----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/18396/#review35227 -----------------------------------------------------------
This patch does as follows: * Change ExprNormalizer to not dissect an aggregation function which has a constant parameter. * Change LocalTajoTestingUtility to add fake statistics data to each TableDesc. * It is required to reproduce the error case. * As far as I know, this bug occurs only when one table large and small table is small. * Later, this fake statistics would be useful for making more various conditions. * Add an unit test to reproduce this bug case in TestCaseByCases. In addition, I've cleaned up some of ExprNormalizer as follows: * Add OpType::isFunction(OpType) method to check if a OpType corresponds a function. * Add OpType::isLiteral(OpType) method to check if a OpType corresponds a literal value (i.e., constant value) * Add OpType::isAggregationFunction method to check if a OpType corresponds either GeneralSetFunction or CountRowFunction * Change ExprNormalizer to use the above methods. So, it is simplified a little bit than before. * Change CountRowFunction to inherent GeneralSetFunction instead of Function - Hyunsik Choi On Feb. 22, 2014, 8:05 p.m., Hyunsik Choi wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/18396/ > ----------------------------------------------------------- > > (Updated Feb. 22, 2014, 8:05 p.m.) > > > Review request for Tajo. > > > Bugs: TAJO-619 > https://issues.apache.org/jira/browse/TAJO-619 > > > Repository: tajo > > > Description > ------- > > {code} > select count(1) from table1 join table2 on table1.name = table2.name; > {code} > > When there are constant values in select list and equi-join keys are text > columns, it causes some errors due to schema mismatch. > > Its cause of this bug is as follows: > * the number 1 results in a generated column name ?literal. > * This name is used in both table1 and table2. Output schemas of table1 and > table2 include ?literal > * The input schema of join has only three columns because ?literal is > duplicated name. > > So, its essential bug may be one of following: > * The constant value is evaluated twice. In other words, literal? occurs in > both table1 and table2. > * Schema class does not allow the same column name. > > > Diffs > ----- > > > tajo-algebra/src/main/java/org/apache/tajo/algebra/CountRowsFunctionExpr.java > cf2efb5e6a52a4ba2f1beff2f40a72f6a9932a80 > tajo-algebra/src/main/java/org/apache/tajo/algebra/FunctionExpr.java > b007362691eb9354d397f105d8d7e59350392de4 > > tajo-algebra/src/main/java/org/apache/tajo/algebra/GeneralSetFunctionExpr.java > c45a58f1a9513ddbe43fa9685d917973807a5a52 > tajo-algebra/src/main/java/org/apache/tajo/algebra/OpType.java > 91f322b08810589b1ad6a4add19872e7150667cb > > tajo-core/tajo-core-backend/src/main/java/org/apache/tajo/engine/planner/ExprNormalizer.java > e7de03f10dbec2ab31ebe6d1baf160a2370149c6 > > tajo-core/tajo-core-backend/src/main/java/org/apache/tajo/engine/planner/LogicalPlanner.java > 756ede4e1d58732ae1079e789e9816bdfa4cd4ca > > tajo-core/tajo-core-backend/src/test/java/org/apache/tajo/LocalTajoTestingUtility.java > 50fed7bd3f33a1305d1a1d325a1d4f8096b8d165 > > tajo-core/tajo-core-backend/src/test/java/org/apache/tajo/engine/query/TestCaseByCases.java > 20b5ac5c1d4fb695707149f9de606fc8ec3882dc > > tajo-core/tajo-core-backend/src/test/resources/queries/TestCaseByCases/testTAJO619Case.sql > PRE-CREATION > > tajo-core/tajo-core-backend/src/test/resources/results/TestCaseByCases/testTAJO619Case.result > PRE-CREATION > > Diff: https://reviews.apache.org/r/18396/diff/ > > > Testing > ------- > > > Thanks, > > Hyunsik Choi > >
