How To Get Number Of Columns in Non-Rectangular Array of Data for textscan Function

Hello,

I am currently using the following line of code for extracting data from a csv file, which contains non-rectangular array of data

C = textscan(fileID, repmat('%f',[1 22]), Delimiter, ',', 'EmptyValue', NaN);

What I would love to do is replace that 22 with a variable. Right now I need to check the csv, count the columns (currently, 22) and enter that number manually. Is there a way I could automate that and replace 22 with a variable, eliminating the need to look in the csv file in the future?

Thanks

Answer 1

[Star Strider]

Since your files don’t appear to have header lines, the xlsread or csvread functions might be more appropriate than textscan to import your files.

[EvanThomas]

Don't those require rectangular arrays of data? I went with textscan only due to the non-rectangular nature of the data, as well as plenty of empty fields. But, maybe that was the wrong way to go. Can the above two functions you mention handle that kind of data?

[Star Strider]

My impression is that they should be able to. I’ve not tried them with your files, so I don’t know how the functions would work with them.

Also, textscan may allow you to read beyond the column limits. When I’ve done that, it simply filled the fields with no data as empty cells (although it is best to fill them with NaN values as you have instructed it to do, since you can then use cell2mat to convert them to numeric matrices). So you may be able to use textscan without alteration, then parse the cell arrays it returns afterwards.

It’s probably best to experiment to see what functions work best with your data. I would try them all on a few representative files to see what the results are.

[EvanThomas]

Thanks StarStrider, I had tried csvread already (my go-to import function on most occasions), but it would not work on this data set. Each column is a different length, so it didn't like it. still, I was hoping for a slick way around that. Not really seeing one, likewise with xlsread. textscan is working perfectly right now, I just know sooner, or later, the number of columns will change, and that hardcoded "22" is going to bite me, while I sit and wonder for a while why my script broke because I had since forgotten about it, hehe

[Star Strider]

My pleasure.

I’ve found textscan to be rather robust to (my) stretching it beyond its intended limits. In any event, you can always comment your code (good practice anyway) to check your textscan call if your code starts behaving strangely.