openjpa-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Patrick Linskey" <plins...@bea.com>
Subject RE: [jira] Commented: (OPENJPA-221) DerbyDictionary doesn't describe a working mapping for char fields.
Date Thu, 19 Apr 2007 07:30:12 GMT
> I don't know how feasible it is, but maybe if openjpa can 
> detect that the column used to store a char is defined as SQL 
> CHAR(1) then the mapping to the database character is used. 
> And we can then depend on the database to signal that it's 
> unhappy with storing a char that doesn't fit.

Historically, our experience has been that:

1. Users tend to use a single format throughout their database, so a
per-configuration setting is Good Enough.

2. It's generally pretty uncommon to store char fields anyways.

-Patrick

-- 
Patrick Linskey
BEA Systems, Inc.
_______________________________________________________________________
Notice:  This email message, together with any attachments, may contain
information  of  BEA Systems,  Inc.,  its subsidiaries  and  affiliated
entities,  that may be confidential,  proprietary,  copyrighted  and/or
legally privileged, and is intended solely for the use of the individual
or entity named in this message. If you are not the intended recipient,
and have received this message in error, please immediately return this
by email and then delete it. 

> -----Original Message-----
> From: Craig.Russell@Sun.COM [mailto:Craig.Russell@Sun.COM] 
> Sent: Wednesday, April 18, 2007 6:05 PM
> To: open-jpa-dev@incubator.apache.org
> Subject: Re: [jira] Commented: (OPENJPA-221) DerbyDictionary 
> doesn't describe a working mapping for char fields.
> 
> First, let me observe that Java char is a single entity that 
> represents one of 65535 or so different items, more if you 
> are trying to represent characters in Chinese that require 
> Unicode-4 byte representation.
> 
> Most databases treat CHAR(1) as a single byte, which is great 
> for ASCII 8 but woefully inadequate for representing Java char.
> 
> 1. For the best accuracy in all databases, mapping char to
> String.valueOf((int)char) is bound to be correct.
> 
> 2. But if the user wants to lose precision and save a few 
> bytes in the database, mapping directly from char to CHAR(1) is ideal.
> 
> It doesn't seem to me that there is a single "best" solution 
> to this issue. Historically, Kodo mapped char to INTEGER or 
> VARCHAR column types and there is nothing intrinsically wrong 
> with this.
> 
> On the other hand, if the column type is CHAR(1) there is no 
> way to represent any of the ASCII characters 0-9;a-z;A-Z so 
> there is something to be said for assuming a different 
> mapping in this case.
> 
> I don't know how feasible it is, but maybe if openjpa can 
> detect that the column used to store a char is defined as SQL 
> CHAR(1) then the mapping to the database character is used. 
> And we can then depend on the database to signal that it's 
> unhappy with storing a char that doesn't fit.
> 
> Craig
> 
> On Apr 18, 2007, at 3:37 PM, Patrick Linskey wrote:
> 
> >> Does that mean you store Strings as arrays of integers by 
> default for 
> >> the same reason?
> >
> > No.
> >
> >> Why is this different?
> >
> > Because we've regularly run into problems with chars, and 
> have found 
> > that mapping them as ints by default gets around the problems.
> >
> >> There aren't any non-incubator openjpa releases yet, so I 
> don't see 
> >> the problem.
> >
> > Any other opinions?
> >
> >> I'd also expect that for preexisting
> >> tables either an INTEGER or CHAR column would work.
> >
> > They do, as long as you configure things correctly.
> >
> >>> openjpa.Log: SQL=TRACE
> >>
> >> Where would I put this so I could see what the unit tests 
> were doing?
> >
> > What are "the unit tests"? Your tests, or the tests in the OpenJPA 
