[jira] [Commented] (HIVE-15331) Decimal multiplication with high precision/scale often returns NULL
[
https://issues.apache.org/jira/browse/HIVE-15331?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15752735#comment-15752735
]
Sergey Shelukhin commented on HIVE-15331:
-
Decimal_precision:
{noformat}
-1234.56789012351524157.87532399036884525225
+1234.56789012351524157.87532399036884525
{noformat}
> Decimal multiplication with high precision/scale often returns NULL
> ---
>
> Key: HIVE-15331
> URL: https://issues.apache.org/jira/browse/HIVE-15331
> Project: Hive
> Issue Type: Bug
> Components: Types
>Reporter: Jason Dere
>Assignee: Jason Dere
> Attachments: HIVE-15331.1.patch, HIVE-15331.2.patch,
> HIVE-15331.3.patch
>
>
> {noformat}
> create temporary table dec (a decimal(38,18));
> insert into dec values(100.0);
> hive> select a*a from dec;
> OK
> NULL
> Time taken: 0.165 seconds, Fetched: 1 row(s)
> {noformat}
> Looks like the reason is because the result of decimal(38,18) *
> decimal(38,18) only has 2 digits of precision for integers:
> {noformat}
> hive> set hive.explain.user=false;
> hive> explain select a*a from dec;
> OK
> STAGE DEPENDENCIES:
> Stage-0 is a root stage
> STAGE PLANS:
> Stage: Stage-0
> Fetch Operator
> limit: -1
> Processor Tree:
> TableScan
> alias: dec
> Select Operator
> expressions: (a * a) (type: decimal(38,36))
> outputColumnNames: _col0
> ListSink
> Time taken: 0.039 seconds, Fetched: 15 row(s)
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (HIVE-15331) Decimal multiplication with high precision/scale often returns NULL
[
https://issues.apache.org/jira/browse/HIVE-15331?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15752726#comment-15752726
]
Jason Dere commented on HIVE-15331:
---
Which file in particular? These are all cases where the resulting
precision/scale exceeds 38 - for example decimal_precision.q.out, the operation
is decimal(20,10) * decimal(20,10), the precision/scale (assuming unlimited
precision/scale) would be decimal(41,20), however we're already at the Hive
limit of 38.
So this is where the updated logic would kick in terms of making the
precision/scale fit within 38 digits.
> Decimal multiplication with high precision/scale often returns NULL
> ---
>
> Key: HIVE-15331
> URL: https://issues.apache.org/jira/browse/HIVE-15331
> Project: Hive
> Issue Type: Bug
> Components: Types
>Reporter: Jason Dere
>Assignee: Jason Dere
> Attachments: HIVE-15331.1.patch, HIVE-15331.2.patch,
> HIVE-15331.3.patch
>
>
> {noformat}
> create temporary table dec (a decimal(38,18));
> insert into dec values(100.0);
> hive> select a*a from dec;
> OK
> NULL
> Time taken: 0.165 seconds, Fetched: 1 row(s)
> {noformat}
> Looks like the reason is because the result of decimal(38,18) *
> decimal(38,18) only has 2 digits of precision for integers:
> {noformat}
> hive> set hive.explain.user=false;
> hive> explain select a*a from dec;
> OK
> STAGE DEPENDENCIES:
> Stage-0 is a root stage
> STAGE PLANS:
> Stage: Stage-0
> Fetch Operator
> limit: -1
> Processor Tree:
> TableScan
> alias: dec
> Select Operator
> expressions: (a * a) (type: decimal(38,36))
> outputColumnNames: _col0
> ListSink
> Time taken: 0.039 seconds, Fetched: 15 row(s)
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (HIVE-15331) Decimal multiplication with high precision/scale often returns NULL
[
https://issues.apache.org/jira/browse/HIVE-15331?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15752628#comment-15752628
]
Sergey Shelukhin commented on HIVE-15331:
-
Looks good; it seems like it reduces precision even in cases where it's not
needed though... is that intended? See out file changes, where last few
(non-zero) digits of the result are rounded away
> Decimal multiplication with high precision/scale often returns NULL
> ---
>
> Key: HIVE-15331
> URL: https://issues.apache.org/jira/browse/HIVE-15331
> Project: Hive
> Issue Type: Bug
> Components: Types
>Reporter: Jason Dere
>Assignee: Jason Dere
> Attachments: HIVE-15331.1.patch, HIVE-15331.2.patch,
