comparing two double variables in arduino

Question

So, I have two double variables, and I want to compare them till 3 decimal places. So, for variables (for example):

double x = 0.695999;
double y = 0.695111;

if I check for (x==y), it should return true (since both are equal till 3 decimal places). Thanks!

How does the compiler and even us know that is your intention? The compiler/us are not physic — Ed Heal, Mar 03 '13 at 07:55
may be, if C has a function like: compare(double v1, double v2, int precision)? Not a psych, but we passing the information to the method parameters? — Darth.Vader, Mar 03 '13 at 07:56

score 7 · Accepted Answer · 2013-03-03T11:00:54.123

7

You can (ab)use integer comparison and truncation:

int is_equal_3decplaces(double a, double b) {
    long long ai = a * 1000;
    long long bi = b * 1000;
    return ai == bi;
}

As @DavidRF's benchmarks have shown it, this solution offers a slight (~40%) improvement in performance compared to calculating the absolute values.

edited Mar 03 '13 at 11:00

answered Mar 03 '13 at 07:54

1

That's an interesting hack! :) – Darth.Vader Mar 03 '13 at 07:57
@DavidRF Yap, I think it may be even better than calling `abs` then comparing the epsilon (not sure, though - it would be interesting to run some benchmarks). – Mar 03 '13 at 09:39
Executed 100000000 times, abs takes 0m9.239s while your solution takes 0m15.394s – David Ranieri Mar 03 '13 at 09:59
1

ops, I was not assigning values to x and y, executed 100000000 times, abs (function used by user1952500) takes 0m1.106s while your solution takes 0m0.742s – David Ranieri Mar 03 '13 at 10:35
@H2CO3, `(DOUBLE_MAX * 1000)` can not overflow `LLONG_MAX`, true? – David Ranieri Mar 03 '13 at 11:35

score 4 · Answer 2 · edited Mar 03 '13 at 08:05

4

You should actually check for the difference being less than a small delta which is pre-configurable, unless you always want it to be checked for three decimal places. For example:

#define epsilon ((double)0.000999)
bool is_approximately_equal(double x, double y)
{
        return (abs(x - y) < epsilon);
}

Note that abs on C++ has the version of abs for double. In C you'll have to do something with -ve values.

edited Mar 03 '13 at 08:05

mttrb

8,297
3
35
57

answered Mar 03 '13 at 07:59

user1952500

6,611
3
24
37

1

Changed. Thanks! Thought the other was more instructive but anyways doesn't change much. – user1952500 Mar 03 '13 at 08:01
@mttrb, right. I mixed c++ and c (std::cmath has the double and float versions). Added a comment on that. – user1952500 Mar 03 '13 at 08:02
I am actually using arduino - I am not sure if it is based on C or C++ and if abs will work for me on arduino... if it doesn't I will just write my own abs() function. – Darth.Vader Mar 03 '13 at 08:07
It looks like `abs` is defined as `#define abs(x) ((x)>0?(x):-(x))` in `Arduino.h` so it should work on the doubles and integers. – mttrb Mar 03 '13 at 08:37

comparing two double variables in arduino

2 Answers2