csvwrite is storing wrong number

I'm using "csvwrite" command for storing a 2d matrix with size 104937*1810. Is there any limitation on this command or csv file? The produced file is 670 MB but in each stored row there are exactly 5 numbers which are not the same as matrix in the matlab. All numbers are int32 and the wrong numbers are in different colomns in each row.

[Walter Roberson]

Would it be possible to xlswrite() or writetable() to a .xlsx file and attach that for us to test with?

Also which release are you using?

Answer 1

[dpb]

>> tstary=int32(randi(2^31,104937,1810));
>> whos tstary
  Name             Size                  Bytes  Class    Attributes
  tstary      104937x1810            759743880  int32              

>> tstary(1:3,1:5)
ans =
  3×10 int32 matrix
   1186639717    717774140   2035668545    471416973    522550194   
   2056308186    151423171    131112521   1937130594   1414722413   
   1917344954   1185569608    203068410    614639920    243357862   
>> csvwrite('tstary.csv',tstary)

Content of tstary.csv from file viewer...

1.1866e+09,7.1777e+08,2.0357e+09,4.7142e+08,5.2255e+08,1.1246e+09,1.5343e+09,3.2607e+08, ...
2.0563e+09,1.5142e+08,1.3111e+08,1.9371e+09,1.4147e+09,7.5605e+08,7.9947e+08,6.188e+08,  ...
1.9173e+09,1.1856e+09,2.0307e+08,6.1464e+08,2.4336e+08,9.5385e+08,1.0362e+09,2.0352e+09, ...
...

The content is exactly what the documentaion for csvwrite says it will be--the data in %g format with a maximum of 5 digits of precision.

The help there indicates to use dlmwrite with optional precision argument, but for such a large file, using a text format is absurd way to store/move/transport data. Use stream or some other form instead.

But, csvread is working as expected.

[Walter Roberson]

Ah, I had understood the question a very different way, because of the discussion of "exactly five" that are wrong and how the wrong numbers are in different columns in different rows.

[dpb]

I misunderstood initially, too, I think, but decided that probably it was a case of the symptom not being recognized as to cause that produced a mistaken interpretation of the cause of the difference and that led to asking the wrong question--or wording the question to be about the wrong problem.

[Abolfazl Taghribi]

I did not mention the "maximum of 5 digits of precision" limitation for csvwrite before. This was the problem. I should use dlmwrite instead. Thanks dpl .