> HIVE-15331.3.patch
>
>
> {noformat}
> create temporary table dec (a decimal(38,18));
> insert into dec values(100.0);
> hive> select a*a from dec;
> OK
> NULL
> Time taken: 0.165 seconds, Fetched: 1 row(s)
> {noformat}
> Looks like the reason is because the result of decimal(38,18) *
> decimal(38,18) only has 2 digits of precision for integers:
> {noformat}
> hive> set hive.explain.user=false;
> hive> explain select a*a from dec;
> OK
> STAGE DEPENDENCIES:
> Stage-0 is a root stage
> STAGE PLANS:
> Stage: Stage-0
> Fetch Operator
> limit: -1
> Processor Tree:
> TableScan
> alias: dec
> Select Operator
> expressions: (a * a) (type: decimal(38,36))
> outputColumnNames: _col0
> ListSink
> Time taken: 0.039 seconds, Fetched: 15 row(s)
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (HIVE-15331) Decimal multiplication with high precision/scale often returns NULL
[
https://issues.apache.org/jira/browse/HIVE-15331?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15752568#comment-15752568
]
Jason Dere commented on HIVE-15331:
---
[~sershe] can you review?
> Decimal multiplication with high precision/scale often returns NULL
> ---
>
> Key: HIVE-15331
> URL: https://issues.apache.org/jira/browse/HIVE-15331
> Project: Hive
> Issue Type: Bug
> Components: Types
>Reporter: Jason Dere
>Assignee: Jason Dere
> Attachments: HIVE-15331.1.patch, HIVE-15331.2.patch,
> HIVE-15331.3.patch
>
>
> {noformat}
> create temporary table dec (a decimal(38,18));
> insert into dec values(100.0);
> hive> select a*a from dec;
> OK
> NULL
> Time taken: 0.165 seconds, Fetched: 1 row(s)
> {noformat}
> Looks like the reason is because the result of decimal(38,18) *
> decimal(38,18) only has 2 digits of precision for integers:
> {noformat}
> hive> set hive.explain.user=false;
> hive> explain select a*a from dec;
> OK
> STAGE DEPENDENCIES:
> Stage-0 is a root stage
> STAGE PLANS:
> Stage: Stage-0
> Fetch Operator
> limit: -1
> Processor Tree:
> TableScan
> alias: dec
> Select Operator
> expressions: (a * a) (type: decimal(38,36))
> outputColumnNames: _col0
> ListSink
> Time taken: 0.039 seconds, Fetched: 15 row(s)
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (HIVE-15331) Decimal multiplication with high precision/scale often returns NULL
[
https://issues.apache.org/jira/browse/HIVE-15331?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15732920#comment-15732920
]
Jason Dere commented on HIVE-15331:
---
I think this is ready for review now.
> Decimal multiplication with high precision/scale often returns NULL
> ---
>
> Key: HIVE-15331
> URL: https://issues.apache.org/jira/browse/HIVE-15331
> Project: Hive
> Issue Type: Bug
> Components: Types
>Reporter: Jason Dere
>Assignee: Jason Dere
> Attachments: HIVE-15331.1.patch, HIVE-15331.2.patch,
> HIVE-15331.3.patch
>
>
> {noformat}
> create temporary table dec (a decimal(38,18));
> insert into dec values(100.0);
> hive> select a*a from dec;
> OK
> NULL
> Time taken: 0.165 seconds, Fetched: 1 row(s)
> {noformat}
> Looks like the reason is because the result of decimal(38,18) *
> decimal(38,18) only has 2 digits of precision for integers:
> {noformat}
> hive> set hive.explain.user=false;
> hive> explain select a*a from dec;
> OK
> STAGE DEPENDENCIES:
> Stage-0 is a root stage
> STAGE PLANS:
> Stage: Stage-0
> Fetch Operator
> limit: -1
> Processor Tree:
> TableScan
> alias: dec
> Select Operator
> expressions: (a * a) (type: decimal(38,36))
> outputColumnNames: _col0
> ListSink
> Time taken: 0.039 seconds, Fetched: 15 row(s)
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (HIVE-15331) Decimal multiplication with high precision/scale often returns NULL
[
https://issues.apache.org/jira/browse/HIVE-15331?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15731436#comment-15731436
]
Hive QA commented on HIVE-15331:
Here are the results of testing the latest attachment:
https://issues.apache.org/jira/secure/attachment/12842239/HIVE-15331.3.patch
{color:green}SUCCESS:{color} +1 due to 2 test(s) being added or modified.
