--- /dev/null
+//\r
+// This software is now distributed according to\r
+// the Lesser Gnu Public License. Please see\r
+// http://www.gnu.org/copyleft/lesser.txt for\r
+// the details.\r
+// -- Happy Computing!\r
+//\r
+package com.stevesoft.pat;\r
+\r
+import java.util.*;\r
+\r
+/** Uses table lookup to match [] type constructs, but\r
+ only if it can use a lookup table 256 bits in size.\r
+ It is impractical to make a table if it is too large.\r
+ */\r
+public class FastBracket extends Bracket {\r
+ int min, max;\r
+ BitSet bs;\r
+ FastBracket(boolean n) { super(n); }\r
+ // This routine can optimize a bracket, possibly\r
+ // it will replace it with a FastBracket.\r
+ static Bracket process(Bracket b,boolean ignc) {\r
+ Vector v = b.v;\r
+ b.pv = null;\r
+ try {\r
+\r
+ // Expand out the vector to make separate\r
+ // entries for other cases if ignoreCase is\r
+ // turned on.\r
+ Vector nv = v;\r
+ if(ignc) {\r
+ nv = new Vector();\r
+ for(int i=0;i<v.size();i++) {\r
+ Pattern p = (Pattern)v.elementAt(i);\r
+ nv.addElement(p);\r
+ if(p instanceof oneChar) {\r
+ oneChar oc = (oneChar)p;\r
+ nv.addElement(new oneChar(oc.altc));\r
+ } else if(p instanceof Range) {\r
+ Range ra = (Range)p;\r
+ nv.addElement(new Range(ra.altlo,ra.althi));\r
+ }\r
+ }\r
+ }\r
+ v = nv;\r
+\r
+ // Bubble sort, make sure elements\r
+ // are in order. This will allow us\r
+ // to merge them.\r
+ for(int i=0;i<v.size()-1;i++) {\r
+ for(int j=0;j<v.size()-1;j++) {\r
+ char c1 = getl(v.elementAt(j));\r
+ char c2 = getl(v.elementAt(j+1));\r
+ if(c2 < c1) {\r
+ Object o = v.elementAt(j);\r
+ v.setElementAt(v.elementAt(j+1),j);\r
+ v.setElementAt(o,j+1);\r
+ }\r
+ }\r
+ }\r
+\r
+ nv = new Vector();\r
+ // merge -- remove overlaps\r
+ Pattern p = (Pattern)v.elementAt(0);\r
+ nv.addElement(p);\r
+ for(int i=1;i<v.size();i++) {\r
+ if(geth(p)+1 >= getl(v.elementAt(i))) {\r
+ Pattern p2 = (Pattern)v.elementAt(i);\r
+ char lo = min(getl(p),getl(p2));\r
+ char hi = max(geth(p),geth(p2));\r
+ nv.setElementAt(p=mkelem(lo,hi),nv.size()-1);\r
+ } else {\r
+ p = (Pattern)v.elementAt(i);\r
+ nv.addElement(p);\r
+ }\r
+ }\r
+\r
+ b.v = v = nv;\r
+ } catch(RegSyntax e) {\r
+ e.printStackTrace();\r
+ }\r
+\r
+ // We don't want these things to be empty.\r
+ Vector negv = neg(v);\r
+ if(v.size()==1) return b;\r
+ if(negv.size()==1) {\r
+ b.v = negv;\r
+ b.neg = !b.neg;\r
+ return b;\r
+ }\r
+\r
+ // Now consider if we can make a FastBracket.\r
+ // Uses a BitSet to do a lookup.\r
+ FastBracket fb = newbrack(v,b.neg);\r
+ if(fb == null)\r
+ fb = newbrack(negv,!b.neg);\r
+ if(fb != null) {\r
+ fb.parent = b.parent;\r
+ fb.next = b.next;\r
+ return fb;\r
+ }\r
+\r
+ // return the normal Bracket.\r
+ return b;\r
+ }\r
+\r
+ // Build a FastBracket and set bits. If this can't\r
+ // be done, return null.\r
+ final static FastBracket newbrack(Vector v,boolean neg) {\r
+ FastBracket fb = new FastBracket(neg);\r
+ fb.v = v;\r
+ if(v.size()==0) return null;\r
+ fb.min = getl(v.elementAt(0));\r
+ fb.max = geth(v.elementAt(v.size()-1));\r
+ if(fb.max-fb.min <= 256) {\r
+ fb.bs = new BitSet(fb.max-fb.min+1);\r
+ for(int i=0;i<v.size();i++) {\r
+ Object o = v.elementAt(i);\r
+ int min0 = getl(o)-fb.min;\r
+ int max0 = geth(o)-fb.min;\r
+ for(int j=min0;j<=max0;j++)\r
+ fb.bs.set(j);\r
+ }\r
+ return fb;\r
+ }\r
+ return null;\r
+ }\r
+\r
+ // Negate a sorted Vector. Applying this\r
+ // operation twice should yield the same Vector\r
+ // back.\r
+ final static Vector neg(Vector v) {\r
+ try {\r
+ Vector nv = new Vector();\r
+ if(v.size()==0) {\r
+ nv.addElement(new Range((char)0,(char)65535));\r
+ return nv;\r
+ }\r
+ int p0 = getl(v.elementAt(0));\r
+ if(p0!=0)\r
+ nv.addElement(mkelem((char)0,(char)(p0-1) ));\r
+ for(int i=0;i<v.size()-1;i++) {\r
+ int hi = getl(v.elementAt(i+1))-1;\r
+ int lo = geth(v.elementAt(i))+1;\r
+ nv.addElement(mkelem((char)lo,(char)hi));\r
+ }\r
+ int pN = geth(v.lastElement());\r
+ if(pN != 65535)\r
+ nv.addElement(mkelem((char)(pN+1),(char)65535));\r
+ return nv;\r
+ } catch(RegSyntax rs) {\r
+ return null;\r
+ }\r
+ }\r
+ // Make either a Range or oneChar Object, depending on which\r
+ // is appropriate.\r
+ final static Pattern mkelem(char lo,char hi) throws RegSyntax {\r
+ return lo==hi ? (Pattern)(new oneChar(lo)) : (Pattern)(new Range(lo,hi));\r
+ }\r
+ static final char min(char a,char b) {\r
+ return a<b ? a : b;\r
+ }\r
+ static final char max(char a,char b) {\r
+ return a>b ? a : b;\r
+ }\r
+\r
+ // getl -- get lower value of Range object,\r
+ // or get value of oneChar object.\r
+ final static char getl(Object o) {\r
+ Pattern p = (Pattern)o;\r
+ if(p instanceof Range)\r
+ return ((Range)p).lo;\r
+ return ((oneChar)p).c;\r
+ }\r
+ // geth -- get higher value of Range object,\r
+ // or get value of oneChar object.\r
+ final static char geth(Object o) {\r
+ Pattern p = (Pattern)o;\r
+ if(p instanceof Range)\r
+ return ((Range)p).hi;\r
+ return ((oneChar)p).c;\r
+ }\r
+\r
+ // This is the easy part!\r
+ public int matchInternal(int pos,Pthings pt) {\r
+ if(pos >= pt.src.length() || Masked(pos,pt)) return -1;\r
+ char c = pt.src.charAt(pos);\r
+ return (neg ^ (c >= min && c <= max && bs.get(c-min)) ) ?\r
+ nextMatch(pos+1,pt) : -1;\r
+ }\r
+}\r