Years ago I was in a job interview where the employer took me to a whiteboard and asked me to implement atoi()
in C. They also asked me to talk about any issues with my implementation. While there is hardly any reason to implement atoi()
yourself nowadays, sometimes it’s good practice to try it anyway.
C Standard Library interface
First lets looks at the function definition:
1 2 3 

The C standard library method atoi()
takes in a string (character array) and returns a number. You’ll actually notice that the man page deprecated atoi()
in favor of strtol()
which uses long instead of int. Our implementation will use a long instead of an int as well:
1


Number Theory
Before we start digging into code, what do we actually mean when we write out a number? Well according to Wikipedia’s entry on numbers:
In this base ten system, the rightmost digit of a natural number has a place value of one, and every other digit has a place value ten times that of the place value of the digit to its right.
Lets look at the numbers 7, 42, and 365 to see how this would work:
7 = 7×1 = 7×10^{0}
42 = 4×10 + 2×1 = 4×10^{1} + 2×10^{0}
365 = 3×100 + 6×10 + 5×1 = 3×10^{2} + 6×10^{1} + 5×10^{0}
General Rule
Each character in the number string can be represented by n×10^{p} where n is the numerical value of the character and p is the place value of the character decreasing from stringLength1 to 0
Getting the Numerical Value for a Character
According to the ASCII Table, numbers 0–9 have the values 48–57 respectively. We can get the numerical position by taking the ASCII value and subtracting 48.
Getting the Place Value of the Character
The place value is interesting to get because while most of our counting increases from left to right, the place value increases from right to left. We can figure out the place value by finding the string length using the function strlen()
1 2 3 

Once we know the string’s length, we can calculate the place value by subtracting the character’s position from left to right from the string’s length. We will also need to subtract 1 because the place value is 0 based—the right most place value is 10^{0}.
First Attempt
1 2 3 4 5 6 7 8 9 10 11 

Lets see how it works:
cc_stringToNumber("1234") 
1234  Correct! 
cc_stringToNumber("0") 
0  Correct! 
cc_stringToNumber("Foo") 
2893  Wrong! 
cc_stringToNumber("42") 
258  Wrong! 
There are 2 issues here. The first is that we’re taking the integer value of each character wether they are a number or not. ‘F’ has an ASCII value of 70 which means we calculate it’s value to be 7048 or 22. ‘o’ has an ASCII value of 111 so we calculate it’s value to be 63. This gives us the overall value: (22 * 10^{2}) + (63 * 10^{1}) + (63 * 10^{0}) = 2893. We need a solution: If we encounter any character < ‘0’ or > ‘9’, we need to return 0 for the value.
Second Attempt
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 

Lets see how it works:
cc_stringToNumber("1234") 
1234  Correct! 
cc_stringToNumber("0") 
0  Correct! 
cc_stringToNumber("Foo") 
0  Correct! 
cc_stringToNumber("42") 
0  Wrong! 
Well, this is an improvement, but we’re not handling negative numbers correctly. Negative numbers are hard for a couple of reason. The 
sign will move the character positions off by one. The minus sign any other place in a number is also not valid numerical number. It either has to be at the beginning or not there at all.
Our solution is to look at the first character: If it’s not a minus sign, we’ll run the function as normal. If it is, we’ll run the function with a subset of the data
Final Attempt
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 

Lets see how it works:
cc_stringToNumber("1234") 
1234  Correct! 
cc_stringToNumber("0") 
0  Correct! 
cc_stringToNumber("Foo") 
0  Correct! 
cc_stringToNumber("42") 
42  Correct! 
You’ll see I split up the function into two different methods. cc_positiveStringToNumber()
assumes that the string is always positive and has our current logic. cc_stringToNumber()
has the logic to handle the negative sign. If it encounters a negative sign as the first character (and has at least one more character after it), then it performs cc_positiveStringToNumber()
with the minus sign stripped out (++string moves the string pointer forward a character).
This also handles just a minus sign and an empty string.
In Conclusion
While it may be fun and educational to implement a standard function, it’s worth noting that our method has a few limitations the C Standard Library functions don’t have. We don’t support the “+” sign for positive number (though this is an easy feature to add). More importantly, strtol()
supports different bases like octal and hexadecimal. Finally, keep in mind that c standard library functions have been tested and optimized for years.