> > project?
> >
> > For your tests, add a <property name="openjpa.Log" 
> value="SQL=TRACE"/> 
> > to your persistence unit. See 
> > http://incubator.apache.org/openjpa/docs/latest/manual/
> > manual.html#ref_g
> > uide_logging.
> >
> >> (1) is a lot more important, but changing the answer to 
> (2) is easier 
> >> and solves my immediate problem.
> >
> > I don't think that 1 is important, since you can trivially set 
> > storeCharsAsNumbers to true. Ditto for 2.
> >
> > -Patrick
> >
> > --
> > Patrick Linskey
> > BEA Systems, Inc.
> > 
> ______________________________________________________________________
> > _
> > Notice:  This email message, together with any attachments, may 
> > contain
> > information  of  BEA Systems,  Inc.,  its subsidiaries  and   
> > affiliated
> > entities,  that may be confidential,  proprietary,  copyrighted   
> > and/or
> > legally privileged, and is intended solely for the use of the 
> > individual or entity named in this message. If you are not the 
> > intended recipient, and have received this message in error, please 
> > immediately return this by email and then delete it.
> >
> >> -----Original Message-----
> >> From: David Jencks [mailto:david_jencks@yahoo.com]
> >> Sent: Wednesday, April 18, 2007 3:20 PM
> >> To: open-jpa-dev@incubator.apache.org
> >> Subject: Re: [jira] Commented: (OPENJPA-221) 
> DerbyDictionary doesn't 
> >> describe a working mapping for char fields.
> >>
> >>
> >> On Apr 18, 2007, at 11:12 AM, Patrick Linskey wrote:
> >>
> >>> Hi,
> >>>
> >>>
> >>>> IIUC derby is a pure java db optimized for use with java
> >> and storing
> >>>> java primitive types basically using java serialization.
> >> Why would
> >>>> openjpa want to store a char in derby as an integer?
> >>>
> >>> "Because we've always done it that way." Is there a reason why we 
> >>> should not be storing chars as numbers? Historically, we've seen 
> >>> problems with comparisons and localization issues when
> >> storing chars
> >>> as chars, which is why we store them as ints by default.
> >>
> >> Does that mean you store Strings as arrays of integers by 
> default for 
> >> the same reason?  Why is this different?
> >>
> >>> Based on the fact that you said
> >>> that the unit tests fail with Derby with that configuration
> >> change, it
> >>> sounds like there are some sorts of issues with char mappings in 
> >>> Derby.
> >>
> >> The unit tests fail with only the 
> storeCharsAsIntegers=false because 
> >> the sql for creating the table is invalid.  They succeed with the 
> >> additional patch to create a CHAR(1) column instead of 
> CHAR(255) for 
> >> a char field.  I'm happy to discuss if creating a  
> CHAR(255) column 
> >> to store a char field is reasonable :-)
> >>>
> >>> Additionally, since we've always done it that way, changing
> >> would mean
> >>> backwards-compatibility problems.
> >>
> >> There aren't any non-incubator openjpa releases yet, so I 
> don't see 
> >> the problem.  I'd also expect that for preexisting tables 
> either an 
> >> INTEGER or CHAR column would work.
> >>
> >>>
> >>>> Why are the current settings correct, despite not 
> working with the 
> >>>> obvious char <> CHAR mapping?
> >>>
> >>> How do you define "not working"? It's my expectation that if the 
> >>> application behaves as expected, then things are working. 
> It sounds 
> >>> like what you're saying is "the default is not what was
> >> expected", not
> >>> that "things don't work".
> >>
> >> I expect that if I have a char field in an object and a 
> preexisting 
> >> table with a CHAR column openjpa will figure out some way to get a 
> >> char from the field to the column and back again without any 
> >> additional configuration, for all databases.  Admittedly 
> my proposed 
