Convert System.Byte[] memory dump to numeric data without fwrite/fread?

Question

Frank Moldenhauer 2019-1-11

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/439391-convert-system-byte-memory-dump-to-numeric-data-without-fwrite-fread

回答： Steven Remington 2019-1-15

Hi,

from a .NET application I get System.Byte[] array data that contain memory dumps with numerical data of several types (int16, int32, ieee float) inside. I want to process these data in Matlab.

>> whos data
  Name      Size            Bytes  Class            Attributes
  data      1x1                 8  System.Byte[]              
>> data
data = 
  Byte[] with properties:
            Length: 65536
        LongLength: 65536
              Rank: 1
          SyncRoot: [1×1 System.Byte[]]
        IsReadOnly: 0
       IsFixedSize: 1
    IsSynchronized: 0

The following code does the job for single precision float data, it returns a 16384 array fdata made from the 4 x 16384 = 65536 bytes (other data types similar):

% convert to matlab array, see
% https://de.mathworks.com/matlabcentral/answers/46827-fast-transfer-of-net-array-to-matlab-array
 bdata = data.uint8;
% write memory dump to file
fid = fopen('tmp.bin', 'w');
fwrite(fid, bdata); 
fclose(fid);
% read file e.g. as ieee single precision float
fid = fopen('tmp.bin', 'r');
fdata = fread(fid, inf, 'single');
fclose(fid);

Is there a way to to this faster in memory without using a temporary file?

Thanks & regards,

Frank

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Frank Moldenhauer 2019-1-14

Meanwhile solved using a C MEX function, but may be there is a trick to do this directly in Matlab?

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Steven Remington 2019-1-15

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/439391-convert-system-byte-memory-dump-to-numeric-data-without-fwrite-fread#answer_356533

There are a couple options. MATLAB has a typecasting function which allows you to convert data types.

More information about that process can be found at the follow doc page:

https://www.mathworks.com/help/matlab/ref/typecast.html

With that you should be able to select columns from your data in order to convert to the correct data type.

Another option would be to use "textscan" with a format to bulk read data of different data types from the file if you have a complex formatted file. Some examples of this is shown below:

https://www.mathworks.com/help/matlab/ref/textscan.html