[jira] [Commented] (HIVE-14077) revert or fix HIVE-13380
[ https://issues.apache.org/jira/browse/HIVE-14077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15939198#comment-15939198 ] Pengcheng Xiong commented on HIVE-14077: [~sershe] and [~xuefuz], will this go into 2.2.0? Thanks > revert or fix HIVE-13380 > > > Key: HIVE-14077 > URL: https://issues.apache.org/jira/browse/HIVE-14077 > Project: Hive > Issue Type: Bug >Reporter: Sergey Shelukhin >Priority: Blocker > > See comments in that JIRA -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (HIVE-14077) revert or fix HIVE-13380
[ https://issues.apache.org/jira/browse/HIVE-14077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15939211#comment-15939211 ] Sergey Shelukhin commented on HIVE-14077: - We reverted it from 2.1, because it breaks TPCDS/H and also ANSI compliance. Some reexamination is necessary, either it should be reverted or fixed. By now I don't remember if there's reason to not revert it, except for b/c (of note is that it was not released in 2.1 so I am not sure what the concern with that was). > revert or fix HIVE-13380 > > > Key: HIVE-14077 > URL: https://issues.apache.org/jira/browse/HIVE-14077 > Project: Hive > Issue Type: Bug >Reporter: Sergey Shelukhin >Priority: Blocker > > See comments in that JIRA -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (HIVE-14077) revert or fix HIVE-13380
[ https://issues.apache.org/jira/browse/HIVE-14077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15941888#comment-15941888 ] Pengcheng Xiong commented on HIVE-14077: Let's move that to 3.0 before you can remember the reason... > revert or fix HIVE-13380 > > > Key: HIVE-14077 > URL: https://issues.apache.org/jira/browse/HIVE-14077 > Project: Hive > Issue Type: Bug >Reporter: Sergey Shelukhin >Priority: Blocker > > See comments in that JIRA -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (HIVE-14077) revert or fix HIVE-13380
[ https://issues.apache.org/jira/browse/HIVE-14077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15942076#comment-15942076 ] Sergey Shelukhin commented on HIVE-14077: - I don't remember the reason for objections. The patch breaks queries in benchmarks and also ANSI. So by default, we should revert. > revert or fix HIVE-13380 > > > Key: HIVE-14077 > URL: https://issues.apache.org/jira/browse/HIVE-14077 > Project: Hive > Issue Type: Bug >Reporter: Sergey Shelukhin >Priority: Blocker > > See comments in that JIRA -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (HIVE-14077) revert or fix HIVE-13380
[ https://issues.apache.org/jira/browse/HIVE-14077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15942077#comment-15942077 ] Pengcheng Xiong commented on HIVE-14077: [~sershe], is there anybody working on this? > revert or fix HIVE-13380 > > > Key: HIVE-14077 > URL: https://issues.apache.org/jira/browse/HIVE-14077 > Project: Hive > Issue Type: Bug >Reporter: Sergey Shelukhin >Priority: Blocker > > See comments in that JIRA -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (HIVE-14077) revert or fix HIVE-13380
[ https://issues.apache.org/jira/browse/HIVE-14077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15942080#comment-15942080 ] Sergey Shelukhin commented on HIVE-14077: - I'll take a look the coming week > revert or fix HIVE-13380 > > > Key: HIVE-14077 > URL: https://issues.apache.org/jira/browse/HIVE-14077 > Project: Hive > Issue Type: Bug >Reporter: Sergey Shelukhin >Assignee: Sergey Shelukhin >Priority: Blocker > > See comments in that JIRA -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (HIVE-14077) revert or fix HIVE-13380
[ https://issues.apache.org/jira/browse/HIVE-14077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15943428#comment-15943428 ] Carter Shanklin commented on HIVE-14077: FWIW SQL standard states any precise numeric type has higher precedence than any approximate numeric type In the SQL:2011 spec see section 9.7, Syntax Rule 9b. So I'm in favor of reverting. > revert or fix HIVE-13380 > > > Key: HIVE-14077 > URL: https://issues.apache.org/jira/browse/HIVE-14077 > Project: Hive > Issue Type: Bug >Reporter: Sergey Shelukhin >Assignee: Sergey Shelukhin >Priority: Blocker > > See comments in that JIRA -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (HIVE-14077) revert or fix HIVE-13380
