

Returns the indexer for a particular column.

This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.

Parameters

column_name The name of the column.

Definition at line 66 of file metadata_impl.hpp.

◆ is_categorical() [1/2]

bool turi::ml_metadata::is_categorical ( size_t column_index ) const

inline

Returns true if the underlying type is treated as a categorical variable, and false otherwise.

Parameters

column_index The index of the column.

Definition at line 288 of file metadata_impl.hpp.

◆ is_categorical() [2/2]

bool turi::ml_metadata::is_categorical ( const std::string & column_name ) const

inline

Returns true if the underlying type is treated as a categorical variable, and false otherwise.

This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.

Parameters

column_name The name of the column.

Definition at line 300 of file metadata_impl.hpp.

◆ is_indexed() [1/2]

bool turi::ml_metadata::is_indexed ( size_t column_index ) const

inline

Returns true if the underlying column type is indexed, and false otherwise. This differs form the is_categorical in that dictionaries are not treated as pure categorical variables, as they have values associated with them, but they are indexed.

Parameters

column_index The index of the column.

Definition at line 311 of file metadata_impl.hpp.

◆ is_indexed() [2/2]

bool turi::ml_metadata::is_indexed ( const std::string & column_name ) const

inline

Returns true if the underlying column type is indexed, and false otherwise. This differs form the is_categorical in that dictionaries are not treated as pure categorical variables, as they have values associated with them, but they are indexed.

This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.

Parameters

column_name The name of the column.

Definition at line 325 of file metadata_impl.hpp.

◆ is_untranslated_column() [1/2]

bool turi::ml_metadata::is_untranslated_column ( size_t column_index ) const

inline

Returns true if the underlying column type is untranslated. This means it will only be available as flexible_type later on.

Parameters

column_index The index of the column.

Definition at line 334 of file metadata_impl.hpp.

◆ is_untranslated_column() [2/2]

bool turi::ml_metadata::is_untranslated_column ( const std::string & column_name ) const

inline

Returns true if the underlying column type is untranslated. This means it will only be available as flexible_type later on.

This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.

Parameters

column_name The name of the column.

Definition at line 345 of file metadata_impl.hpp.

◆ load()

void turi::ml_metadata::load ( turi::iarchive & iarc )

Serialization – load.

◆ nd_column_shape() [1/2]

const flex_nd_vec::index_range_type & turi::ml_metadata::nd_column_shape ( size_t column_index ) const

inline

If the type of the column is an ND vector, returns the shape of the nd_vector held by that coulmn.

Parameters

column_index The index of the column.

Returns the current nd column shape of the columns

Definition at line 158 of file metadata_impl.hpp.

◆ nd_column_shape() [2/2]

const flex_nd_vec::index_range_type & turi::ml_metadata::nd_column_shape ( const std::string & column_name ) const

inline

If the type of the column is an ND vector, returns the shape of the nd_vector held by that coulmn.

Parameters

column_index The index of the column.

Returns the current nd column shape of the columns

Definition at line 164 of file metadata_impl.hpp.

◆ num_columns()

size_t turi::ml_metadata::num_columns ( bool include_untranslated_columns = true ) const

inline

Returns the number of columns present.

Definition at line 85 of file metadata_impl.hpp.

◆ num_dimensions()

size_t turi::ml_metadata::num_dimensions ( ) const

inline

Returns the number of distinct dimensions, including all categorical features, etc.

Returns the number of distinct dimensions, including all categorical features.

Definition at line 116 of file metadata_impl.hpp.

◆ num_untranslated_columns()

size_t turi::ml_metadata::num_untranslated_columns ( ) const

inline

Returns the number of untranslated columns present.

Definition at line 94 of file metadata_impl.hpp.

◆ save()

void turi::ml_metadata::save ( turi::oarchive & oarc ) const

Serialization – save.

◆ set_training_index_sizes_to_current_column_sizes()

void turi::ml_metadata::set_training_index_sizes_to_current_column_sizes ( )

Sets the values of all future calls to index_size() to return the column_size values currently present in the indexers. This is done automatically at the end of fill(), but it can be useful if more is done to the indexers after that that is still considered part of training.

