com.google.javascript.jscomp.regex.CharRanges Maven / Gradle / Ivy
Show all versions of closure-compiler-unshaded Show documentation
/*
* Copyright 2011 The Closure Compiler Authors.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package com.google.javascript.jscomp.regex;
import static java.lang.Math.max;
import static java.lang.Math.min;
import java.util.Arrays;
/**
* An immutable sparse bitset that deals well where the data is chunky:
* where P(bit[x+1] == bit[x]). E.g. [101,102,103,104,105,1001,1002,1003,1004]
* is chunky.
*/
final class CharRanges {
/**
* A strictly increasing set of bit indices where even members are the
* inclusive starts of ranges, and odd members are the exclusive ends.
*
* E.g., { 1, 5, 6, 10 } represents the set ( 1, 2, 3, 4, 6, 7, 8, 9 ).
*/
private final int[] ranges;
public static final CharRanges EMPTY = new CharRanges(new int[0]);
public static final CharRanges ALL_CODE_UNITS
= new CharRanges(new int[] { 0, 0x10000 });
public static CharRanges inclusive(int start, int end) {
if (start > end) {
throw new IndexOutOfBoundsException(start + " > " + end);
}
return new CharRanges(new int[] { start, end + 1 });
}
/**
* Returns an instance containing all and only the given members.
*/
public static CharRanges withMembers(int... members) {
return new CharRanges(intArrayToRanges(members));
}
/**
* Returns an instance containing the given ranges.
* @param ranges An even-length ordered sequence of non-overlapping,
* non-contiguous, [inclusive start, exclusive end) ranges.
*/
public static CharRanges withRanges(int... ranges) {
if ((ranges.length & 1) != 0) { throw new IllegalArgumentException(); }
for (int i = 1; i < ranges.length; ++i) {
if (ranges[i] <= ranges[i - 1]) {
throw new IllegalArgumentException(ranges[i] + " > " + ranges[i - 1]);
}
}
return new CharRanges(ranges);
}
private CharRanges(int[] ranges) {
this.ranges = ranges;
}
private static int[] intArrayToRanges(int[] members) {
int nMembers = members.length;
if (nMembers == 0) {
return new int[0];
}
Arrays.sort(members);
// Count the number of runs.
int nRuns = 1;
for (int i = 1; i < nMembers; ++i) {
int current = members[i];
int last = members[i - 1];
if (current == last) { continue; }
if (current != last + 1) { ++nRuns; }
}
int[] ranges = new int[nRuns * 2];
ranges[0] = members[0];
int k = 0;
for (int i = 1; k + 2 < ranges.length; ++i) {
int current = members[i];
int last = members[i - 1];
if (current == last) { continue; }
if (current != last + 1) {
ranges[++k] = last + 1; // add 1 to make end exclusive
ranges[++k] = current;
}
}
ranges[++k] = members[nMembers - 1] + 1; // add 1 to make end exclusive
return ranges;
}
public boolean contains(int bit) {
return (Arrays.binarySearch(ranges, bit) & 1) == 0;
// By the contract of Arrays.binarySearch, its result is either the position
// of bit in ranges or it is the bitwise inverse of the position of the
// least element greater than bit.
// Two cases
// case (idx >= 0)
// We ended up exactly on a range boundary.
// Starts are inclusive and ends are both exclusive, so this contains
// bit iff idx is even.
//
// case (idx < 0)
// If the least element greater than bit is an odd element,
// then bit must be greater than a start and less than an end, so
// contained.
//
// If bit is greater than all elements, then idx will be past the end of
// the array, and will be even since ranges.length is even.
//
// Otherwise, bit must be in the space between two runs, so not
// contained.
//
// In all cases, oddness is equivalent to containedness.
// Those two cases lead to
// idx >= 0 ? ((idx & 1) == 0) : ((~idx & 1) == 1)
// But ~n & bit == bit <=> n & bit == 0, so
// idx >= 0 ? ((idx & 1) == 0) : ((~idx & 1) == 1)
// => idx >= 0 ? ((idx & 1) == 0) : ((idx & 1) == 0)
// => (idx & 1) == 0
}
public boolean isEmpty() {
return ranges.length == 0;
}
public int getNumRanges() { return ranges.length >> 1; }
public int start(int i) { return ranges[i << 1]; }
public int end(int i) { return ranges[(i << 1) | 1]; }
public CharRanges union(CharRanges other) {
// Index of the input ranges
int[] q = this.ranges;
int[] r = other.ranges;
// Lengths of the inputs
int m = q.length;
int n = r.length;
if (m == 0) { return other; }
if (n == 0) { return this; }
// The output array. The length is m+n in the worst case when all the
// ranges in a are disjoint from the ranges in b.