[ https://issues.apache.org/jira/browse/HIVE-14077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15944193#comment-15944193 ] Sergey Shelukhin commented on HIVE-14077: - I've read the original discussions on multiple JIRAs... I'm going to see if the combination of HIVE-13380 and HIVE-13945 produces the correct behavior for literals. [~cartershanklin] does it apply to literals, or also in other cases e.g. arithmetic? cc [~ashutoshc] > revert or fix HIVE-13380 > > > Key: HIVE-14077 > URL: https://issues.apache.org/jira/browse/HIVE-14077 > Project: Hive > Issue Type: Bug >Reporter: Sergey Shelukhin >Assignee: Sergey Shelukhin >Priority: Blocker > > See comments in that JIRA -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (HIVE-14077) revert or fix HIVE-13380
[ https://issues.apache.org/jira/browse/HIVE-14077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15944197#comment-15944197 ] Ashutosh Chauhan commented on HIVE-14077: - My assumption this is a non-issue after HIVE-13945 > revert or fix HIVE-13380 > > > Key: HIVE-14077 > URL: https://issues.apache.org/jira/browse/HIVE-14077 > Project: Hive > Issue Type: Bug >Reporter: Sergey Shelukhin >Assignee: Sergey Shelukhin >Priority: Blocker > > See comments in that JIRA -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (HIVE-14077) revert or fix HIVE-13380
[ https://issues.apache.org/jira/browse/HIVE-14077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15944211#comment-15944211 ] Sergey Shelukhin commented on HIVE-14077: - May be an issue, I thought decimal + double in explicitly typed case should go to double, but I'm not sure based on the above comment > revert or fix HIVE-13380 > > > Key: HIVE-14077 > URL: https://issues.apache.org/jira/browse/HIVE-14077 > Project: Hive > Issue Type: Bug >Reporter: Sergey Shelukhin >Assignee: Sergey Shelukhin >Priority: Blocker > > See comments in that JIRA -- This message was sent by Atlassian JIRA (v6.3.15#6346)
[jira] [Commented] (HIVE-14077) revert or fix HIVE-13380
[ https://issues.apache.org/jira/browse/HIVE-14077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15345212#comment-15345212 ] Ashutosh Chauhan commented on HIVE-14077: - I have noted this on HIVE-13380, pasting it here again on course of action I am suggesting: bq. I think correct course of action here is to treat literal constants as big decimal instead of current choice of double. That will get us consistent behavior. I checked on MySQL and its doing the same thing. > revert or fix HIVE-13380 > > > Key: HIVE-14077 > URL: https://issues.apache.org/jira/browse/HIVE-14077 > Project: Hive > Issue Type: Bug >Reporter: Sergey Shelukhin >Priority: Blocker > > See comments in that JIRA -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (HIVE-14077) revert or fix HIVE-13380
[ https://issues.apache.org/jira/browse/HIVE-14077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15345221#comment-15345221 ] Sergey Shelukhin commented on HIVE-14077: - Heh, I am about to do that as part of some other bugfix > revert or fix HIVE-13380 > > > Key: HIVE-14077 > URL: https://issues.apache.org/jira/browse/HIVE-14077 > Project: Hive > Issue Type: Bug >Reporter: Sergey Shelukhin >Priority: Blocker > > See comments in that JIRA -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (HIVE-14077) revert or fix HIVE-13380
[ https://issues.apache.org/jira/browse/HIVE-14077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15345490#comment-15345490 ] Sergey Shelukhin commented on HIVE-14077: - I'll use this JIRA to add a test. > revert or fix HIVE-13380 > > > Key: HIVE-14077 > URL: https://issues.apache.org/jira/browse/HIVE-14077 > Project: Hive > Issue Type: Bug >Reporter: Sergey Shelukhin >Priority: Blocker > > See comments in that JIRA -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (HIVE-14077) revert or fix HIVE-13380