> >> fix only does this for derby, and by changing the default 
> mapping for 
> >> chars for derby.
> >>
> >> I additionally expect that if openjpa creates a schema for 
> me for a 
> >> database with default utf support it will map a char field 
> to a CHAR 
> >> column.  I wouldn't necessarily expect this for a database that by 
> >> default doesn't have utf columns.
> >>
> >>>
> >>>> I haven't found the magic setting so I can see what 
> table is being 
> >>>> created for the unit tests
> >>>
> >>> openjpa.Log: SQL=TRACE
> >>
> >> Where would I put this so I could see what the unit tests 
> were doing?
> >>
> >> I think there are 2 issues here:
> >>
> >> 1. should openjpa be able to use a preexisting CHAR column for 
> >> storing a char, no matter what the storeCharsAsInteger setting is?
> >> 2. should the settings for derby be storeCharsAsInteger = false or 
> >> true?
> >>
> >> (1) is a lot more important, but changing the answer to 
> (2) is easier 
> >> and solves my immediate problem.
> >>
> >> thanks
> >> david jencks
> >>
> >>>
> >>> -Patrick
> >>>
> >>> --
> >>> Patrick Linskey
> >>> BEA Systems, Inc.
> >>>
> >> 
> _____________________________________________________________________
> >> _
> >>> _
> >>> Notice:  This email message, together with any attachments, may 
> >>> contain information  of  BEA Systems,  Inc.,  its 
> subsidiaries  and 
> >>> affiliated entities,  that may be confidential,  proprietary,  
> >>> copyrighted and/or legally privileged, and is intended solely for 
> >>> the use of the individual or entity named in this message. If you 
> >>> are not the intended recipient, and have received this message in 
> >>> error, please immediately return this by email and then delete it.
> >>>
> >>>> -----Original Message-----
> >>>> From: David Jencks [mailto:david_jencks@yahoo.com]
> >>>> Sent: Wednesday, April 18, 2007 10:53 AM
> >>>> To: open-jpa-dev@incubator.apache.org
> >>>> Subject: Re: [jira] Commented: (OPENJPA-221)
> >> DerbyDictionary doesn't
> >>>> describe a working mapping for char fields.
> >>>>
> >>>> I'm not understanding something, maybe someone could 
> explain, and 
> >>>> obviously the comments I suggested in DBDictionary are 
> completely 
> >>>> wrong, although I sure don't see why.
> >>>>
> >>>> IIUC derby is a pure java db optimized for use with java
> >> and storing
> >>>> java primitive types basically using java serialization.
> >> Why would
> >>>> openjpa want to store a char in derby as an integer?  
> Why are the 
> >>>> current settings correct, despite not working with the
> >> obvious char
> >>>> <> CHAR mapping?  I haven't found the magic setting so I
> >> can see what
> >>>> table is being created for the unit tests, but I'm 
> pretty sure it 
> >>>> isn't creating a CHAR column for the char field in the allTypes 
> >>>> object.
> >>>>
> >>>> I assumed the problems I ran into were a result of no one having 
> >>>> tested this code path, but you appear to be saying that
> >> the current
> >>>> code is more correct than my proposal.  I'd really like to
> >> know why.
> >>>>
> >>>> On Apr 18, 2007, at 10:18 AM, Patrick Linskey (JIRA) wrote:
> >>>>
> >>>>>
> >>>>>     [ https://issues.apache.org/jira/browse/OPENJPA-221?
> >>>>> page=com.atlassian.jira.plugin.system.issuetabpanels:comment-
> >>>>> tabpanel#action_12489820 ]
> >>>>>
> >>>>> Patrick Linskey commented on OPENJPA-221:
> >>>>> -----------------------------------------
> >>>>>
> >>>>> It's not surprising that the OpenJPA tests
> >>>> storeCharsAsNumbers to be
> >>>>> true.
> >>>>
> >>>> maybe to you :-)  I still find it extremely surprising, 
> and can't 
> >>>> imagine any reason why you'd want to do this.
> >>>>>
> >>>>> I was referring to your test environment. Rather than
> >> changing the
