Hello Clinton, many thanks for your answers. They made a lot of things much clearer to me - "the whitepaper" set a lot of expectations in the minds of many people (including me), so that's why I asked those questions. I am sure someone will ask me those questions at the presentation, so I want to be prepared. I found some more configuration options in the trunk, but for the presentation I am limited to 3-beta-9 where they didnť work. But it's good to know, that those issues got answered. My final question - would you encourage using annotations after all? From the limited experience with iBatis 3 I have, they could be used only to simple cases, but they make the code fragmented among annotations and XML. I can imagine that this could be very confusing in the large scale, so maybe I should recommend to attendees of the presentation to better stick with good old XML. What do you think?
Thanks, Jan Offtopic: P.S.: you seem very resignated to the Java - does the folks from Microsoft got you? :) 2010/2/18 Clinton Begin <clinton.be...@gmail.com> > Hi Jan, > > > >> expectations come from iBatis 3 Whitepaper > > Just to be clear, it wasn't a "white paper" it was a community whiteboard > for brainstorming ideas. So it contained goals and ideas, but there was > never a commitment to any of them. > > > >> Multilevel Configuration > > Ultimately I decided not to implement it this way because it became hard to > trace where behavior was configured. To fully understand how a statement > might behave or perform, you would have had to look in 3 places. I wasn't > satisfied with that, so I made it to be either XML or Annotations per > statement. You can happily mix mappers with annotations or xml statements, > but each statement must use either one or the other. > > > >> to state all properties in order to alter the single one > > That's configurable. See the docs under the automappingBehavior setting (I > think that's what it's called). > > >>annotations > > You think it's unpleasant? You should have seen me the day that I had to > remove WORKING CODE from my system because Sun changed the behavior of Java > Annotations between Java 1.5 and 1.6. > > In a nutshell, circular references are not supported in the Annotations > framework. All of your best hopes and dreams of having something better are > dashed by this one limitation. There was a workaround in Java 1.5, but Sun > considered it a compiler bug, so they "fixed" the workaround in JDK 1.6. If > you want to see it, you can dig back through a few hundred commits and see > where I removed that behavior. > > A shame. If this was C# I would have been able to fully realize the vision > of where we wanted to take the annotation (or attributes in C#) approach. > > **>> 3 combinations of annotations and xml > > In general, I recommend using annotations for simple things, and use XML > for anything more complicated. Sorry, but Java annotations are just too > messy, limited and ugly. There's a million reasons why we had to scale back > the behavior and limit it to what you see now. If you want to understand > why, I encourage you to try to implement some of your suggestions. You'll > find the same challenges. > > >> 4 automatic (generated) SQL > > Yes, I completely ditched the idea of generating SQL. It's just a road > that I wasn't prepared to take iBATIS down. > > >> 5 dynamic sql in annotations > > Yes, SelectProvider and the SQL Builder framework are the only way (other > than writing your own). I wasn't about to start putting XML embedded in SQL > embedded in strings embedded in annotations... that's out of hand and bad > practice IMHO. > > I appreciate your questions, but please realize that everything was > considered. Some things were left out for very good design reasons, others > because of limitations in Java and others we might just not have had a > chance to implement yet. > > Clinton > > 2010/2/18 Ing. Jan Novotný <novotn...@gmail.com> > >> Hello, >> >> >> I am going to have a presentation about iBatis 3 at Czech JUG. I have a >> 3-year experience with iBatis-2 and now I am doing little research on what's >> new in upcoming iBatis 3 release. Most of my expectations come from iBatis 3 >> Whitepaper at the iBatis Confluence site. Now I am testing the new version >> and there are some things I didn't expect. I will be very pleased if anyone >> capable could answer me following questions so that I may forward them to >> the attendees of CZ JUG: >> >> * 1) multilevel configuration* >> Firstly I thought that there would be a chance to override less >> prioritized level of configuration. Meaning, that if I have SQL represented >> in an @Select annotation and I create XML statement in mapper file with the >> same name, this XML declaration will take over the annotation and will be >> used instead. Reality is that iBatis will fail to start complaining about >> duplicate declaration. As I discovered, I can use only one or the other way >> - I can combine XML file with annotation based Java DAO, but they must not >> overlap in the sense of statement declaration. Is that right? Why is that >> done this way? >> Next I thought there could be some means to explicitly say, that >> some parts of the lower configuration level fit for me and therefore ease my >> life a little. Let say I have ResultMap statement in XML configuration and >> 95% of its properties is fine by convention, but I need only to specify more >> single property or declare association / collection mapping. In that case I >> have to state all properties in order to alter the single one that is needed >> to be extra taken care of (or at least these are my findings from the tests >> I wrote). What I really want to say is: use convention to all properties >> except those that I explicitly configure in XML - that is not possible, is >> it? >> >> * 2) annotations* >> Another unpleasant surprise was, that I am not able to get along >> with annotations to specify <collection> mapping - annotation @Collection is >> missing. Also @Association is not present - but as I discovered, it's >> possible to write @Result annotations in the way that associated object gets >> mapped. In documentation I found a statement, that annotations in Java >> aren't powerfull enough to make some things possible, but I can't make up >> why @Collection annotation is not possible to be made. >> >> * 3) combination of XML and annotations* >> I tried combination of XML and annotations (due to lack of >> @Collection annotation) to specify @Select annotation but specify output as >> a ResultMap mapping in the XML configuration. But that's not possible too. >> Mixing configuration from annotations and XML is not possible in any way I >> guess. Ok, but is there any way how to reuse my annotations, to get the same >> behaviour as reusing ResultMap mapping in different statements of XML >> configuration?! >> >> * 4) automatic SQL implementation* >> I didn't find any piece of this functionality. May I take it, that >> this was completely discarded? >> >> * 5) annotations and dynamic SQL* >> Finally I didn't find any clue how to use dynamic SQL in >> annotations. Does that mean that @SelectProviders are the only way how to >> make dynamic queries by the Java code? >> >> Thanks for your time and answers I can use in my presentation if anyone >> asks. Please, don't be offended by my questions, generally I think iBatis is >> a great DAO framework. >> >> Cheers Jan >> >> -- >> -------------------------------------------------------------- >> Ing. Jan Novotný >> @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ >> http://blog.novoj.net >> Myšlenky dne otce Fura >> -------------------------------------------------------------- >> > > -- -------------------------------------------------------------- Ing. Jan Novotný @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ http://blog.novoj.net Myšlenky dne otce Fura --------------------------------------------------------------