int[] out = new int[m + n];
// Indexes into the various arrays
int i = 0;
int j = 0;
int k = 0;
// Since there are three arrays, and indices into them the following
// should never occur in this function:
// (1) q[j] or q[k] -- q is indexed by i
// (2) r[i] or r[k] -- r is indexed by j
// (3) out[i] or out[j] -- out is indexed by k
// (4) i < n or j < m -- index compared to wrong limit
// This loop exits because we always increment at least one of i,j.
while (i < m && j < n) {
// Range starts and ends.
int a0 = q[i];
int a1 = q[i + 1];
int b0 = r[j];
int b1 = r[j + 1];
if (a1 < b0) { // [a0, a1) ends before [b0, b1) starts
out[k++] = a0;
out[k++] = a1;
i += 2;
} else if (b1 < a0) { // [b0, b1) ends before [a0, a1) starts
out[k++] = b0;
out[k++] = b1;
j += 2;
} else { // ranges overlap
// We need to compute a new range based on the set of ranges that
// transitively overlap.
// AAAAAAAAA AAA
// BBB BBB* BBB
// In the range above, the start comes from one set, and the end from
// another. The range with the asterisk next to it is subsumed entirely
// by a range from the other, and so not all ranges on the input
// contribute a value to the output.
// The last BBB run serves only as a bridge -- it overlaps two
// disjoint ranges in the other one so establishes that they
// transitively overlap.
int start = min(a0, b0);
// Guess at the end, and lookahead to come up with a more complete
// estimate.
int end = max(a1, b1);
i += 2;
j += 2;
while (i < m || j < n) {
if (i < m && q[i] <= end) {
end = max(end, q[i + 1]);
i += 2;
} else if (j < n && r[j] <= end) {
end = max(end, r[j + 1]);
j += 2;
} else {
break;
}
}
out[k++] = start;
out[k++] = end;
}
}
// There may be unprocessed ranges at the end of one of the inputs.
if (i < m) {
System.arraycopy(q, i, out, k, m - i);
k += m - i;
} else if (j < n) {
System.arraycopy(r, j, out, k, n - j);
k += n - j;
}
// We guessed at the output length above. Cut off the tail.
if (k != out.length) {
int[] clipped = Arrays.copyOf(out, k);
out = clipped;
}
return new CharRanges(out);
}
public CharRanges intersection(CharRanges other) {
int[] aRanges = ranges;
int[] bRanges = other.ranges;
int aLen = aRanges.length;
int bLen = bRanges.length;
if (aLen == 0) { return this; }
if (bLen == 0) { return other; }
int aIdx = 0;
int bIdx = 0;
int[] intersection = new int[min(aLen, bLen)];
int intersectionIdx = 0;
int pos = min(aRanges[0], bRanges[0]);
while (aIdx < aLen && bIdx < bLen) {
if (aRanges[aIdx + 1] <= pos) {
aIdx += 2;
} else if (bRanges[bIdx + 1] <= pos) {
bIdx += 2;
} else {
int start = max(aRanges[aIdx], bRanges[bIdx]);
if (pos < start) { // Advance to start of common block.
pos = start;
} else {
// Now we know that pos is less than the ends of the two ranges and
// greater or equal to the starts of the two ranges.
int end = min(aRanges[aIdx + 1], bRanges[bIdx + 1]);
if (intersectionIdx != 0 && pos == intersection[intersectionIdx - 1]) {
intersection[intersectionIdx - 1] = end;
} else {
if (intersectionIdx == intersection.length) {
int[] newArr = new int[intersectionIdx * 2];
System.arraycopy(intersection, 0, newArr, 0, intersectionIdx);
intersection = newArr;
}
intersection[intersectionIdx++] = pos;
intersection[intersectionIdx++] = end;
}
pos = end;
}
}
}
if (intersectionIdx != intersection.length) {
int[] newArr = Arrays.copyOf(intersection, intersectionIdx);
intersection = newArr;
}
return new CharRanges(intersection);
}
public CharRanges difference(CharRanges subtrahendRanges) {
// difference = minuend - subtrahend
int[] minuend = this.ranges;
int[] subtrahend = subtrahendRanges.ranges;
int mn = minuend.length;
int sn = subtrahend.length;
if (mn == 0 || sn == 0) { return this; }
int[] difference = new int[minuend.length];
// Indices into minuend.ranges, subtrahend.ranges, and difference.
