Weird .csv file plot problem

Hey, When I plot from a csv file like so:

 K=csvread(b);
 plot(K);

I get the correct outcome but the x axis is not the first column as I would like it to be. So I tried this and it didn't work for some reason:

M=csvread(a);
x=M(:,1);
y=M(:,2);
l=M(:,3);
figure (1)
plot(x,y,'g');
hold on 
plot(x,l,'r');

I do get a graph with the correct x axis, but the graph is missing many points unlike the previous way where it had all 700000 points. Do I have too many columns for the second way to work? I am using MATLAB 2013b. Thanks

[Image Analyst]

You forgot to attach the files you're using for the badly-named a and b. Please attach them so we can try it and see what the difference is.

[Haris Kioleidis]

Ok sorry, I thought it could be some known limitation/problem. Here is the file. A and B is practically the same and the problem happens with all my .csv files.

Answer 1

[dpb]

The first syntax per the documentation plots all columns of K versus ordinal number

You're on the right track with the second; you don't need the temporaries nor two calls to plot() however...

M=csvread(...;
plot(M(:,1), M(:,2:end))

will plot all columns from 2nd on vs the first; plot will use all points other than any which are NaN that are silently ignored; probably the difference in what you think is number of plotted points is that in the first case the x-axis scale will go to cover 1:N whereas in the second it will be scaled to the value of the elements in the first column which one presumes are not 1:length(M).

What does

sum(isfinite(M))

return at the command line??

Beyond that we'd have to see the data to be able to say anything more...

NB: There's little real point in plotting 700K points, even the most high resolution and largest monitor won't have more than a few K pixels so there's nowhere near enough to display all the points without serious aliasing. Decimating or decimating with bounding or the like would be a logical thing to do here...

ADDENDUM

The problem is that you didn't save enough precision in the input file to record the data accurately.

>> arrayfun(@(i) length(unique(m(:,i))),[1:3],'uniform',1)
ans =
      5140          60          35
>>

Whether the 2nd/3rd columns are supposed to be more than such a few different values I don't know, but clearly the first column was supposed to be a time vector and it simply isn't precise enough for the size of the data.

Don't know who wrote the file, but there's where the basic problem comes from, not from plot()

[Haris Kioleidis]

I get this at the command line: ans =

      700012      700012      700012

Also, I uploaded the file. When I plot with the first syntax I get decent enough resolution and can work some things out but the x axis, as you said, is the ordinal number which I can't use. Thanks

[dpb]

>> x=linspace(3,1E6,700000).';
>> y=randn(siz e(x));
>> m=[x y y+10];
>> hL=plot(m(:,1), m(:,2:end));
>> x=get(hL,'XData');
>> cellfun(@length,x)
ans =
      700000
      700000
>>

plot() put all 700K on there...what does that do on your system?

[Haris Kioleidis]

Here are the two graphs of the same file. Full is with

L=csvread(a);
plot(L);

and missing with

L=csvread(a);
plot(L(:,1), L(:,2:end))

[dpb]

And I downloaded you file and plotted it and did the same test...

subplot(2,1,1)
hL1=plot(m);
x1=get(hL1,'XData');
cellfun(@length,x1)
ans =
    700012
    700012
    700012
subplot(2,1,2)
hL2=plot(m(:,1),m(:,2:end));
x2=get(hL2,'XData');
cellfun(@length,x2)
ans =
    700012
    700012
>>

which shows it plotted all points either way...the appearance is a result of the fact you didn't save the data with sufficient precision for the first column--

u=unique(m(:,1));
numel(u)
ans =
      5140

length(m(:,1))/numel(u)
ans =
136.1891

so there are on average 136 values plotted on top of each other with m(:,1) as the x vector.

[Haris Kioleidis]

Yeah, got it. Thanks a lot for the help.