SAS中DO WHILE与DO UNTIL循环为何会在应终止后多执行一次循环体？

Hey there! This is a super common gotcha with floating-point number precision—let’s break down exactly what’s happening with both of your loops.

Problem 1: DO WHILE Loop Outputting 1.1 #

First, let’s recap your code:

data do_while; 
    a = 0; 
    do while (a<1); 
        a = a+0.1; 
        output; 
    end; 
run;

You’d expect the loop to stop when a hits 1.0, right? Here’s why it doesn’t:

• The value 0.1 can’t be represented exactly as a binary floating-point number (the format SAS uses for numeric variables). It’s an infinite repeating fraction in binary, so every time you add 0.1, you’re actually adding a tiny approximation of 0.1.
• After 10 additions, instead of getting exactly 1.0, you end up with a value very slightly less than 1.0 (something like 0.9999999999999999).
• Since this value still satisfies a<1, the loop runs one more time: a becomes 0.9999999999999999 + 0.1 = 1.0999999999999999, which SAS displays as 1.1 in your dataset. Then the loop checks a<1 again, which is false, so it terminates.

Problem 2: DO UNTIL Loop Also Outputting 1.1 #

Your DO UNTIL code:

data do_until; 
    b=0; 
    do until (b>=1); 
        b=b+0.1; 
        output; 
    end; 
run;

Your logic about DO UNTIL (execute first, check condition second) is totally correct—here’s why you still get 1.1:

• Again, floating-point precision is the issue. After 9 additions, b isn’t exactly 0.9—it’s a tiny approximation slightly less than 0.9.
• Adding 0.1 gives you a value still slightly less than 1.0 (like 0.9999999999999999). Since b>=1 is still false, the loop runs again: b becomes 1.0999999999999999 (displayed as 1.1), you output it, then the condition b>=1 finally evaluates to true, and the loop stops.

How to Fix This #

To avoid floating-point precision headaches in loops like this, try one of these approaches:

1. Loop with integers instead: Calculate your decimal value using an integer counter—this eliminates approximation errors entirely:

   data do_while_fixed;
       do i = 1 to 10;
           a = 0.1 * i;
           output;
       end;
   run;
   

2. Use rounding in your condition: If you need to stick with decimal increments, round the value when checking the condition to account for tiny precision gaps:

   data do_while_fixed;
       a = 0;
       do while (round(a, 0.1) < 1);
           a = a + 0.1;
           output;
       end;
   run;
   

You can verify the exact stored values by adding a put statement to your original code, like put a best32.;—this will show you the full precision of the number SAS is actually using, not just the rounded display value.

内容的提问来源于stack exchange，提问作者PoorDataAnalyst