Mercurial > dive4elements > river
view flys-artifacts/src/main/java/de/intevation/flys/artifacts/math/WKmsOperation.java @ 5779:ebec12def170
Datacage: Add a pool of builders to make it multi threadable.
XML DOM is not thread safe. Therefore the old implementation only allowed one thread
to use the builder at a time. As the complexity of the configuration
has increased over time this has become a bottleneck of the whole application
because it took quiet some time to build a result. Furthermore the builder code path
is visited very frequent. So many concurrent requests were piled up
resulting in long waits for the users.
To mitigate this problem a round robin pool of builders is used now.
Each of the pooled builders has an independent copy of the XML template
and can be run in parallel.
The number of builders is determined by the system property
'flys.datacage.pool.size'. It defaults to 4.
author | Sascha L. Teichmann <teichmann@intevation.de> |
---|---|
date | Sun, 21 Apr 2013 12:48:09 +0200 |
parents | 5642a83420f2 |
children |
line wrap: on
line source
package de.intevation.flys.artifacts.math; import de.intevation.flys.artifacts.model.WKms; import de.intevation.flys.artifacts.model.WKmsImpl; import java.util.Arrays; public abstract class WKmsOperation { public static final double EPSILON = 1e-6; public static final class KmW implements Comparable<KmW> { protected double km; protected double w; public KmW(double km, double w) { this.km = km; this.w = w; } public int compareTo(KmW other) { return km < other.km ? -1 : km > other.km ? +1 : 0; } public boolean kmEquals(KmW other) { return Math.abs(km - other.km) < EPSILON; } public double subtract(KmW other) { return w - other.w; } } // class KmW public static final WKmsOperation SUBTRACTION = new WKmsOperation() { @Override public WKms operate(WKms a, WKms b) { return subtract(a, b); } }; protected WKmsOperation() { } public abstract WKms operate(WKms a, WKms b); /** * Subtract two series from each other, interpolate values * missing in one series in the other. */ public static WKms subtract(WKms minuend, WKms subtrahend) { int M = minuend .size(); int S = subtrahend.size(); // Don't subtract empty sets if (M < 1 || S < 1) { return new WKmsImpl(); } KmW [] ms = new KmW[M]; KmW [] ss = new KmW[S]; for (int i = 0; i < M; ++i) { ms[i] = new KmW(minuend.getKm(i), minuend.getW(i)); } for (int i = 0; i < S; ++i) { ss[i] = new KmW(subtrahend.getKm(i), subtrahend.getW(i)); } Arrays.sort(ms); Arrays.sort(ss); // no overlap -> empty result set if (ms[0].km > ss[S-1].km || ss[0].km > ms[M-1].km) { return new WKmsImpl(); } WKmsImpl result = new WKmsImpl(); int mi = 0; int si = 0; OUT: while (mi < M && si < S) { KmW m = ms[mi]; KmW s = ss[si]; if (m.km + EPSILON < s.km) { // minuend is before subtrahend while (ms[mi].km + EPSILON < s.km) { if (++mi >= M) { break OUT; } } if (ms[mi].km + EPSILON > s.km) { double mw = Linear.linear( s.km, ms[mi-1].km, ms[mi].km, ms[mi-1].w, ms[mi].w); result.add(s.km, mw - s.w); ++si; } else { // s.km == ms[mi].km result.add(s.km, ms[mi].subtract(s)); ++mi; ++si; } } else if (m.km > s.km + EPSILON) { // subtrahend is before minuend while (m.km > ss[si].km + EPSILON) { if (++si >= S) { break OUT; } } if (ss[si].km + EPSILON > m.km) { double sw = Linear.linear( m.km, ss[si-1].km, ss[si].km, ss[si-1].w, ss[si].w); result.add(m.km, m.w - sw); } else { // ss[si].km == m.km result.add(m.km, m.subtract(ss[si])); ++mi; ++si; } } else { // m.km == s.km result.add(s.km, m.subtract(s)); ++mi; ++si; } } return result; } } // vim:set ts=4 sw=4 si et sta sts=4 fenc=utf-8 :