{color:red}ERROR:{color} -1 due to 9 failed/errored test(s), 10782 tests
executed
*Failed tests:*
{noformat}
org.apache.hadoop.hive.cli.TestCliDriver.testCliDriver[sample2] (batchId=5)
org.apache.hadoop.hive.cli.TestCliDriver.testCliDriver[sample4] (batchId=15)
org.apache.hadoop.hive.cli.TestCliDriver.testCliDriver[sample6] (batchId=61)
org.apache.hadoop.hive.cli.TestCliDriver.testCliDriver[sample7] (batchId=60)
org.apache.hadoop.hive.cli.TestCliDriver.testCliDriver[sample9] (batchId=38)
org.apache.hadoop.hive.cli.TestMiniLlapCliDriver.testCliDriver[orc_ppd_schema_evol_3a]
(batchId=134)
org.apache.hadoop.hive.cli.TestMiniLlapCliDriver.testCliDriver[transform_ppr2]
(batchId=134)
org.apache.hadoop.hive.cli.TestMiniLlapLocalCliDriver.testCliDriver[stats_based_fetch_decision]
(batchId=150)
org.apache.hadoop.hive.cli.TestMiniTezCliDriver.testCliDriver[explainanalyze_2]
(batchId=92)
{noformat}
Test results: https://builds.apache.org/job/PreCommit-HIVE-Build/2481/testReport
Console output: https://builds.apache.org/job/PreCommit-HIVE-Build/2481/console
Test logs: http://104.198.109.242/logs/PreCommit-HIVE-Build-2481/
Messages:
{noformat}
Executing org.apache.hive.ptest.execution.TestCheckPhase
Executing org.apache.hive.ptest.execution.PrepPhase
Executing org.apache.hive.ptest.execution.ExecutionPhase
Executing org.apache.hive.ptest.execution.ReportingPhase
Tests exited with: TestsFailedException: 9 tests failed
{noformat}
This message is automatically generated.
ATTACHMENT ID: 12842239 - PreCommit-HIVE-Build
> Decimal multiplication with high precision/scale often returns NULL
> ---
>
> Key: HIVE-15331
> URL: https://issues.apache.org/jira/browse/HIVE-15331
> Project: Hive
> Issue Type: Bug
> Components: Types
>Reporter: Jason Dere
>Assignee: Jason Dere
> Attachments: HIVE-15331.1.patch, HIVE-15331.2.patch,
> HIVE-15331.3.patch
>
>
> {noformat}
> create temporary table dec (a decimal(38,18));
> insert into dec values(100.0);
> hive> select a*a from dec;
> OK
> NULL
> Time taken: 0.165 seconds, Fetched: 1 row(s)
> {noformat}
> Looks like the reason is because the result of decimal(38,18) *
> decimal(38,18) only has 2 digits of precision for integers:
> {noformat}
> hive> set hive.explain.user=false;
> hive> explain select a*a from dec;
> OK
> STAGE DEPENDENCIES:
> Stage-0 is a root stage
> STAGE PLANS:
> Stage: Stage-0
> Fetch Operator
> limit: -1
> Processor Tree:
> TableScan
> alias: dec
> Select Operator
> expressions: (a * a) (type: decimal(38,36))
> outputColumnNames: _col0
> ListSink
> Time taken: 0.039 seconds, Fetched: 15 row(s)
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (HIVE-15331) Decimal multiplication with high precision/scale often returns NULL
[
https://issues.apache.org/jira/browse/HIVE-15331?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15726941#comment-15726941
]
Hive QA commented on HIVE-15331:
Here are the results of testing the latest attachment:
https://issues.apache.org/jira/secure/attachment/12842017/HIVE-15331.2.patch
{color:green}SUCCESS:{color} +1 due to 2 test(s) being added or modified.
