npy.h File Reference

Definitions for reading and writing NPY and NPZ files The libnpy library provides a means to read and write NPY and NPY files from C++. methods for reading and writing the numpy lib (NPY) format. The implementation is based upon the description available at: https://docs.scipy.org/doc/numpy/reference/generated/numpy.lib.format.html The NPZ implementation draws heavily from the PKZIP Application note: https://pkware.cachefly.net/webdocs/casestudies/APPNOTE.TXT. More...

#include <algorithm>
#include <cassert>
#include <cstdint>
#include <filesystem>
#include <fstream>
#include <map>
#include <sstream>
#include <stdexcept>
#include <string>
#include <vector>

Include dependency graph for npy.h:

Go to the source code of this file.

Classes
struct	npy::header_info
	Class representing the header info for an NPY file. More...
struct	npy::file_entry
	Struct representing a file in the NPZ archive. More...
class	npy::npzstringwriter
	Class which handles writing of an NPZ to an in-memory string stream. More...
class	npy::npzfilewriter
	Class which handles writing of an NPZ archive to disk. More...
class	npy::npzstringreader
	Class handling reading of an NPZ from an in-memory string stream. More...
class	npy::npzfilereader
	Class handling reading of an NPZ from a file on disk. More...
class	npy::tensor< T >
	The default tensor class. More...

Namespaces
namespace	npy

Macros
#define	NPY_VERSION_MAJOR   2
#define	NPY_VERSION_MINOR   0
#define	NPY_VERSION_PATCH   0
#define	NPY_VERSION_STRING   "2.0.0"

Enumerations
enum class	npy::endian_t : char { npy::NATIVE , npy::BIG , npy::LITTLE }
	Enumeration which represents a type of endianness. More...
enum class	npy::data_type_t : char {   npy::INT8 , npy::UINT8 , npy::INT16 , npy::UINT16 ,   npy::INT32 , npy::UINT32 , npy::INT64 , npy::UINT64 ,   npy::FLOAT32 , npy::FLOAT64 , npy::COMPLEX64 , npy::COMPLEX128 ,   npy::UNICODE_STRING }
	This enum represents the different types of tensor data that can be stored. More...
enum class	npy::compression_method_t : std::uint16_t { npy::STORED = 0 , npy::DEFLATED = 8 }
	Enumeration indicating the compression method to use for data in the NPZ archive. More...

Functions
endian_t	npy::native_endian ()
	This function will return the endianness of the current hardware.
const std::string &	npy::to_dtype (data_type_t dtype, endian_t endian=endian_t::NATIVE)
	Convert a data type and endianness to a NPY dtype string.
const std::pair< data_type_t, endian_t > &	npy::from_dtype (const std::string &dtype)
std::ostream &	npy::operator<< (std::ostream &os, const endian_t &obj)
std::ostream &	npy::operator<< (std::ostream &os, const data_type_t &obj)
template<typename CHAR >
void	npy::write_npy_header (std::basic_ostream< CHAR > &output, const std::string &dtype, bool fortran_order, const std::vector< size_t > &shape)
	Writes an NPY header to the provided stream.
template<typename T , typename CHAR >
void	npy::write_values (std::basic_ostream< CHAR > &output, const T *data_ptr, size_t num_elements, endian_t endianness)
	Write values to the provided stream.
template<typename T , typename CHAR >
void	npy::save (std::basic_ostream< CHAR > &output, const T &tensor, endian_t endianness=npy::endian_t::NATIVE)
	Saves a tensor to the provided stream.
template<typename T , template< typename > class TENSOR, typename CHAR >
void	npy::save (std::basic_ostream< CHAR > &output, const TENSOR< T > &tensor, endian_t endianness=npy::endian_t::NATIVE)
	Saves a tensor to the provided stream.
template<typename T >
void	npy::save (const std::string &path, T &tensor, endian_t endianness=npy::endian_t::NATIVE)
	Saves a tensor to the provided location on disk.
template<typename T , template< typename > class TENSOR>
void	npy::save (const std::string &path, T &tensor, endian_t endianness=npy::endian_t::NATIVE)
	Saves a tensor to the provided location on disk.
template<typename CHAR >
header_info	npy::read_npy_header (std::basic_istream< CHAR > &input)
	Read an NPY header from the provided stream.
template<typename T , typename CHAR >
void	npy::read_values (std::basic_istream< CHAR > &input, T *data_ptr, size_t num_elements, const header_info &info)
	Read values from the provided stream.
template<typename T , typename CHAR >
T	npy::load (std::basic_istream< CHAR > &input)
template<typename T >
T	npy::load (const std::string &path)
	Loads a tensor in NPY format from the specified location on the disk. The type of the tensor must match the data to be read.
template<typename T , template< typename > class TENSOR>
TENSOR< T >	npy::load (const std::string &path)
	Loads a tensor in NPY format from the specified location on the disk. The type of the tensor must match the data to be read.
template<typename CHAR >
header_info	npy::peek (std::basic_istream< CHAR > &input)
	Return the header information for an NPY file.
header_info	npy::peek (const std::string &path)
	Return the header information for an NPY file.

Variables
const int	STATIC_HEADER_LENGTH = 10

Detailed Description

Definitions for reading and writing NPY and NPZ files The libnpy library provides a means to read and write NPY and NPY files from C++. methods for reading and writing the numpy lib (NPY) format. The implementation is based upon the description available at: https://docs.scipy.org/doc/numpy/reference/generated/numpy.lib.format.html The NPZ implementation draws heavily from the PKZIP Application note: https://pkware.cachefly.net/webdocs/casestudies/APPNOTE.TXT.

---

Copyright (C) 2021 Matthew Johnson

For conditions of distribution and use, see copyright notice in LICENSE

---

Macro Definition Documentation

NPY_VERSION_MAJOR

#define NPY_VERSION_MAJOR   2

NPY_VERSION_MINOR

#define NPY_VERSION_MINOR   0

NPY_VERSION_PATCH

#define NPY_VERSION_PATCH   0

NPY_VERSION_STRING

#define NPY_VERSION_STRING   "2.0.0"

Variable Documentation

STATIC_HEADER_LENGTH

const int STATIC_HEADER_LENGTH = 10