我有一个这样的输入数据结构:
{
{ {x1,y1,z1,p1}, {x1,y2,z1,p1}, {x1,y3,z1,p1}, {x1,y4,z1,p1} },
{ {x1,y1,z2,p2}, {x1,y2,z2,p2}, {x1,y3,z2,p2}, {x1,y4,z2,p2} },
{ {x1,y1,z3,p3}, {x1,y2,z3,p3}, {x1,y3,z3,p3}, {x1,y4,z3,p3} }
}
为了完整理解问题,每个项目都是一个布尔表达式(类型为 ITree),例如 nn < 42。
我想要结束的是显示每个位置的树的数量:
顶树:
根 | x1 / | \ z1 z2 z3 | | | 一(0) 一(4) 一(8)
树一:
根(偏移) - 偏移 = o // \ \ y1 y2 y3 y4 | | | | b(o) b(o+1) b(o+2) b(o+3)
树 b:
根(偏移) - 偏移 = o / | \ p1 p2 p3 | | | o o+1 o+2
因此,如果我有一个包含 {x1,y2,z1,p1} 将评估为 true 的值的列表,我可以很容易地看到它在单元格 1 中(实际上在 0,1 中),并且包含 {x1, y2,z1,p2} 它不在任何单元格中。
我已经构建了一个功能实现,但是它很慢:
public MultiIfTreeList Compile(List<List<ITree>> input, out List<MultiIfTreeList> all, out List<int> orderOfLeafs ) {
//{ { {x,y} }, { {z} } } -> {0:{x.ToString:x, y.ToString:y}, 1:{z.ToString:z}}
List<Tuple<int, Dictionary<string, ITree>>> andList = ConvertToDictionaryAndFlatten(input);
orderOfLeafs = new List<int>();
all = new List<MultiIfTreeList>();
return BuildIfTree(andList, orderOfLeafs, all);
}
private MultiIfTreeList BuildIfTree(List<Tuple<int, Dictionary<string, ITree>>> andList, List<int> orderOfLeafs, List<MultiIfTreeList> all)
{
if (andList.Count == 0)
return null;
var children = new List<MultiIfTree>();
while (andList.Count > 0)
{
//count number of occurances of each statement, ie x1=5 and x2=3 and find the highest one
Dictionary<string, int> counts = new Dictionary<string, int>();
foreach (var exp1 in andList.SelectMany(exp => exp.Item2.Keys))
if (!counts.ContainsKey(exp1))
counts[exp1] = 1;
else
counts[exp1]++;
var maxcount = counts.Max(x => x.Value);
if (maxcount == 1) //OPTIMIZATION: then all are different and we can just do them one at a time
{
foreach (var lst in andList)
{
var item = lst.Item2.First();
lst.Item2.Remove(item.Key);
var idx = orderOfLeafs.Count;
children.Add(
new MultiIfTree
{
Value = item.Key,
Ast = item.Value,
LeafsThis = 0,
Children = BuildIfTree(new List<Tuple<int, Dictionary<string, ITree>>> { lst }, orderOfLeafs, all),
ExpireCountWithChildren = orderOfLeafs.Count - idx,
});
}
andList.Clear();
}
else
{
//Make lists of where each statement can be found
foreach (var kvp in counts.Where(x => x.Value == maxcount))
{
var max = kvp.Key;
var expireindex = expire.Count;
ITree exp = null;
var listWithMax = new List<Tuple<int, Dictionary<string, ITree>>>();
var listWithoutMax = new List<Tuple<int, Dictionary<string, ITree>>>();
foreach (var lst in andList)
{
var copy = new Dictionary<string, ITree>(lst.Item2);
var item = new Tuple<int, Dictionary<string, ITree>>(lst.Item1, copy);
if (copy.ContainsKey(max))
{
exp = copy[max];
copy.Remove(max);
if (copy.Count == 0)
expire.Add(lst.Item1);
else
listWithMax.Add(item);
}
else
listWithoutMax.Add(item);
}
if (exp != null)
children.Add(
new MultiIfTree
{
Value = max,
Ast = exp,
LeafsThis = orderOfLeafs.Count - idx,
Children = BuildIfTree(listWithMax, expire, all),
LeafCountWithChildren = orderOfLeafs.Count - idx,
});
andList = listWithoutMax;
}
}
}
var tree = new MultiIfTreeList(children);
all.Add(tree);
return tree;
}
在一个输入有 25008 个列表,每个列表有 4 个表达式,这在我的机器上大约需要 800 毫秒。
编辑1: 要真正获得顶级树a和b,只需对它们进行散列处理,在我的特殊情况下,那些25008 * 4最终成为24000棵树,这些树被散列到760棵独特的树。这部分只需要 21 毫秒