{color:red}ERROR:{color} -1 due to 16 failed/errored test(s), 10775 tests
executed
*Failed tests:*
{noformat}
org.apache.hadoop.hive.cli.TestCliDriver.testCliDriver[auto_sortmerge_join_2]
(batchId=44)
org.apache.hadoop.hive.cli.TestCliDriver.testCliDriver[decimal_precision]
(batchId=47)
org.apache.hadoop.hive.cli.TestCliDriver.testCliDriver[decimal_udf] (batchId=8)
org.apache.hadoop.hive.cli.TestCliDriver.testCliDriver[sample2] (batchId=5)
org.apache.hadoop.hive.cli.TestCliDriver.testCliDriver[sample4] (batchId=15)
org.apache.hadoop.hive.cli.TestCliDriver.testCliDriver[sample6] (batchId=61)
org.apache.hadoop.hive.cli.TestCliDriver.testCliDriver[sample7] (batchId=60)
org.apache.hadoop.hive.cli.TestCliDriver.testCliDriver[sample9] (batchId=38)
org.apache.hadoop.hive.cli.TestCliDriver.testCliDriver[vector_decimal_expressions]
(batchId=48)
org.apache.hadoop.hive.cli.TestCliDriver.testCliDriver[vector_decimal_precision]
(batchId=45)
org.apache.hadoop.hive.cli.TestMiniLlapCliDriver.testCliDriver[transform_ppr2]
(batchId=134)
org.apache.hadoop.hive.cli.TestMiniLlapLocalCliDriver.testCliDriver[stats_based_fetch_decision]
(batchId=150)
org.apache.hadoop.hive.cli.TestMiniLlapLocalCliDriver.testCliDriver[vector_decimal_expressions]
(batchId=146)
org.apache.hadoop.hive.cli.TestMiniLlapLocalCliDriver.testCliDriver[vector_decimal_precision]
(batchId=146)
org.apache.hadoop.hive.cli.TestMiniLlapLocalCliDriver.testCliDriver[vector_decimal_udf]
(batchId=152)
org.apache.hadoop.hive.cli.TestMiniTezCliDriver.testCliDriver[explainanalyze_2]
(batchId=92)
{noformat}
Test results: https://builds.apache.org/job/PreCommit-HIVE-Build/2450/testReport
Console output: https://builds.apache.org/job/PreCommit-HIVE-Build/2450/console
Test logs: http://104.198.109.242/logs/PreCommit-HIVE-Build-2450/
Messages:
{noformat}
Executing org.apache.hive.ptest.execution.TestCheckPhase
Executing org.apache.hive.ptest.execution.PrepPhase
Executing org.apache.hive.ptest.execution.ExecutionPhase
Executing org.apache.hive.ptest.execution.ReportingPhase
Tests exited with: TestsFailedException: 16 tests failed
{noformat}
This message is automatically generated.
ATTACHMENT ID: 12842017 - PreCommit-HIVE-Build
> Decimal multiplication with high precision/scale often returns NULL
> ---
>
> Key: HIVE-15331
> URL: https://issues.apache.org/jira/browse/HIVE-15331
> Project: Hive
> Issue Type: Bug
> Components: Types
>Reporter: Jason Dere
>Assignee: Jason Dere
> Attachments: HIVE-15331.1.patch, HIVE-15331.2.patch
>
>
> {noformat}
> create temporary table dec (a decimal(38,18));
> insert into dec values(100.0);
> hive> select a*a from dec;
> OK
> NULL
> Time taken: 0.165 seconds, Fetched: 1 row(s)
> {noformat}
> Looks like the reason is because the result of decimal(38,18) *
> decimal(38,18) only has 2 digits of precision for integers:
> {noformat}
> hive> set hive.explain.user=false;
> hive> explain select a*a from dec;
> OK
> STAGE DEPENDENCIES:
> Stage-0 is a root stage
> STAGE PLANS:
> Stage: Stage-0
> Fetch Operator
> limit: -1
> Processor Tree:
> TableScan
> alias: dec
> Select Operator
> expressions: (a * a) (type: decimal(38,36))
> outputColumnNames: _col0
> ListSink
> Time taken: 0.039 seconds, Fetched: 15 row(s)
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (HIVE-15331) Decimal multiplication with high precision/scale often returns NULL
[
https://issues.apache.org/jira/browse/HIVE-15331?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15713158#comment-15713158
]
Jason Dere commented on HIVE-15331:
---
It cannot be bullet proof - even the Msft blogpost admits this. However I do
think that we can do a little better here than the existing behavior, which
punts multiplication of any value over 100.