◆ statistics() [1/2]

const std::shared_ptr< ml_data_internal::column_statistics > & turi::ml_metadata::statistics ( size_t column_index ) const

inline

Returns the statistics for a particular column.

Parameters

column_index The index of the column.

Definition at line 75 of file metadata_impl.hpp.

◆ statistics() [2/2]

const std::shared_ptr< ml_data_internal::column_statistics > & turi::ml_metadata::statistics ( const std::string & column_name ) const

inline

Returns the statistics for a particular column.

This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.

Parameters

column_name The name of the column.

Definition at line 80 of file metadata_impl.hpp.

◆ target_column_mode()

ml_column_mode turi::ml_metadata::target_column_mode ( ) const

inline

Returns the mode of the target column. See ml_data_column_modes.hpp for details on the column modes.

Definition at line 250 of file metadata_impl.hpp.

◆ target_column_name()

const std::string & turi::ml_metadata::target_column_name ( ) const

inline

Returns the name of the column at column_index.

Returns the size of the columns in the metadata that were present at train time.

Definition at line 133 of file metadata_impl.hpp.

◆ target_column_size()

size_t turi::ml_metadata::target_column_size ( ) const

inline

Returns the current index size of the columns in the metadata.

Definition at line 171 of file metadata_impl.hpp.

◆ target_column_type()

flex_type_enum turi::ml_metadata::target_column_type ( ) const

inline

Returns the mode of the target column. See ml_data_column_modes.hpp for details on the column modes.

Returns the type of the target column.

Definition at line 278 of file metadata_impl.hpp.

◆ target_index_size()

size_t turi::ml_metadata::target_index_size ( ) const

inline

Returns the index size of the columns in the metadata that were present at train time. Index size differs from column size in that column_size may grow on test, but index_size is constant.

Definition at line 200 of file metadata_impl.hpp.

◆ target_is_categorical()

bool turi::ml_metadata::target_is_categorical ( ) const

inline

Returns true if the underlying target type is treated as a categorical variable, and false otherwise.

This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.

Parameters

column_name The name of the column.

Returns true if the underlying type is treated as a categorical variable, and false otherwise.

Definition at line 353 of file metadata_impl.hpp.

◆ target_is_indexed()

bool turi::ml_metadata::target_is_indexed ( ) const

inline

Returns true if the underlying target type is indexed, and false otherwise. This differs form the is_categorical in that dictionaries are not treated as pure categorical variables, as they have values associated with them, but they are indexed.

Returns true if the underlying type is indexed, and false otherwise. This differs form the is_categorical in that dictionaries are not treated as pure categorical variables, as they have values associated with them, but they are indexed.

Definition at line 363 of file metadata_impl.hpp.

The documentation for this class was generated from the following files:

ml/ml_data/metadata.hpp
ml/ml_data/metadata_impl.hpp

Public Member Functions

Detailed Description

Member Function Documentation

◆ column_index()

◆ column_mode() [1/2]

◆ column_mode() [2/2]

◆ column_name()

◆ column_names()

◆ column_size()

◆ column_type() [1/2]

◆ column_type() [2/2]

◆ contains_column()

◆ feature_name()

◆ feature_names()

◆ get_column_metadata()

◆ get_version()

◆ global_index_offset() [1/2]

◆ global_index_offset() [2/2]

◆ has_target()

◆ has_translated_columns()

◆ has_untranslated_columns()

◆ index_size() [1/2]

◆ index_size() [2/2]

◆ indexer() [1/2]

◆ indexer() [2/2]

◆ is_categorical() [1/2]

◆ is_categorical() [2/2]

◆ is_indexed() [1/2]

◆ is_indexed() [2/2]

◆ is_untranslated_column() [1/2]

◆ is_untranslated_column() [2/2]

◆ load()

◆ nd_column_shape() [1/2]

◆ nd_column_shape() [2/2]

◆ num_columns()

◆ num_dimensions()

◆ num_untranslated_columns()

◆ save()

◆ set_training_index_sizes_to_current_column_sizes()

◆ statistics() [1/2]

◆ statistics() [2/2]

◆ target_column_mode()

◆ target_column_name()

◆ target_column_size()

◆ target_column_type()

◆ target_index_size()

◆ target_is_categorical()

◆ target_is_indexed()