Github user JamesRTaylor commented on a diff in the pull request: https://github.com/apache/phoenix/pull/208#discussion_r78280999 --- Diff: phoenix-core/src/main/java/org/apache/phoenix/compile/ExpressionCompiler.java --- @@ -523,7 +523,12 @@ public Expression visitLeave(LikeParseNode node, List<Expression> children) thro byte[] wildcard = {StringUtil.MULTI_CHAR_LIKE}; StringUtil.fill(nullExpressionString, 0, pattern.length(), wildcard, 0, 1, false); if (pattern.equals(new String (nullExpressionString))) { - return IsNullExpression.create(lhs, true, context.getTempPtr()); --- End diff -- I think the original code, optimizing to COL IS NOT NULL is not correct, because if COL is null, it'll evaluate to false when it should evaluate to null. Instead, I think it should become COL >= CAST(KeyRange.IS_NOT_NULL_RANGE AS VARCHAR) and for the negate case, it'd be COL < CAST(KeyRange.IS_NOT_NULL_RANGE AS VARCHAR). That way in any case, if COL is null, the expression will evaluate to null. If COL is any string then the first expression will evaluate to true and the latter to false. The code will look something like this: // Declare this as a static constant at top of file private static final Expression NOT_NULL_STRING = LiteralExpression.newConstant(PVarchar.INSTANCE.toObject(KeyRange.IS_NOT_NULL_RANGE.getLowerRange())); List<Expression> compareChildren = Arrays.asList(lhs, NOT_NULL_STRING); return new ComparisonExpression.create(compareChildren, node.isNegate() ? CompareOp.LESS : CompareOp.GREATER_OR_EQUAL); Then make sure we have tests around COL being null too.
--- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. ---