Method HLL

src/main/java/net/agkn/hll/HLL.java:181–238

NOTE: Arguments here are named and structured identically to those in the PostgreSQL implementation, which can be found here . @param log2m log-base-2 of th

(final int log2m, final int regwidth, final int expthresh, final boolean sparseon, final HLLType type)

     *        start at. This cannot be <code>null</code>.
     */
    public HLL(final int log2m, final int regwidth, final int expthresh, final boolean sparseon, final HLLType type) {
        this.log2m = log2m;
        if((log2m < MINIMUM_LOG2M_PARAM) || (log2m > MAXIMUM_LOG2M_PARAM)) {
            throw new IllegalArgumentException("'log2m' must be at least " + MINIMUM_LOG2M_PARAM + " and at most " + MAXIMUM_LOG2M_PARAM + " (was: " + log2m + ")");
        }

        this.regwidth = regwidth;
        if((regwidth < MINIMUM_REGWIDTH_PARAM) || (regwidth > MAXIMUM_REGWIDTH_PARAM)) {
            throw new IllegalArgumentException("'regwidth' must be at least " + MINIMUM_REGWIDTH_PARAM + " and at most " + MAXIMUM_REGWIDTH_PARAM + " (was: " + regwidth + ")");
        }

        this.m = (1 << log2m);
        this.mBitsMask = m - 1;
        this.valueMask = (1 << regwidth) - 1;
        this.pwMaxMask = HLLUtil.pwMaxMask(regwidth);
        this.alphaMSquared = HLLUtil.alphaMSquared(m);
        this.smallEstimatorCutoff = HLLUtil.smallEstimatorCutoff(m);
        this.largeEstimatorCutoff = HLLUtil.largeEstimatorCutoff(log2m, regwidth);

        if(expthresh == -1) {
            this.explicitAuto = true;
            this.explicitOff = false;

            // NOTE:  This math matches the size calculation in the PostgreSQL impl.
            final long fullRepresentationSize = (this.regwidth * (long)this.m + 7/*round up to next whole byte*/)/Byte.SIZE;
            final int numLongs = (int)(fullRepresentationSize / 8/*integer division to round down*/);

            if(numLongs > MAXIMUM_EXPLICIT_THRESHOLD) {
                this.explicitThreshold = MAXIMUM_EXPLICIT_THRESHOLD;
            } else {
                this.explicitThreshold = numLongs;
            }
        } else if(expthresh == 0) {
            this.explicitAuto = false;
            this.explicitOff = true;
            this.explicitThreshold = 0;
        } else if((expthresh > 0) && (expthresh <= MAXIMUM_EXPTHRESH_PARAM)){
            this.explicitAuto = false;
            this.explicitOff = false;
            this.explicitThreshold = (1 << (expthresh - 1));
        } else {
            throw new IllegalArgumentException("'expthresh' must be at least " + MINIMUM_EXPTHRESH_PARAM + " and at most " + MAXIMUM_EXPTHRESH_PARAM + " (was: " + expthresh + ")");
        }

        this.shortWordLength = (regwidth + log2m);
        this.sparseOff = !sparseon;
        if(this.sparseOff) {
            this.sparseThreshold = 0;
        } else {
            // TODO improve this cutoff to include the cost overhead of Java
            //      members/objects
            final int largestPow2LessThanCutoff =
                    (int)NumberUtil.log2((this.m * this.regwidth) / this.shortWordLength);
            this.sparseThreshold = (1 << largestPow2LessThanCutoff);
        }

        initializeStorage(type);
    }