[
https://issues.apache.org/jira/browse/PHOENIX-2845?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15262609#comment-15262609
]
Sergey Soldatov commented on PHOENIX-2845:
------------------------------------------
[~jamestaylor] Well, we can add a check that the length is greater than 3 since
Joda time has precision of milliseconds. Joda time adds milliseconds to the
long value. Timestamp keeps all fractions of the second in nanos. For the case
of .1234 it will works in the following way:
1. Joda-time will strip {{4}} and add {{123}} to the final long value.
2. Timestamp constructor will strip {{123}} from the long value and add
{{123000000}} to the nanos value
3. my code will parse {{1234}} and set nanos value to {{123400000}}
> Timestamp ignores nanos in literal expressions
> ----------------------------------------------
>
> Key: PHOENIX-2845
> URL: https://issues.apache.org/jira/browse/PHOENIX-2845
> Project: Phoenix
> Issue Type: Bug
> Affects Versions: 4.7.0
> Reporter: Sergey Soldatov
> Assignee: Sergey Soldatov
> Attachments: PHOENIX-2845-1.patch, PHOENIX-2845-2.patch
>
>
> Phoenix strips nanos in Timestamp. Simple test case:
> {noformat}
> create table x (id bigint primary key, t timestamp);
> upsert into x values (1, TIMESTAMP'2015-01-01 00:00:00.123456');
> upsert into x values (2, TIMESTAMP'2015-01-01 00:00:00.123457');
> upsert into x values (3, TIMESTAMP'2015-01-01 00:00:00.123999');
> {noformat}
> HBase scan:
> {noformat}
> hbase(main):007:0> scan 'X'
> ROW COLUMN+CELL
> \x80\x00\x00\x00\x00\x00\x00\x01 column=0:T, timestamp=1460646668779,
> value=\x80\x00\x01J\xA2\xCA\xB0{\x00\x00\x00\x00
> \x80\x00\x00\x00\x00\x00\x00\x02 column=0:T, timestamp=1460646673369,
> value=\x80\x00\x01J\xA2\xCA\xB0{\x00\x00\x00\x00
> \x80\x00\x00\x00\x00\x00\x00\x03 column=0:T, timestamp=1460646680294,
> value=\x80\x00\x01J\xA2\xCA\xB0{\x00\x00\x00\x00
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
