Bin Shi created PHOENIX-5176: -------------------------------- Summary: KeyRange.compareUpperRange(KeyRang 1, KeyRang 2) returns wrong result when two key ranges have the same upper bound values but one is inclusive and another is exclusive Key: PHOENIX-5176 URL: https://issues.apache.org/jira/browse/PHOENIX-5176 Project: Phoenix Issue Type: Bug Reporter: Bin Shi
In KeyRange.java, {color:#262626} public static int compareUpperRange(KeyRange rowKeyRange1, KeyRange rowKeyRange2) {{color} {color:#262626} int result = Boolean.compare(rowKeyRange1.upperUnbound(), rowKeyRange2.upperUnbound());{color} {color:#262626} if (result != 0) {{color} {color:#262626} return result;{color} {color:#262626} }{color} {color:#262626} result = Bytes.BYTES_COMPARATOR.compare(rowKeyRange1.getUpperRange(), rowKeyRange2.getUpperRange());{color} {color:#262626} if (result != 0) {{color} {color:#262626} return result;{color} {color:#262626} }{color} {color:#262626} return Boolean.compare(*rowKeyRange2*.isUpperInclusive(), *rowKeyRange1*.isUpperInclusive());{color} {color:#262626} }{color} {color:#262626} {color} {color:#262626}The last line in yellow color should be "{color}return Boolean.compare(*rowKeyRange1*.isUpperInclusive(), *rowKeyRange2*.isUpperInclusive());". Given rowKeyRange1 [3, 5) and rowKeyRange2 [3, 5], the function should return -1, but now it returns 1 due to the bug I mentioned. The KeyRange.compareUpperRange is only used in KeyRange.intersect(List<KeyRange> rowKeyRanges1, List<KeyRange> rowKeyRanges2). Given rowKeyRanges1 \{[3, 5), [5, 6)} and rowKeyRanges2\{[3, 5], [6, 7]}, the function should return \{[3, 5), [5, 5]}, i.e., \{[3, 5]}, but it seems that now it returns \{[3,5)} due to the bug. -- This message was sent by Atlassian JIRA (v7.6.3#76005)