> >>>>> default behavior of the DerbyDictionary in code, it seems more 
> >>>>> appropriate to use the built-in configuration option to
> >>>> toggle it for
> >>>>> your application.
> >>>>>
> >>>>> It sounds like you're reluctant to do this since you don't
> >>>> have easy
> >>>>> access to modify the persistence.xml files. Happily, if
> >> you drop a
> >>>>> file conforming to the persistence.xml schema into META-INF/ 
> >>>>> openjpa.xml, OpenJPA will load the settings in the
> >>>> properties in the
> >>>>> first PU in that file as defaults for all PUs.
> >>>>>
> >>>>> What happens if you put the DBDictionary stanza that I 
> mentioned 
> >>>>> earlier into a META-INF/openjpa.xml file?
> >>>>
> >>>> Won't this change the behavior for all databases, not just
> >> the derby
> >>>> dictionary?  I'd prefer to
> >>>>
> >>>> (a) understand why these settings as are they are
> >>>> (b) make all the db-specific dictionaries work 
> unmodified with all 
> >>>> reasonable mappings.
> >>>>
> >>>> thanks
> >>>> david jencks
> >>>>
> >>>>
> >>>>>
> >>>>>> DerbyDictionary doesn't describe a working mapping for
> >> char fields.
> >>>>>>
> >> -------------------------------------------------------------------
> >>>>>>
> >>>>>>                 Key: OPENJPA-221
> >>>>>>                 URL: https://issues.apache.org/jira/browse/
> >>>>>> OPENJPA-221
> >>>>>>             Project: OpenJPA
> >>>>>>          Issue Type: Bug
> >>>>>>          Components: sql
> >>>>>>    Affects Versions: 0.9.7
> >>>>>>            Reporter: David Jencks
> >>>>>>         Attachments: OPENJPA-221.patch
> >>>>>>
> >>>>>>
> >>>>>> If a class has a char field mapped to CHAR or CHAR(1) 
> in a derby 
> >>>>>> database, the derby dictionary sets up a mapping to an
> >>>> integer column
> >>>>>> which doesn't work.  openjpa tries to store e.g. the
> >>>> string "97" for
> >>>>>> the char 'a' which results in a truncation error.
> >>>>>
> >>>>> --
> >>>>> This message is automatically generated by JIRA.
> >>>>> -
> >>>>> You can reply to this email to add a comment to the 
> issue online.
> >>>>>
> >>>>
> >>>>
> >>>
> >>> Notice:  This email message, together with any attachments, may 
> >>> contain information  of  BEA Systems,  Inc.,  its 
> subsidiaries and  
> >>> affiliated entities,  that may be confidential,
> >> proprietary,
> >>> copyrighted  and/or legally privileged, and is intended
> >> solely for the
> >>> use of the individual or entity named in this message. If
> >> you are not
> >>> the intended recipient, and have received this message in error, 
> >>> please immediately return this by email and then delete it.
> >>
> >>
> >
> > Notice:  This email message, together with any attachments, may  
> > contain information  of  BEA Systems,  Inc.,  its subsidiaries   
> > and  affiliated entities,  that may be confidential,  
> proprietary,   
> > copyrighted  and/or legally privileged, and is intended 
> solely for the 
> > use of the individual or entity named in this message. If 
> you are not 
> > the intended recipient, and have received this message in error, 
> > please immediately return this by email and then delete it.
> 
> Craig Russell
> Architect, Sun Java Enterprise System http://java.sun.com/products/jdo
> 408 276-5638 mailto:Craig.Russell@sun.com P.S. A good JDO? O, Gasp!
> 
> 

Notice:  This email message, together with any attachments, may contain information  of  BEA
Systems,  Inc.,  its subsidiaries  and  affiliated entities,  that may be confidential,  proprietary,
 copyrighted  and/or legally privileged, and is intended solely for the use of the individual
or entity named in this message. If you are not the intended recipient, and have received
this message in error, please immediately return this by email and then delete it.

Mime
View raw message