# Why matrix return all 'false/0' when there are a lot of similar values between matrix and a column vector?

1 view (last 30 days)
Zhou Ci on 6 Sep 2021
Commented: Zhou Ci on 6 Sep 2021
Hi all,
I have a text file which I imported in MATLAB and now it's in the form of table. It has latitude longitude values which I want to compare with other lat lon values. Data from text file is attached.
Secondly, there is a .hdf file containing lat lon values in the form of matrix. I used interpolation because size of column vector (to be matched) in text file is 7570x1. Initial size of lat lon from hdf image is:
%initial size
lat_hdf = 406x270
lon_hdf = 406x270
%interpolation
lat1 = imresize(lat_hdf,[7570 270],'bilinear');
lon1 = imresize(lon_hdf,[7570 270],'bilinear');
The problem is when I want to match the text file lat lon with lat1 and lon1 it returns a matrix with all 0. Means all false.
However there are alot of values that are similar in these two. I don't know what's the issue.
##### 2 CommentsShowHide 1 older comment
Zhou Ci on 6 Sep 2021
lat_txt = 7570x1 double;
lon_txt = 7570x1 double;
Mat = (lat1==lat_txt);

KSSV on 6 Sep 2021
You should not compare the floting point numbers like that. You should check the difference between them to be less than some tolerance.
tol = 10^-3 ;
Mat = (lat-lat_txt)<=tol;

Walter Roberson on 6 Sep 2021
Mat = (lat1==lat_txt);
The == operator is only bit-for-bit identical values (except that -0 is considered to be equal to 0). But you did
lat1 = imresize(lat_hdf,[7570 270],'bilinear');
so the values in lat1 are computed values -- and they might only be "close", or they might have slightly different round-off in the last bit or two.
Never use == to compare for equality with computed floating point values, except for cases where the value being compared is extracted from the other value. For example it is valid to do
a = rand(1,10);
maxa = max(a);
a == maxa
because maxa will be a bit-for-bit equivalent to some value in a.
It is common to see people do things like
b = 0:.1:1;
b*10 == 3
and be surprised that it is not true for any positions in b. 0.1 is not exactly representable in double precision, and the : operator does not interpret the operation as requesting b = (0:10)/10
Anyhow, you should use ismembertol() instead of ==
Zhou Ci on 6 Sep 2021
Thank you for this explanation.