textscan: strings and variables

Hi,

I have some csv file which contains numbers and strings (example below)

The column headers are in seperate file from the data, but can be merged to the same one.

I'd like to import the data with textscan (as I have old version of MATLAB-R14) and assign it to some variable (which is the column name) for further analysis.

How can I do that?

For numbers I use something like

eval (column name = [column data])

For time: as it with same length - I also don't have any issue (as I can split the data into 3 coulns and later handle it with time related functions).

But what about other ones where the string values are with variable length (like RATE or STATUS columns as in the example)?

[Stephen23]

Much more robust approaches:

• use READTABLE (then the variable/column names can be almost anything)
• use a structure (then the fieldnames can be any valid MATLAB name)

The current approach is inefficient, liable to bugs, turns of many code-helper tools, and is fragile: consider what would happen if the header text is empty or "QUIT" or "1" or "A-B".

Such fragile approaches are best avoided.

[michael]

@Stephen is READTABLE available in R14?

I need the solution to be fast enought as my tables are with many colkumns and hundred of thouthand of rows.

[Walter Roberson]

https://blogs.mathworks.com/loren/2005/12/13/use-dynamic-field-references/

Dynamic field names existed in your release (R13 introduction), and for earlier releases use setfield()

Answer 1

[Stephen23]

Perhaps something like this, a more robust approach using a structure (this still requires that the headers are valid MATLAB names):

opt = {'Delimiter',','};
fid = fopen('test.txt','rt');
hdr = fgetl(fid);
tmp = textscan(fid,'%s%f%f%s%f%s',opt{:});
fclose(fid);
hdr = regexp(hdr,'[^,]+','match');
out = cell2struct(tmp,hdr,2)

out = struct with fields:
      TIME: {3×1 cell}
       LAT: [3×1 double]
       LON: [3×1 double]
    STATUS: {3×1 cell}
       ALT: [3×1 double]
      RATE: {3×1 cell}

out.TIME

ans = 3×1 cell array
    {'06:02:00'}
    {'06:02:01'}
    {'06:02:02'}

out.LAT

ans = 3×1

30.0010 30.0020 30.0040

out.LON

ans = 3×1

31.0020 31.0030 31.0040

[Walter Roberson]

after getting hdr, pass it through genvarname(), which will convert invalid names (such as containing spaces or punctuation) into valid names, and will further modify names to ensure that they are unique (in case two columns has the same name, one will be given a suffix)