> Decimal multiplication with high precision/scale often returns NULL
> ---
>
> Key: HIVE-15331
> URL: https://issues.apache.org/jira/browse/HIVE-15331
> Project: Hive
> Issue Type: Bug
> Components: Types
>Reporter: Jason Dere
>Assignee: Jason Dere
>
> {noformat}
> create temporary table dec (a decimal(38,18));
> insert into dec values(100.0);
> hive> select a*a from dec;
> OK
> NULL
> Time taken: 0.165 seconds, Fetched: 1 row(s)
> {noformat}
> Looks like the reason is because the result of decimal(38,18) *
> decimal(38,18) only has 2 digits of precision for integers:
> {noformat}
> hive> set hive.explain.user=false;
> hive> explain select a*a from dec;
> OK
> STAGE DEPENDENCIES:
> Stage-0 is a root stage
> STAGE PLANS:
> Stage: Stage-0
> Fetch Operator
> limit: -1
> Processor Tree:
> TableScan
> alias: dec
> Select Operator
> expressions: (a * a) (type: decimal(38,36))
> outputColumnNames: _col0
> ListSink
> Time taken: 0.039 seconds, Fetched: 15 row(s)
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (HIVE-15331) Decimal multiplication with high precision/scale often returns NULL
[
https://issues.apache.org/jira/browse/HIVE-15331?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15713110#comment-15713110
]
Xuefu Zhang commented on HIVE-15331:
I dont' think this can be made bullet proof. I can always construct an example
that shows a rule that fails to accommodate.
> Decimal multiplication with high precision/scale often returns NULL
> ---
>
> Key: HIVE-15331
> URL: https://issues.apache.org/jira/browse/HIVE-15331
> Project: Hive
> Issue Type: Bug
> Components: Types
>Reporter: Jason Dere
>Assignee: Jason Dere
>
> {noformat}
> create temporary table dec (a decimal(38,18));
> insert into dec values(100.0);
> hive> select a*a from dec;
> OK
> NULL
> Time taken: 0.165 seconds, Fetched: 1 row(s)
> {noformat}
> Looks like the reason is because the result of decimal(38,18) *
> decimal(38,18) only has 2 digits of precision for integers:
> {noformat}
> hive> set hive.explain.user=false;
> hive> explain select a*a from dec;
> OK
> STAGE DEPENDENCIES:
> Stage-0 is a root stage
> STAGE PLANS:
> Stage: Stage-0
> Fetch Operator
> limit: -1
> Processor Tree:
> TableScan
> alias: dec
> Select Operator
> expressions: (a * a) (type: decimal(38,36))
> outputColumnNames: _col0
> ListSink
> Time taken: 0.039 seconds, Fetched: 15 row(s)
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (HIVE-15331) Decimal multiplication with high precision/scale often returns NULL
[
https://issues.apache.org/jira/browse/HIVE-15331?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15713011#comment-15713011
]
Jason Dere commented on HIVE-15331:
---
Yes the user ideally should qualify the type, but they often don't. And with
the way this works now decimal multiplication is pretty much unusable without
casts.
