CODEWITHSUNDEEP
javaclojurehashmapjava-interop
Joseph Yourine
Joseph Yourine

Reputation: 1331

Clojure : java interop, why does this code do not work?

I coded a class using Apache Maths (whatever) and I have a weird glitch (or maybe there is a problem in my Java code ?).

here is the code

package distributions;

import java.util.HashMap;
import java.util.Map;
import java.util.Random;
import org.apache.commons.math3.distribution.AbstractIntegerDistribution;
import org.apache.commons.math3.util.FastMath;

public class CustomDistribution extends AbstractIntegerDistribution {

    long[] v;
    double[] p;
    double[] cp;
    Map<Long, Double> mp = new HashMap<Long, Double>();
    Map<Long, Double> mcp = new HashMap<Long, Double>();

    public CustomDistribution (long[] values, double[] probabilities) {
        v = values;
        p = probabilities;

        int len = values.length;

        for (int i = 0; i < len; i++) {
            mp.put(v[i], p[i]);
        }

        cp = new double[len];
        cp[0] = p[0];
        for (int i = 1; i < len; i++) {
            cp[i] = cp[i-1] + p[i];
        }

        for (int i = 0; i < len; i++) {
            mcp.put(v[i], cp[i]);
        }
    }

      public Map<Long, Double> getMCP () {return mcp;}

    @Override
    public double cumulativeProbability(int v) {
        return mcp.get(v);
    }

    // Not implemented, can be done in Clojure code
    @Override
    public double getNumericalMean() {
        return 0;
    }

    // Not implemented, can be done in Clojure code
    @Override
    public double getNumericalVariance() {
        return 0;
    }

    @Override
    public int getSupportLowerBound() {
        return 0;
    }

    @Override
    public int getSupportUpperBound() {
        return 0;
    }

    @Override
    public boolean isSupportConnected() {
        return false;
    }

    @Override
    public double probability(int v) {
        return mp.get(v);
    }

    // Uses a naive search implementation, should be ok due to data size
    public int sample () {

        double r = FastMath.random();
        int len = p.length;
        Boolean flag = false;
        int i = 0;
        int result = -1;

        while (i < len && flag == false) {
            if (cp[i] < r) {
                i = i + 1;
            }
            else {
                result = (int) v[i];
                flag = true;
            }
        }
        return result;
    }
    }

Clojure wrapper :

(defn create-distribution
  "Creates a distribution from empirical data"
  [data]
  (let [values (long-array (data :values))
        probabilities (double-array (data :probabilities))]
    (CustomDistribution. values probabilities)))

(create-distribution {:values [1 2 3 4] :probabilities [0.3 0.2 0.2 0.3]})

Basically it is a distribution with data inside in two forms : HashMaps and Arrays.

In Clojure, I tried :

(.probability dist 4)
(.probability dist (int 4))

Both returns NullPointerException, NativeMethodAccessorImpl..blabla Same for the .cumulativeProbabilityMethod (on the other hand, .sample works fine so it may be a problem of HashMap)

I figured out that maybe in the constructor mcp and mp (that are HashMaps) were not properly calculated in the constructor but when I tried :

(.get (.getMCP dist) 4)

it returns me the correct result. MCP hashmap is in fact the result I expected.

So why does my code does not work ? My two lines are "functionally" the same. If I did a typo, I do not see it. Maybe it's because of the immutability thing ?

Thanks !

Upvotes: 0

Views: 124

Answers (1)

Joseph Yourine
Joseph Yourine

Reputation: 1331

Ok so I figured out the problem, I forgot that Java was not as flexible as Clojure with numeric primitives.

This code works :

@Override
public double cumulativeProbability(int v) {
    return mcp.get((long) v);
}

@Override
public double probability(int v) {
    return mp.get((long) v);
}

In Clojure code, the translation of a double is automatic. In Java I passed an Integer where the HasMap expected a Long

Upvotes: 1

Related Questions