Convert CSV to TXT without changing data properties

Hello,

I have a CSV file that contains random values, all of them being decimals. Some of them have one decimal value while others have 6 or 7 decimals. When I convert them to txt, values are stored as follows.

How can I convert the csv without changing the actual value and keeping the precise decimal points (comma delimited)?

Thanks in Advance!

[Stephen23]

Change the file extension.

[Jake]

Well. CSV file stores data as shown above. When I choose a cell, the exact value is shown above but the rows store data as powers. (ex: 3.51E+01, while the value is 35.11967)

When I change the extension, txt file shows data as I showed in the question :(

I'm trying MATLAB because I might have to deal with more than one file that has a lot of data.

[Walter Roberson]

Csv does not store as scientific notation and pop up to full decimal places. However, Excel does that, and in MATLAB if you view the variable with the variable browser matlab does that unless you set the variable browser to use g format. Preferences can change the default browser format.

Answer 1

[Walter Roberson]

You do what I already told you to do, and even provided code details for: manipulate the file as text instead of as numeric.

I guarantee you that if you treat the file as numeric that you cannot meet your goal of preserving the original number of decimal places -- not unless each different column has its own fixed number of decimal places. Which I can see from your sample is not the case.

https://www.mathworks.com/matlabcentral/answers/510021-read-and-write-csv-files-without-changing-properties#answer_419381

[Jake]

That question was not entirely related to this but this is absolutely true! I should've been more cautious because your approach is much better than what I was trying.

Thanks a lot, Walter!

#TIL with the experts here is always amazing.

Answer 2

[Piyush Lakhani]

Hi James,

Following way might be work for you.

x=csvread('filename.csv','Row_offset','column_offset'); %if your file has text then give offset for column or row

content=sprintf('%f, %f, %f, %f, %f, %f, %f \n',x); % put %f number of times same as number of columns,
%  Insted of %f you can set the decimals you wants like eg. %5.6f

fId='file.txt'
fId = fopen(fId, 'w') ;
fwrite( fId, content ) ;
fclose( fId ) ;

[Jake]

This seems to work as I desired! Thank you so much.

I'm following this while adding a more effective approach based on Walter Roberson's advice.

Thanks again!

[Walter Roberson]

No this does not preserve decimal places!! %f format defaults to %.6f which gives exactly 6 places after the decimal. Yes you can adjust field by field but you would have to know the field width ahead of time and it would have to be consistent for any one column. However, it isn't. In particular, the original csv has trailing 0 suppression on so for example 1.2300 appears in the file as 1.23 even if other entries in the same column use 4 decimal places. But James wants to preserve decimal places exactly as in the original file, so if he knew that a column had up to 4 decimal places, it would not be acceptable to him to output all the values in that column with exactly 4 decimal places: he wants the entry with two decimal places to show up with 2 decimal places.

The only effective way to handle this is to treat the input as text so that you can handle the varying number of decimal places and you can handle entries that end in 0 that is considered significant and must be preserved.

It is possible that have files in which each entry in a column has a different number of decimal places and trailing 0 that are in the data are important. most files can be handled through a fixed number of decimal places for any given column, but sometimes the number of trailing 0s is important. Not often; thinking of them as being important is much more often a mistake of interpretation.

One case where the number of training zeros is important is in expressing physical constants, in which case the number of digits written gives information about the precision. 2.5400 for a constant gives different information than 2.54 does.

Another case where trailing 0 is important is in some expressions of binary fractions. For example 101.11100 provides information about the number of bits being used in a way that 101.111 would not.

But as I said earlier, more often insisting on preservation of decimal places is a mistake compared to using a fixed number of decimal places (either total or for the fraction)

[Jake]

Walter is right. However, the suggested answer is also right because it works based on what I provided in my question. But yes, It has to be changed as I process files/data with different or random number of decimal points.

What I have came across so far helped me a lot and specially these comments are gold! I'm following Walter's advice, as I mentioned earlier. As Walter suggested, I have to preserve the decimal places and thus, I should process the data as text and not as numeric values.

Thank you again!

[Jake]

Turns out that the code only works for one line(row) of data :)

As Walter mentioned, "you can adjust field by field but you would have to know the field width ahead of time and it would have to be consistent for any one column". In my case, however, I know the field width so I can adjust the code based on that. which is the following part.

content=sprintf('%f, %f, %f, %f, %f, %f, %f \n',x); 

as,

content=sprintf('%.0f, %.0f, %.1f, %.1f, %.1f, %.6f, %.6f \n',x); 

But, it doesn't work as desired when the number of rows are more than one. (I'm attaching the sample file for anyone's reference)

Thank you very much for your time.

[Walter Roberson]

content=sprintf('%.0f, %.0f, %.1f, %.1f, %.1f, %.6f, %.6f \n',x.');

to deal with multiple rows