Why does this gives a runtime floating point error?

Question

I am doing an assignment for school which introduced hashmaps, and so I am creating a templated class for a hashmap that uses the std::hash function. The problem that I am having comes in my insert function, which is shown below:

template <class K, class V>
void HashMap<K, V>::insert(K key, V value)
{
    std::hash<std::string> stringHash;
    int intKey = stringHash(key);
    int bucket = intKey % this->size();
    map[bucket].push_back(std::pair<K, V>(key, value));
}

My error occurs at the line: int bucket = intKey % this->size();.

I don't quite understand why this would give a floating point error since I am doing my work entirely in integers. With the key "banana" and the value 3, the hashed int is 2068534322. In the case where this->size is 5, the modulo should be evaluated as 2.

So, why exactly am I getting a floating point error?

EDIT 1: I also tried this->size() replaced with a hard-coded 5 (which is what this->size should evaluate to), so the this->size isn't having a problem evaluating with a 0.

If any more information is required to help, please feel free to leave me a comment. I didn't want to give so much information that it would be overwhelming. — shermanzach, Jun 02 '14 at 20:07
What's a "floating point error"? Please describe in detail the error you encounter. — , Jun 02 '14 at 20:09
You might assert that `this->size();` isn't `0` when performing this operation! — πάντα ῥεῖ, Jun 02 '14 at 20:09
I will throw in the assert for later testing, but I have updated the original post with an explanation about the % 0. — shermanzach, Jun 02 '14 at 20:13
This is trivially easy for you to debug. Don't be helpless. If the line `int bucket = intKey % this->size();` results in a runtime error, use your debugging tools to determine the values that you pass to the `%` operator. It really could not be easier. Once you know those values, produce a simple SSCCE to illustrate the problem. — David Heffernan, Jun 02 '14 at 20:16
Any chance this http://stackoverflow.com/questions/12570374/floating-point-exception-sigfpe-on-int-main-return0 is relevant? I have seen weird floating errors when libraries do not agree. — Sean Perry, Jun 02 '14 at 20:20
std::hash will give you a `size_t` not an `int`. You probably do not want to calculate -1245433%5 and then use the result as the index in `map[bucket]` — nos, Jun 02 '14 at 20:31

πάντα ῥεῖ · Accepted Answer · 2014-06-02T20:24:01.063

4

You do a modulo (== division) operation, so you need to assure your denominator isn't zero

template <class K, class V>
void HashMap<K, V>::insert(K key, V value)
{
    std::hash<std::string> stringHash;
    int intKey = stringHash(key);
    int bucket = this->size() ? intKey % this->size() : intKey; 
       // or whatever makes sense to assign for the latter condition
    map[bucket].push_back(std::pair<K, V>(key, value));
}

Or at least place an assert statement when doing this to track where wrong calls came from:

std::assert(this->size());
int bucket = intKey % this->size();

edited Jun 02 '14 at 20:24

answered Jun 02 '14 at 20:18

πάντα ῥεῖ

1
13
116
190

If `size()` is the size of the underlying array and equals 0, that means there is no `map` to put stuff into. IOW it's not as easy as substituting a different value for `bucket`. – Jun 02 '14 at 20:35

Why does this gives a runtime floating point error?

1 Answers1