[ 
https://issues.apache.org/jira/browse/OPENJPA-2555?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Ancoron Luciferis updated OPENJPA-2555:
---------------------------------------
    Attachment: openjpa-trunk-Enhance-timestamp-precision-handling.patch
                openjpa-2.3.x-Enhance-timestamp-precision-handling.patch
                openjpa-2.2.x-Enhance-timestamp-precision-handling.patch

Attached a new set of patches:
* For branch 2.2.x: [^openjpa-2.2.x-Enhance-timestamp-precision-handling.patch]
* For branch 2.3.x: [^openjpa-2.3.x-Enhance-timestamp-precision-handling.patch]
* For trunk: [^openjpa-trunk-Enhance-timestamp-precision-handling.patch]

\\
With these new patches I've now also done the following:
# Handle MySQL timestamp precision declaration (it requires a precision to be 
specified), which means that if no JPA {{\@Column(scale = N)}} is specified, 
the default precision being generated in the create table statement will be "6" 
for MySQL server 5.6.4+
# As MySQL before version 5.6.4 does not support *any* timestamp precision, 
I've explicitly disabled precision by changing the 
{{DBDictionary#datePrecision}} field to {{SEC}} (always rounded to seconds)
# Refactored the change introduced by OPENJPA-2476 (introducing JDBC-only logic 
into the kernel doesn't made any sense to me) and split the logic:
#* timestamp rounding logic went into the {{TimestampHelper}} inside Maven 
project "openjpa-lib"
#* version column handling is now done by {{TimestampVersionStrategy}} inside 
Maven project "openjpa-jdbc"
# Modified test {{TestTimestampOptLockEx}} to also run on PostgreSQL and MySQL 
(also account for MySQL version < 5.6.4)
# Introduced test  {{TestTimestampOptLockEx2}} which basically is the same test 
as before, but using a different entity with timestamp precision specified 
using {{\@Column(scale = 4)}} to make sure that non-millisecond rounding is 
correct


> Timestamp precision from manual schema not respected
> ----------------------------------------------------
>
>                 Key: OPENJPA-2555
>                 URL: https://issues.apache.org/jira/browse/OPENJPA-2555
>             Project: OpenJPA
>          Issue Type: Bug
>          Components: jdbc, jpa, sql
>    Affects Versions: 2.2.2, 2.3.0
>            Reporter: Ancoron Luciferis
>             Fix For: 2.2.3, 2.3.1, 2.4.0
>
>         Attachments: 2.2.x-Enable-timestamp-precision-handling.patch, 
> 2.3.x-Enable-timestamp-precision-handling.patch, 
> openjpa-2.2.x-Enhance-timestamp-precision-handling.patch, 
> openjpa-2.3.x-Enhance-timestamp-precision-handling.patch, 
> openjpa-trunk-Enhance-timestamp-precision-handling.patch, 
> trunk-Enable-timestamp-precision-handling.patch
>
>
> The use cases here are the following:
> # JPA entities are to-be-created for an existing database schema which 
> includes several timestamp columns with explicit precision
> # A developer wants to specify timestamp precision inside JPA entities to 
> better specify column data type information for the generated schema
> \\
> In both cases, the result will be that any query executed for a timestamp 
> column that is configured for less than millisecond precision (e.g. deci- or 
> centi-seconds) will fail to find appropriate rows.
> One of the reasons for that is that the precision used for rounding a 
> timestamp value before it goes into a query is configured for a whole 
> database type (using the dictionary) or the whole persistence context (using 
> the configuration parameter).
> This makes it impossible to have different column configurations, e.g. some 
> without any precision declaration (where it's not important) but some with.
> In addition, the default precision for the standard timestamp data type is 6 
> (microseconds), which is not respected by some databases (most prominently 
> MySQL, which defaults to a precision of "0" instead).
> However, even if respected, when using timestamps generated by the database 
> itself, which include the relevant precision, using those values for later 
> comparison often fails because of precision mismatch and also for different 
> behavior of different databases regarding fractional handling and the way how 
> comparisons on timestamps work.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to