int mIdx = 0;
int sIdx = 0;
int dIdx = 0;
int pos = minuend[0];
while (mIdx < mn) {
if (pos >= minuend[mIdx + 1]) {
mIdx += 2;
} else if (pos < minuend[mIdx]) {
// Skip gaps in the minuend.
pos = minuend[mIdx];
} else if (sIdx < sn && pos >= subtrahend[sIdx]) {
// Skip over a removed part.
pos = subtrahend[sIdx + 1];
sIdx += 2;
} else {
// Now we know that pos is between [minuend[i], minuend[i + 1])
// and outside [subtrahend[j], subtrahend[j + 1]).
int end = sIdx < sn ? min(minuend[mIdx + 1], subtrahend[sIdx]) : minuend[mIdx + 1];
if (dIdx != 0 && difference[dIdx - 1] == pos) {
difference[dIdx - 1] = pos;
} else {
if (dIdx == difference.length) {
int[] newArr = new int[dIdx * 2];
System.arraycopy(difference, 0, newArr, 0, dIdx);
difference = newArr;
}
difference[dIdx++] = pos;
difference[dIdx++] = end;
}
pos = end;
}
}
if (dIdx != difference.length) {
int[] newArr = Arrays.copyOf(difference, dIdx);
difference = newArr;
}
return new CharRanges(difference);
}
public boolean containsAll(CharRanges sub) {
int[] superRanges = this.ranges;
int[] subRanges = sub.ranges;
int superIdx = 0;
int subIdx = 0;
int superLen = superRanges.length;
int subLen = subRanges.length;
while (subIdx < subLen) {
if (superIdx == superLen) {
return false;
}
if (superRanges[superIdx + 1] <= subRanges[subIdx]) {
// Super range ends before subRange starts.
superIdx += 2;
} else if (superRanges[superIdx] > subRanges[subIdx]) {
// Uncontained portion at start of sub range.
return false;
} else if (superRanges[superIdx + 1] >= subRanges[subIdx + 1]) {
// A sub range is completely contained in the super range.
// We know this because of the above condition and we have already
// ruled out that subRanges[subIdx] < superRanges[superIdx].
subIdx += 2;
} else {
// Uncontained portion at end of sub range.
return false;
}
}
return subIdx == subLen;
}
/**
* Shifts the bits matched by the given delta.
* So if this has the bits (a, b, c, ..., z) set then the result has the bits
* ((a - delta), (b - delta), (c - delta), ...., (z - delta)) set.
*
* @throws IndexOutOfBoundsException if shifting by delta would cause an
* overflow or underflow in a 32 bit {@code signed int} range boundary.
* Since the end boundaries of ranges are exclusive, even if there is no
* range containing {@link Integer#MAX_VALUE}, shifting by a delta of 1
* can cause an overflow.
*/
public CharRanges shift(int delta) {
int n = ranges.length;
if (delta == 0 || n == 0) { return this; }
// Test overflow/underflow
if (delta < 0) {
long lmin = ranges[0] + delta;
if (lmin < Integer.MIN_VALUE) { throw new IndexOutOfBoundsException(); }
} else {
long lmax = ranges[n - 1] + delta;
if (lmax > Integer.MAX_VALUE) { throw new IndexOutOfBoundsException(); }
}
// Create a shifted range.
int[] shiftedRanges = new int[n];
for (int i = n; --i >= 0;) {
shiftedRanges[i] = ranges[i] + delta;
}
return new CharRanges(shiftedRanges);
}
@Override
public String toString() {
StringBuilder sb = new StringBuilder();
sb.append('[');
for (int i = 0; i < ranges.length; ++i) {
if ((i & 1) != 0 && ranges[i] == ranges[i - 1] + 1) { continue; }
if (i != 0) { sb.append((i & 1) == 0 ? ' ' : '-'); }
sb.append("0x").append(Integer.toString(ranges[i] - (i & 1), 16));
}
sb.append(']');
return sb.toString();
}
@Override
public boolean equals(Object o) {
if (!(o instanceof CharRanges)) { return false; }
return Arrays.equals(this.ranges, ((CharRanges) o).ranges);
}
@Override
public int hashCode() {
int hc = 0;
for (int i = 0, n = min(16, ranges.length); i < n; ++i) {
hc = (hc << 2) + ranges[i];
}
return hc;
}
}