Daniel Dai created PIG-4910:
-------------------------------
Summary: Assert wrongly pushed up in optimizer
Key: PIG-4910
URL: https://issues.apache.org/jira/browse/PIG-4910
Project: Pig
Issue Type: Bug
Reporter: Daniel Dai
Assignee: Daniel Dai
Fix For: 0.17.0
The following script fail:
{code}
TEST_DATA = LOAD 'input' USING PigStorage() AS (c1:int);
GR = FOREACH (GROUP TEST_DATA BY c1) GENERATE group as c1,
COUNT_STAR(TEST_DATA) as count, TEST_DATA;
ROWS_WITH_C1_EQUALS_ZERO = FILTER GR BY count > 1L;
ROWS_WITH_C1_EQUALS_ZERO_FLATTENED = FOREACH ROWS_WITH_C1_EQUALS_ZERO GENERATE
FLATTEN($0);
-- Assert shouldn't fail as it should be applied after group by but because
assert is getting pushed to mapper, it is failing.
ASSERT ROWS_WITH_C1_EQUALS_ZERO_FLATTENED BY c1 == 0, 'Should have never seen
this message, assert has a bug.';
DUMP ROWS_WITH_C1_EQUALS_ZERO_FLATTENED;
{code}
input:
0
0
1
The reason is assert is pushed before FILTER:
{code}
ROWS_WITH_C1_EQUALS_ZERO_FLATTENED: (Name: LOStore Schema: c1#14:int)
|
|---ROWS_WITH_C1_EQUALS_ZERO_FLATTENED: (Name: LOForEach Schema: c1#14:int)
| |
| (Name: LOGenerate[true] Schema:
c1#14:int)ColumnPrune:InputUids=[14]ColumnPrune:OutputUids=[14]
| | |
| | c1:(Name: Project Type: int Uid: 14 Input: 0 Column: (*))
| |
| |---(Name: LOInnerLoad[0] Schema: c1#14:int)
|
|---ROWS_WITH_C1_EQUALS_ZERO: (Name: LOFilter Schema:
c1#14:int,count#31:long,ROWS_WITH_C1_EQUALS_ZERO_FLATTENED#29:bag{#30:tuple(c1#14:int)})
| |
| (Name: GreaterThan Type: boolean Uid: 33)
| |
| |---count:(Name: Project Type: long Uid: 31 Input: 0 Column: 1)
| |
| |---(Name: Constant Type: long Uid: 32)
|
|---GR: (Name: LOForEach Schema:
c1#14:int,count#31:long,ROWS_WITH_C1_EQUALS_ZERO_FLATTENED#29:bag{#30:tuple(c1#14:int)})
| |
| (Name: LOGenerate[false,false,false] Schema:
c1#14:int,count#31:long,ROWS_WITH_C1_EQUALS_ZERO_FLATTENED#29:bag{#30:tuple(c1#14:int)})ColumnPrune:InputUids=[29,
14]ColumnPrune:OutputUids=[14, 31]
| | |
| | group:(Name: Project Type: int Uid: 14 Input: 0 Column: (*))
| | |
| | (Name: UserFunc(org.apache.pig.builtin.COUNT_STAR) Type:
long Uid: 31)
| | |
| | |---ROWS_WITH_C1_EQUALS_ZERO_FLATTENED:(Name: Project Type:
bag Uid: 29 Input: 1 Column: (*))
| | |
| | ROWS_WITH_C1_EQUALS_ZERO_FLATTENED:(Name: Project Type: bag
Uid: 29 Input: 2 Column: (*))
| |
| |---(Name: LOInnerLoad[0] Schema: group#14:int)
| |
| |---ROWS_WITH_C1_EQUALS_ZERO_FLATTENED: (Name: LOInnerLoad[1]
Schema: c1#14:int)
| |
| |---ROWS_WITH_C1_EQUALS_ZERO_FLATTENED: (Name: LOInnerLoad[1]
Schema: c1#14:int)
|
|---1-3: (Name: LOCogroup Schema:
group#14:int,ROWS_WITH_C1_EQUALS_ZERO_FLATTENED#29:bag{#44:tuple(c1#14:int)})
| |
| c1:(Name: Project Type: int Uid: 14 Input: 0 Column: 0)
|
|---ROWS_WITH_C1_EQUALS_ZERO_FLATTENED: (Name: LOFilter Schema:
c1#14:int)
| |
| (Name: UserFunc(org.apache.pig.builtin.Assert) Type:
boolean Uid: 40)
| |
| |---(Name: BinCond Type: boolean Uid: 38)
| | |
| | |---(Name: Equal Type: boolean Uid: 35)
| | | |
| | | |---c1:(Name: Project Type: int Uid: 14 Input:
0 Column: 0)
| | | |
| | | |---(Name: Constant Type: int Uid: 34)
| | |
| | |---(Name: Constant Type: boolean Uid: 36)
| | |
| | |---(Name: Constant Type: boolean Uid: 37)
| |
| |---(Name: Constant Type: chararray Uid: 39)
|
|---TEST_DATA: (Name: LOForEach Schema: c1#14:int)
| |
| (Name: LOGenerate[false] Schema:
c1#14:int)ColumnPrune:InputUids=[14]ColumnPrune:OutputUids=[14]
| | |
| | (Name: Cast Type: int Uid: 14)
| | |
| | |---c1:(Name: Project Type: bytearray Uid: 14
Input: 0 Column: (*))
| |
| |---(Name: LOInnerLoad[0] Schema: c1#14:bytearray)
|
|---TEST_DATA: (Name: LOLoad Schema:
c1#14:bytearray)RequiredFields:null
{code}
Runs fine by turning off PushUpFilter ("-t PushUpFilter").
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
