sscanf读取的数据不正确 [英] inaccurate reading of data by sscanf

问题描述

在Matlab中，可以使用str2double或str2num和cellfun的组合将单元格数组中的数字的字符串表示形式转换为双精度数组中的数字表示形式. 最近的帖子建议结合使用str2mat和sscanf来加快转换速度.但是，这种方法对于某些数据集似乎无效，例如:

In Matlab, conversion of string representations of numbers in a cell array into numeric representations in a double array can be performed with str2double or a combination of str2num and cellfun. A recent post suggested using a combination of str2mat and sscanf to achieve faster conversion. However, this approach seems to fail for certain data sets, such as the following:

x={'98.78743';'99.39717';'99.93578';'100.40125';'100.79166';'101.10525';'101.34037';'101.49553';'101.56939';'101.56072';'101.4685';'101.29184';'101.03002';'100.68249';'100.24887';'99.72897';'99.12274';'98.43036';'97.65215';'96.78864';'95.84054'};
y=sscanf(str2mat(x).','%f');
[str2double(x),y,str2double(x)-y]
ans =

    98.7874   98.7874         0
    99.3972   99.3972         0
    99.9358   99.9358         0
   100.4013  100.4013   -0.0000
   100.7917    0.7917  100.0000
   101.1052    0.1053  101.0000
   101.3404    0.3404  101.0000
   101.4955    0.4955  101.0000
   101.5694    0.5694  101.0000
   101.5607    0.5607  101.0000
   101.4685    0.4685  101.0000
   101.2918  101.2918   -0.0000
   101.0300    0.0300  101.0000
   100.6825    0.6825  100.0000
   100.2489    0.2489  100.0000
    99.7290    0.7290   99.0000
    99.1227   99.1227         0
    98.4304   98.4304         0
    97.6522   97.6522         0
    96.7886   96.7886         0
    95.8405   95.8405         0

以上代码块的最后一行显示了原始单元格数组的双重表示形式，sscanf转换后的形式以及两者之间的区别.为了忠实地进行转换，第三列中的所有条目都将恰好为0.许多值在转换时不正确地消除了最高有效位.这种错误行为的基础是什么，可以对语句做些微修改以产生准确的转换吗?通常，我尝试转换的数字具有不同的长度(小数点前和小数点后两者)，因此，任何依赖单元格数组中具有相同长度的所有数字的方法都将失败.最后，我的MATLAB版本是2013年前的版本，因此我不能依赖上述帖子中建议的strjoin方法.

The final line in the above code block shows the double representation of the original cell array, the sscanf-converted form, and the difference between the two. For a faithful conversion, all entries in the third column would be precisely 0. Many of the values are converted with improper elimination of the most significant digits. What is the basis for this erroneous behavior and can the statement be modified slightly to produce an accurate conversion? In general, the numbers which I am trying to convert are of different lengths (both pre- and post-decimal point), so any approach that relies on all numbers in the cell array having the same length are bound to fail. Finally, my version of MATLAB is pre-2013, so I cannot rely on the strjoin approach that was also suggested in the aforementioned post.

推荐答案

尝试以下方法:

y=sscanf(str2mat(x).','%9f')

9实际上对应于str2mat(x).'

>> str2mat(x).'

ans =

999111111111111999999
899000000000000998765
...001111111100......
739............714678
893471345542062223584
775090496669384820280
417115059081028973165
378262353758048746544
   5657392 4297      

因此，sscanf将搜索9个浮点，并在找到空格时停止.您可以根据数据定义9个值.

Hence, sscanf will search for 9 floats and stop when it finds a space. You could define the 9 values depending on your data.

一个例子:

x={'98.7';'99.397';'99.93578';'100.4';'100.79166';'101.10'};
>> str2mat(x).'

ans =

999111
899000
...001
739...
 93471
 75 90
  7 1 
  8 6 
    6 

最大长度仍然是9.答案仍然正确.

Here the maximum length is still 9. And the answer is still correct.

y=sscanf(str2mat(x).','%9f')

>> y =

   98.7000
   99.3970
   99.9358
  100.4000
  100.7917
  101.1000

要确定%'number'f的随机长度，请使用:

For determining the random length for the %'number'f, use:

size(str2mat(x).',1)

另一个技巧:

xMaxLength = num2str(size(str2mat(x).',1));
y = sscanf(str2mat(x).',['%' xMaxLength 'f']);

为了只计算一次str2mat:

xx = str2mat(x).';
xxMaxLength = num2str(size(xx,1));
y = sscanf(xx,['%' xxMaxLength 'f']);

这篇关于sscanf读取的数据不正确的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！