高效將Matlab Fread轉換為python

Question

我需要將Matlab fread轉換為python，尤其是允許讀取到2d數組中並在讀取時跳過數據。 我提出了以下建議，但我想可能會有更高效和“ pythonic”的方法來做到這一點（我絕不是程序員）。 有什么建議嗎？ 請注意，我無法讀取整個文件，然后對數組進行二次采樣，因為要讀取的文件太大。

def FromFileSkip(fid, count=1, skip=0, dtype=np.float32):
    if np.ndim(count)==0:
        if skip>=0:
            data = np.zeros(count, dtype=dtype)
            k = 0
            while k<count:
                data[k] = np.fromfile(fid, count=1, dtype=dtype)
                fid.seek(skip, 1)
                k +=1
            return data
    elif np.ndim(count)==1:
        if skip>0:
            data = np.zeros(count, dtype=dtype)
            k = 0
            while k<count[1]:
                data[:,k] = np.fromfile(fid, count=count[0], dtype=dtype)
                fid.seek(skip, 1)
                k +=1
            return data
    else:
        raise ValueError('File can be read only into 1d or 2d arrays')

Answer 1

這或多或少是您所擁有的，也許只是一點點清潔。

def fromfileskip(fid,shape,counts,skip,dtype):
  """
  fid    : file object,    Should be open binary file.
  shape  : tuple of ints,  This is the desired shape of each data block.
           For a 2d array with xdim,ydim = 3000,2000 and xdim = fastest 
           dimension, then shape = (2000,3000).
  counts : int, Number of times to read a data block.
  skip   : int, Number of bytes to skip between reads.
  dtype  : np.dtype object, Type of each binary element.
  """
  data = np.zeros((counts,)  + shape)
  for c in xrange(counts):
    block = np.fromfile(fid,dtype=np.float32,count=np.product(shape))
    data[c] = block.reshape(shape)
    fid.seek( fid.tell() + skip)

  return data