python npz转csv

Posted 2023-04-23 myrj

import numpy as np
import pandas as pd
file = np.load(\'x_train.npz\',allow_pickle=True)
#查看其中的npy文件名，其实npz文件就是由多个npy文件压缩而成的，winrar即可解压缩
for ii in file:
    print(ii)
##indices
##indptr
##format
##shape
##data
##    file.files
##[\'indices\', \'indptr\', \'format\', \'shape\', \'data\']
#解压缩后：
npfile=np.load("C:\\python37\\x_train\\data.npy")

npfile=np.load("C:\\\\python37\\\\x_train\\\\data.npy")
npp=pd.DataFrame(data=npfile)
npp.to_csv("data.csv")

columns = file[\'indices\']
values = file[\'data\']
np.savetxt("mm.csv",file,delimiter=\',\',fmt=\'%f\')

##for key,value in file.items():
##    print(key,value)
    
    #np.savetxt("nn"+key+".csv",values)
##
##from pandas import DataFrame
###pd.DataFrame(values,columns=columns)
##a=DataFrame(values,columns=columns)
###a.to_excel(\'国民经济核算季度数据.xlsx\',index=None)
##a.to_csv(\'test.csv\',index=None)

 

python 逐步写入`npz`文件以节省内存

import zipfile
import io
import typing

import numpy as np


class IncrementalNpzWriter:
    """
    Write data to npz file incrementally rather than compute all and write
    once, as in ``np.save``. This class can be used with ``contextlib.closing``
    to ensure closed after usage.
    """
    def __init__(self, tofile: str, mode: str = 'x'):
        """
        :param tofile: the ``npz`` file to write
        :param mode: must be one of {'x', 'w', 'a'}. See
               https://docs.python.org/3/library/zipfile.html for detail
        """
        assert mode in 'xwa', str(mode)
        self.tofile = zipfile.ZipFile(tofile, mode=mode,
                                      compression=zipfile.ZIP_DEFLATED)

    def write(self, key: str, data: typing.Union[np.ndarray, bytes],
              is_npy_data: bool = True) -> None:
        """
        :param key: the name of data to write
        :param data: the data
        :param is_npy_data: if ``True``, ".npz" will be appended to ``key``,
               and ``data`` will be serialized by ``np.save``;
               otherwise, ``key`` will be treated as is, and ``data`` will be
               treated as binary data
        :raise KeyError: if the transformed ``key`` (as per ``is_npy_data``)
               already exists in ``self.tofile``
        """
        if key in self.tofile.namelist():
            raise KeyError('Duplicate key "{}" already exists in "{}"'
                           .format(key, self.tofile.filename))
        self.update(key, data, is_npy_data=is_npy_data)

    def update(self, key: str, data: typing.Union[np.ndarray, bytes],
               is_npy_data: bool = True) -> None:
        """
        Same as ``self.write`` but overwrite existing data of name ``key``.

        :param key: the name of data to write
        :param data: the data
        :param is_npy_data: if ``True``, ".npz" will be appended to ``key``,
               and ``data`` will be serialized by ``np.save``;
               otherwise, ``key`` will be treated as is, and ``data`` will be
               treated as binary data
        """
        kwargs = {
            'mode': 'w',
            'force_zip64': True,
        }
        if is_npy_data:
            key += '.npy'
            with io.BytesIO() as cbuf:
                np.save(cbuf, data)
                cbuf.seek(0)
                with self.tofile.open(key, **kwargs) as outfile:
                    shutil.copyfileobj(cbuf, outfile)
        else:
            with self.tofile.open(key, **kwargs) as outfile:
                outfile.write(data)

    def close(self):
        if self.tofile is not None:
            self.tofile.close()
            self.tofile = None