[ https://issues.apache.org/jira/browse/HIVE-14077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15375512#comment-15375512 ] Xuefu Zhang commented on HIVE-14077: Sorry for chiming in late, but I'm wonder what's focal debate here? As [~jdere] pointed out, Hive has a specific decimal syntax, and a numeric such as 0.06 is treated as double. I don't see anything wrong with that. Making 0.06 literal as decimal is another concern for b/c. Regarding operation ordering, I think we all agree that nonexact type operating with exact type produces nonexact time. This should apply to comparison as well. However, b/c is a valid concern and should be called out as pointed out in HIVE-13380. > revert or fix HIVE-13380 > > > Key: HIVE-14077 > URL: https://issues.apache.org/jira/browse/HIVE-14077 > Project: Hive > Issue Type: Bug >Reporter: Sergey Shelukhin >Priority: Blocker > > See comments in that JIRA -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (HIVE-14077) revert or fix HIVE-13380
[ https://issues.apache.org/jira/browse/HIVE-14077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15375515#comment-15375515 ] Sergey Shelukhin commented on HIVE-14077: - Well, per ANSI literals like 0.06 should be treated as decimal. Treating them double in filters also causes surprising query results. In fact, HIVE-13945 changed the treatment on master. I think the idea of this JIRA was now to add a test. We also need to check that all the active branches either have both or neither of these patches, for consistency. > revert or fix HIVE-13380 > > > Key: HIVE-14077 > URL: https://issues.apache.org/jira/browse/HIVE-14077 > Project: Hive > Issue Type: Bug >Reporter: Sergey Shelukhin >Priority: Blocker > > See comments in that JIRA -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (HIVE-14077) revert or fix HIVE-13380
[ https://issues.apache.org/jira/browse/HIVE-14077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15375524#comment-15375524 ] Xuefu Zhang commented on HIVE-14077: Thanks for the clarification. Regarding ANSI, we don't claim that Hive is ANSI compliant. Hive is already treating 0.06 as double always, and I'd say b/c is more important than a motivation to make it ANSI-compliant. Again, Hive uses 0.06BD as decimal literal. > revert or fix HIVE-13380 > > > Key: HIVE-14077 > URL: https://issues.apache.org/jira/browse/HIVE-14077 > Project: Hive > Issue Type: Bug >Reporter: Sergey Shelukhin >Priority: Blocker > > See comments in that JIRA -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (HIVE-14077) revert or fix HIVE-13380
[ https://issues.apache.org/jira/browse/HIVE-14077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15375529#comment-15375529 ] Sergey Shelukhin commented on HIVE-14077: - Well, HIVE-13945 has the motivation to change it... :) And there are other similar issues (that's why we reverted HIVE-13380 from 2.1 before 2.1 release - it broke some TPCDS/H queries due to double arithmetic). Also I thought Hive does strive to be ANSI compliant? :) Wrt special form, that is preserved, I also added 0.06D for double. > revert or fix HIVE-13380 > > > Key: HIVE-14077 > URL: https://issues.apache.org/jira/browse/HIVE-14077 > Project: Hive > Issue Type: Bug >Reporter: Sergey Shelukhin >Priority: Blocker > > See comments in that JIRA -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (HIVE-14077) revert or fix HIVE-13380
[ https://issues.apache.org/jira/browse/HIVE-14077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15375902#comment-15375902 ] Xuefu Zhang commented on HIVE-14077: RE. Also I thought Hive does strive to be ANSI compliant Yeah, but there are limits on what we can do, especially when b/c is a concern. > revert or fix HIVE-13380 > > > Key: HIVE-14077 > URL: https://issues.apache.org/jira/browse/HIVE-14077 > Project: Hive > Issue Type: Bug >Reporter: Sergey Shelukhin >Priority: Blocker > > See comments in that JIRA -- This message was sent by Atlassian JIRA (v6.3.4#6332)