diff -r 000000000000 -r 6474c204b198 intl/icu/source/tools/toolutil/denseranges.cpp
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/intl/icu/source/tools/toolutil/denseranges.cpp	Wed Dec 31 06:09:35 2014 +0100
@@ -0,0 +1,158 @@
+/*
+*******************************************************************************
+*   Copyright (C) 2010, International Business Machines
+*   Corporation and others.  All Rights Reserved.
+*******************************************************************************
+*   file name:  denseranges.cpp
+*   encoding:   US-ASCII
+*   tab size:   8 (not used)
+*   indentation:4
+*
+*   created on: 2010sep25
+*   created by: Markus W. Scherer
+*
+* Helper code for finding a small number of dense ranges.
+*/
+
+#include "unicode/utypes.h"
+#include "denseranges.h"
+
+// Definitions in the anonymous namespace are invisible outside this file.
+namespace {
+
+/**
+ * Collect up to 15 range gaps and sort them by ascending gap size.
+ */
+class LargestGaps {
+public:
+    LargestGaps(int32_t max) : maxLength(max<=kCapacity ? max : kCapacity), length(0) {}
+
+    void add(int32_t gapStart, int64_t gapLength) {
+        int32_t i=length;
+        while(i>0 && gapLength>gapLengths[i-1]) {
+            --i;
+        }
+        if(i<maxLength) {
+            // The new gap is now one of the maxLength largest.
+            // Insert the new gap, moving up smaller ones of the previous
+            // length largest.
+            int32_t j= length<maxLength ? length++ : maxLength-1;
+            while(j>i) {
+                gapStarts[j]=gapStarts[j-1];
+                gapLengths[j]=gapLengths[j-1];
+                --j;
+            }
+            gapStarts[i]=gapStart;
+            gapLengths[i]=gapLength;
+        }
+    }
+
+    void truncate(int32_t newLength) {
+        if(newLength<length) {
+            length=newLength;
+        }
+    }
+
+    int32_t count() const { return length; }
+    int32_t gapStart(int32_t i) const { return gapStarts[i]; }
+    int64_t gapLength(int32_t i) const { return gapLengths[i]; }
+
+    int32_t firstAfter(int32_t value) const {
+        if(length==0) {
+            return -1;
+        }
+        int32_t minValue=0;
+        int32_t minIndex=-1;
+        for(int32_t i=0; i<length; ++i) {
+            if(value<gapStarts[i] && (minIndex<0 || gapStarts[i]<minValue)) {
+                minValue=gapStarts[i];
+                minIndex=i;
+            }
+        }
+        return minIndex;
+    }
+
+private:
+    static const int32_t kCapacity=15;
+
+    int32_t maxLength;
+    int32_t length;
+    int32_t gapStarts[kCapacity];
+    int64_t gapLengths[kCapacity];
+};
+
+}  // namespace
+
+/**
+ * Does it make sense to write 1..capacity ranges?
+ * Returns 0 if not, otherwise the number of ranges.
+ * @param values Sorted array of signed-integer values.
+ * @param length Number of values.
+ * @param density Minimum average range density, in 256th. (0x100=100%=perfectly dense.)
+ *                Should be 0x80..0x100, must be 1..0x100.
+ * @param ranges Output ranges array.
+ * @param capacity Maximum number of ranges.
+ * @return Minimum number of ranges (at most capacity) that have the desired density,
+ *         or 0 if that density cannot be achieved.
+ */
+U_CAPI int32_t U_EXPORT2
+uprv_makeDenseRanges(const int32_t values[], int32_t length,
+                     int32_t density,
+                     int32_t ranges[][2], int32_t capacity) {
+    if(length<=2) {
+        return 0;
+    }
+    int32_t minValue=values[0];
+    int32_t maxValue=values[length-1];  // Assume minValue<=maxValue.
+    // Use int64_t variables for intermediate-value precision and to avoid
+    // signed-int32_t overflow of maxValue-minValue.
+    int64_t maxLength=(int64_t)maxValue-(int64_t)minValue+1;
+    if(length>=(density*maxLength)/0x100) {
+        // Use one range.
+        ranges[0][0]=minValue;
+        ranges[0][1]=maxValue;
+        return 1;
+    }
+    if(length<=4) {
+        return 0;
+    }
+    // See if we can split [minValue, maxValue] into 2..capacity ranges,
+    // divided by the 1..(capacity-1) largest gaps.
+    LargestGaps gaps(capacity-1);
+    int32_t i;
+    int32_t expectedValue=minValue;
+    for(i=1; i<length; ++i) {
+        ++expectedValue;
+        int32_t actualValue=values[i];
+        if(expectedValue!=actualValue) {
+            gaps.add(expectedValue, (int64_t)actualValue-(int64_t)expectedValue);
+            expectedValue=actualValue;
+        }
+    }
+    // We know gaps.count()>=1 because we have fewer values (length) than
+    // the length of the [minValue..maxValue] range (maxLength).
+    // (Otherwise we would have returned with the one range above.)
+    int32_t num;
+    for(i=0, num=2;; ++i, ++num) {
+        if(i>=gaps.count()) {
+            // The values are too sparse for capacity or fewer ranges
+            // of the requested density.
+            return 0;
+        }
+        maxLength-=gaps.gapLength(i);
+        if(length>num*2 && length>=(density*maxLength)/0x100) {
+            break;
+        }
+    }
+    // Use the num ranges with the num-1 largest gaps.
+    gaps.truncate(num-1);
+    ranges[0][0]=minValue;
+    for(i=0; i<=num-2; ++i) {
+        int32_t gapIndex=gaps.firstAfter(minValue);
+        int32_t gapStart=gaps.gapStart(gapIndex);
+        ranges[i][1]=gapStart-1;
+        ranges[i+1][0]=minValue=(int32_t)(gapStart+gaps.gapLength(gapIndex));
+    }
+    ranges[num-1][1]=maxValue;
+    return num;
+}