/*******************************************************************************
 * Copyright (c) 2003, 2005 IBM Corporation and others.
 * All rights reserved. This program and the accompanying materials
 * are made available under the terms of the Eclipse Public License v1.0
 * which accompanies this distribution, and is available at
 * http://www.eclipse.org/legal/epl-v10.html
 *
 * Contributors:
 *     IBM Corporation - initial API and implementation
 * Port to the D programming language:
 *     Frank Benoit <benoit@tionex.de>
 *******************************************************************************/
module dwtx.draw2d.graph.RankSorter;

import dwt.dwthelper.utils;
import dwtx.dwtxhelper.Collection;
import dwtx.dwtxhelper.Random;
import dwtx.draw2d.graph.Node;
import dwtx.draw2d.graph.Rank;
import dwtx.draw2d.graph.DirectedGraph;
import dwtx.draw2d.graph.EdgeList;
import dwtx.draw2d.graph.Edge;
import dwtx.draw2d.graph.GraphUtilities;

/**
 * Sorts Ranks during the up and down sweeps of the MinCross visitor.
 * @author Randy Hudson
 * @since 2.1.2
 */
class RankSorter {

Random flipflop;
Node node;
double rankSize, prevRankSize, nextRankSize;
int currentRow;
Rank rank;
double progress;
DirectedGraph g;

public this(){
    flipflop = new Random(3);
}

protected void assignIncomingSortValues() {
    rankSize = rank.total;
    prevRankSize = g.ranks.getRank(currentRow - 1).total;
    if (currentRow < g.ranks.size() - 1)
        nextRankSize = g.ranks.getRank(currentRow + 1).total;
    for (int n = 0; n < rank.count(); n++) {
        node = rank.getNode(n);
        sortValueIncoming();
    }
}

protected void assignOutgoingSortValues() {
    rankSize = rank.total;
    prevRankSize = g.ranks.getRank(currentRow + 1).total;
    if (currentRow > 1)
        nextRankSize = g.ranks.getRank(currentRow - 1).total;

    for (int n = 0; n < rank.count(); n++) {
        node = rank.getNode(n);
        sortValueOutgoing();
    }
}

double evaluateNodeIncoming() {
    bool change = false;
    EdgeList incoming = node.incoming;
    do {
        change = false;
        for (int i = 0; i < incoming.size() - 1; i++) {
            if (incoming.getSourceIndex(i) > incoming.getSourceIndex(i + 1)) {
                Edge e = incoming.getEdge(i);
                incoming.set(i, incoming.get(i + 1));
                incoming.set(i + 1, e);
                change = true;
            }
        }
    } while (change);

    int n = incoming.size();
    if (n is 0) {
        return node.index * prevRankSize / rankSize;
    }
    if (n % 2 is 1)
        return incoming.getSourceIndex(n / 2);

    int l = incoming.getSourceIndex(n / 2 - 1);
    int r = incoming.getSourceIndex(n / 2);
    if (progress >= 0.8 && n > 2) {
        int dl = l - incoming.getSourceIndex(0);
        int dr = incoming.getSourceIndex(n - 1) - r;
        if (dl < dr)
            return l;
        if (dl > dr)
            return r;
    }
    if (progress > 0.25 && progress < 0.75) {
        if (flipflop.nextBoolean())
            return (l + l + r) / 3.0;
        else
            return (r + r + l) / 3.0;
    }
    return (l + r) / 2.0;
}

double evaluateNodeOutgoing() {
    bool change = false;
    EdgeList outgoing = node.outgoing;
    do {
        change = false;
        for (int i = 0; i < outgoing.size() - 1; i++) {
            if (outgoing.getTargetIndex(i) > outgoing.getTargetIndex(i + 1)) {
                Edge e = outgoing.getEdge(i);
                outgoing.set(i, outgoing.get(i + 1));
                outgoing.set(i + 1, e);
                change = true;
            }
        }
    } while (change);

    int n = outgoing.size();
    if (n is 0)
        return node.index * prevRankSize / rankSize;
    if (n % 2 is 1)
        return outgoing.getTargetIndex(n / 2);
    int l = outgoing.getTargetIndex(n / 2 - 1);
    int r = outgoing.getTargetIndex(n / 2);
    if (progress >= 0.8 && n > 2) {
        int dl = l - outgoing.getTargetIndex(0);
        int dr = outgoing.getTargetIndex(n - 1) - r;
        if (dl < dr)
            return l;
        if (dl > dr)
            return r;
    }
    if (progress > 0.25 && progress < 0.75) {
        if (flipflop.nextBoolean())
            return (l + l + r) / 3.0;
        else
            return (r + r + l) / 3.0;
    }
    return (l + r) / 2.0;
}

public void sortRankIncoming(DirectedGraph g, Rank rank, int row, double progress) {
    this.currentRow = row;
    this.rank = rank;
    this.progress = progress;
    assignIncomingSortValues();
    sort();
    postSort();
}

public void init(DirectedGraph g) {
    this.g = g;
    for (int i = 0; i < g.ranks.size(); i++) {
        rank = g.ranks.getRank(i);

        //Sort the ranks based on their constraints. Constraints are preserved throughout.
        Collections.sort(rank, new class() Comparator {
            public int compare(Object left, Object right) {
                return (cast(Node)left).rowOrder - (cast(Node)right).rowOrder;
            }
        });
        postSort();
    }
}

void optimize(DirectedGraph g) {
}

protected void postSort() {
    rank.assignIndices();
}

void sort() {
    bool change;
    do {
        change = false;
        for (int i = 0; i < rank.size() - 1; i++)
            change |= swap(i);
        if (!change)
            break;
        change = false;
        for (int i = rank.size() - 2; i >= 0; i--)
            change |= swap(i);
    } while (change);
}

bool swap(int i) {
    Node left = rank.getNode(i);
    Node right = rank.getNode(i + 1);
    if (GraphUtilities.isConstrained(left, right))
        return false;
    if (left.sortValue <= right.sortValue)
        return false;
    rank.set(i, right);
    rank.set(i + 1, left);
    return true;
}

public void sortRankOutgoing(DirectedGraph g, Rank rank, int row, double progress) {
    this.currentRow = row;
    this.rank = rank;
    this.progress = progress;
    assignOutgoingSortValues();
    sort();
    postSort();
}

void sortValueIncoming() {
    node.sortValue = evaluateNodeIncoming();
    //$TODO restore this optimization
//  if (progress is 0.0 && !(node instanceof VirtualNode))
//      node.sortValue = -1;
    double value = evaluateNodeOutgoing();
    if (value < 0)
        value = node.index * nextRankSize / rankSize;
    node.sortValue += value * progress;
//  if (progress < 0.7 && node.sortValue !is -1)
//      node.sortValue += Math.random() * rankSize / (5 + 8 * progress);
}

void sortValueOutgoing() {
    node.sortValue = evaluateNodeOutgoing();
    //$TODO restore this optimization
//  if (progress is 0.0 && !(node instanceof VirtualNode))
//      node.sortValue = -1;
    double value = evaluateNodeIncoming();
    if (value < 0)
        value = node.index * nextRankSize / rankSize;
    node.sortValue += value * progress;
//  if (progress < 0.7 && node.sortValue !is -1)
//      node.sortValue += Math.random() * rankSize / (5 + 8 * progress);
}

}