The precision/scale rules for division look like they have been changed because
of similar issues (/HIVE-5866). We could try to match what was done there in
multiplication, but I think it would be better to try to emulate the SQL Server
behavior given that we were using their precision/scale rules
> Decimal multiplication with high precision/scale often returns NULL
> ---
>
> Key: HIVE-15331
> URL: https://issues.apache.org/jira/browse/HIVE-15331
> Project: Hive
> Issue Type: Bug
> Components: Types
>Reporter: Jason Dere
>Assignee: Jason Dere
>
> {noformat}
> create temporary table dec (a decimal(38,18));
> insert into dec values(100.0);
> hive> select a*a from dec;
> OK
> NULL
> Time taken: 0.165 seconds, Fetched: 1 row(s)
> {noformat}
> Looks like the reason is because the result of decimal(38,18) *
> decimal(38,18) only has 2 digits of precision for integers:
> {noformat}
> hive> set hive.explain.user=false;
> hive> explain select a*a from dec;
> OK
> STAGE DEPENDENCIES:
> Stage-0 is a root stage
> STAGE PLANS:
> Stage: Stage-0
> Fetch Operator
> limit: -1
> Processor Tree:
> TableScan
> alias: dec
> Select Operator
> expressions: (a * a) (type: decimal(38,36))
> outputColumnNames: _col0
> ListSink
> Time taken: 0.039 seconds, Fetched: 15 row(s)
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (HIVE-15331) Decimal multiplication with high precision/scale often returns NULL
[
https://issues.apache.org/jira/browse/HIVE-15331?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15712964#comment-15712964
]
Xuefu Zhang commented on HIVE-15331:
There are more than a couple of JIRAs on this topic. I think it can be argued
both ways regarding which (int part vs dec part) to favor. As pointed by the
blog post, user needs to quality the type matching the actually data. When not
possible, use double/float instead.
What is missing in Hive is a mode in which error is thrown instead of null, but
that's not decimal type specific.
> Decimal multiplication with high precision/scale often returns NULL
> ---
>
> Key: HIVE-15331
> URL: https://issues.apache.org/jira/browse/HIVE-15331
> Project: Hive
> Issue Type: Bug
> Components: Types
>Reporter: Jason Dere
>Assignee: Jason Dere
>
> {noformat}
> create temporary table dec (a decimal(38,18));
> insert into dec values(100.0);
> hive> select a*a from dec;
> OK
> NULL
> Time taken: 0.165 seconds, Fetched: 1 row(s)
> {noformat}
> Looks like the reason is because the result of decimal(38,18) *
> decimal(38,18) only has 2 digits of precision for integers:
> {noformat}
> hive> set hive.explain.user=false;
> hive> explain select a*a from dec;
> OK
> STAGE DEPENDENCIES:
> Stage-0 is a root stage
> STAGE PLANS:
> Stage: Stage-0
> Fetch Operator
> limit: -1
> Processor Tree:
> TableScan
> alias: dec
> Select Operator
> expressions: (a * a) (type: decimal(38,36))
> outputColumnNames: _col0
> ListSink
> Time taken: 0.039 seconds, Fetched: 15 row(s)
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (HIVE-15331) Decimal multiplication with high precision/scale often returns NULL
[
https://issues.apache.org/jira/browse/HIVE-15331?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15712840#comment-15712840
]
Jason Dere commented on HIVE-15331:
---
While Hive is using similar precision/scale rules as SQL Server for arithmetic
operations (https://msdn.microsoft.com/en-us/library/ms190476.aspx), things
seem to be breaking down once we hit the max precision of 38. Looks like this
sheds a bit of light on how SQL Server handles things in that case:
https://blogs.msdn.microsoft.com/sqlprogrammability/2006/03/29/multiplication-and-division-with-numerics
- basically it gives preference to the integer portion rather than the scale
(Hive does the opposite), and setting a minimum scale of 6.
cc [~xuefuz]
> Decimal multiplication with high precision/scale often returns NULL
> ---
>
> Key: HIVE-15331
> URL: https://issues.apache.org/jira/browse/HIVE-15331
> Project: Hive
> Issue Type: Bug
> Components: Types
>Reporter: Jason Dere
>Assignee: Jason Dere
>
> {noformat}
> create temporary table dec (a decimal(38,18));
> insert into dec values(100.0);
> hive> select a*a from dec;
> OK
> NULL
> Time taken: 0.165 seconds, Fetched: 1 row(s)
> {noformat}
> Looks like the reason is because the result of decimal(38,18) *
> decimal(38,18) only has 2 digits of precision for integers:
> {noformat}
> hive> set hive.explain.user=false;
> hive> explain select a*a from dec;
> OK
> STAGE DEPENDENCIES:
> Stage-0 is a root stage
> STAGE PLANS:
> Stage: Stage-0
> Fetch Operator
> limit: -1
> Processor Tree:
> TableScan
> alias: dec
> Select Operator
> expressions: (a * a) (type: decimal(38,36))
> outputColumnNames: _col0
> ListSink
> Time taken: 0.039 seconds, Fetched: 15 row(s)
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
