Using textscan to read header with mixed formats

20 次查看（过去 30 天）

显示更早的评论

Henk-Jan Ramaker 2021-12-15

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1611510-using-textscan-to-read-header-with-mixed-formats

评论： dpb 2021-12-15

testFile.csv

在 MATLAB Online 中打开

I have a *.csv file I want to import (see attachment). I read part of testFile.csv as follows:

fileID = fopen(file);
X = textscan(fileID,'%f %f %f %f','Delimiter',',','Headerlines',22)
fclose(fileID);
X = cell2mat(X);
DATA.waveAxis = X(:,1)';
DATA.absorbanceSpectrum = X(:,2)';
DATA.backgroundReference = X(:,3)';
DATA.sampleSignal =  X(:,4)';

That works well, but I haven't retreived all the information from testFile.csv yet.

That is, I would also like to add the first 20 rows of the testFile.csv to my structure "DATA". For instance, I want to add header information to DATA such that it looks like "DATA.method = Column1" (first header line) or "DATA.serialNumber = 5490232" (10th header line).

However, the first 20 header lines have different formats, so I find it very difficult to write a piece of needy & speedy code to do the job. Therefore, any help is greatly appreciated!

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

dpb 2021-12-15

Wouldn't be too bad to parse and create a structure with dynamic fields names the specific file -- generically, this could be a pain given the what appear to be superfluous fields in the header data -- unless there's some known key about which has how many fields.

For example, starts off with a single piece of data in the second field for Method, Date-Time, Version, Temp, ... until get to 'Shift Vector Coefficients' which also appears to have an array of three doubles in the second data field -- except they're separated by the same delimiter as is used in the other records so there are instead five delimiters instead of only three. How to handle that will simply have to have a look up of what to do when get a given record.

Then, to add confusion, "Section 1,," only has two and no leading name field -- are there other sections in a real file, maybe, trailing after the first. But, there doesn't seem to be any indicator by which to determine how long a section of data might be...

You'll just have to have logic to treat the records by type...if it is a fixed header where it always has the same header information, then it's tedious to do once, but is fairly straightforward. If you have to be able to recognize any number of header records that could have any name string, that'll be tougher to deal with.

请先登录，再进行评论。

请先登录，再回答此问题。