[ 
https://issues.apache.org/jira/browse/ORC-554?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16929089#comment-16929089
 ] 

Laszlo Bodor edited comment on ORC-554 at 9/13/19 10:26 AM:
------------------------------------------------------------

meanwhile I realized that float's precision would not be enough for timestamps, 
so it couldn't be supported for timestamps, is it right? [~owen.omalley], 
[~prasanth_j]

I'll close this as invalid

{code}
original timestamp string: 2015-11-29 12:34:56.1
readFloat:

buffer:
[0]     -30     
[1]     -75     
[2]     -84     
[3]     78      

bits: 1319941602
float: 1.44880051E9

result with second and nano loss: 2015-11-29 12:35:12.0


readDouble:

buffer:
[0]     102     
[1]     102     
[2]     6       
[3]     60      
[4]     -68     
[5]     -106    
[6]     -43     
[7]     65      

bits: 4743863517709756006
double: 1.4488004961E9
result: 2015-11-29 12:34:56.1
{code}


was (Author: abstractdog):
meanwhile I realized that float's precision would not be enough for timestamps:

{code}
original timestamp string: 2015-11-29 12:34:56.1
readFloat:

buffer:
[0]     -30     
[1]     -75     
[2]     -84     
[3]     78      

bits: 1319941602
float: 1.44880051E9

result with second and nano loss: 2015-11-29 12:35:12.0


readDouble:

buffer:
[0]     102     
[1]     102     
[2]     6       
[3]     60      
[4]     -68     
[5]     -106    
[6]     -43     
[7]     65      

bits: 4743863517709756006
double: 1.4488004961E9
result: 2015-11-29 12:34:56.1
{code}

> Float to timestamp schema evolution handles time/nanoseconds incorrectly
> ------------------------------------------------------------------------
>
>                 Key: ORC-554
>                 URL: https://issues.apache.org/jira/browse/ORC-554
>             Project: ORC
>          Issue Type: Bug
>            Reporter: Laszlo Bodor
>            Assignee: Laszlo Bodor
>            Priority: Major
>         Attachments: ORC-554.repro.patch
>
>
> During ORC-539 I found a reproducible issue:  [^ORC-554.repro.patch] 
> seems like float to timestamp schema evolution loses nanoseconds somewhere, 
> moreover, seconds are not correct in the result TimestampColumnVector
> https://issues.apache.org/jira/browse/ORC-539?focusedCommentId=16928308&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-16928308



--
This message was sent by Atlassian Jira
(v8.3.2#803003)

Reply via email to