Separate CSV to stucture array by a specified field

Greetings!

I have a CSV file with the following contents:

"x", "y", "ZLEVEL"
15566.071598095354, 654.75138415898618, A
15730.023876579013, 660.38875580774038, A
15772.038140720921, 665.0524486236709, A
15861.67702561093, 669.09878141965692, B
15964.77015716501, 675.05706329613804, B
16054.020545106847, 683.1063847011992, B
16118.583970776759, 690.87616533615892, B
16231.136948002502, 695.99923622422557, C
16344.66294729244, 700.25827295850286, C
16400.372270382126, 705.20613829250306, C
16486.242397353868, 709.25627232144143, C
16564.589258300024, 715.96303745505884, C
16624.046537611983, 723.13271402537612, C
16752.728644676041, 727.80842778700935, C
16860.248868173338, 732.05753736629867, C
16906.759450074052, 736.81366143388732, C

I intent to create a 3 by 1 struct array with two fields XYData (matrix) and ZLEVEL (string), which holds the CSV data separated by ZLEVEL field and XYData matrix holds X and Y in its original CSV order.

Can you suggest me a nifty way (using built-in MATLAB functions) to do this please?

Thank you.

Answer 1

[Walter Roberson]

If you have R2014b or later you can use tableread() and then group the entries by ZLevel.

But that's because you asked for nifty, not for efficient.

[Andrey Kazak]

I was also thinking about using Table objects to solve this task. Can you share a snippet for grouping the entries of a table by ZLEVEL please?

[Andrey Kazak]

Here is how one could achieve the goal. Let's assume that the table has name Main and we want to filter x and y values which have ZLEVEL equal to C:

 Main{Main.ZLEVEL == 'C',{'x', 'y'}

And that's it! Very simple and convenient with help of Table object. As for data processing speed I found that even a large CSV file (100 MB) is being imported (at least on my average PC) fairly quickly.

Answer 2

[Guillaume]

A table probably makes the reading of the file easier but I'm not sure it helps much after that. I don't think unstack is going to work for this.

It's still somewhat simple to do:

%read data with textscan
fid = fopen('somefile.txt', 'rt');
c = textscan(fid, '%f, %f, %s', 'HeaderLines', 1);  
fclose(fid);
%or as a table
t = readtable('somefile.txt');

%partition the data, 
%using a table here. replace t.x, t.y, t.ZLEVEL, by c{1}, c{2}, c{3} if using textscan
[levels, ~, pos] = unique(t.ZLEVEL);  %get partitioning
xygrouped = accumarray(pos, 1:numel(pos), [], @(v) {[t.x(v), t.y(v)]}) %defer to accumarray to do the grouping
sarray = struct('XYData', xygrouped, 'ZLEVEL', levels)  %and convert to struct

[Andrey Kazak]

Table objects worked very fine for my case.

1) convert ZLEVEL to categorical array:

Main.ZLEVEL = categorical(Main.ZLEVEL);

2) get categories of ZLEVEL

ZCats = categories(Main.ZLEVEL)

3) loop through ZCats and do what you need.

[Guillaume]

I thought your objective was to create a structure. Note that my example does use a table object.

Your step 1) and 2) are the same as my unique line.

Your step 3) is the same as my accumarray line.