The new parser does type checking on comparisons, and e.Reverse is not comparable to a.Forward (the first is of type A, the second of type E)
On Apr 15, 12:21 am, Ayende Rahien <[email protected]> wrote: > Steve, > That is great job. > Looking at the test results, I am fine with breaking all of them, except > for: > "from E e, A a where e.Reverse = a.Forward and a = ?" > > Looking at this, I am not sure _why_ this should generate a _parser_ error. > From syntax perspective, it seems to be valid. > > My point of view is in two points: > a) if we match Hibernate's behavior, we are good, because that is our spec, > for most things > b) we _can_ break backward compat in small ways. We are going to ship both > parsers for at least a while, and give the user the configuration option to > select one. That means that if a user is hit by a backward compat bug here, > they can just revert to the old parser, and continue using that. That is an > acceptable solution from my point of view. > > > > On Tue, Apr 14, 2009 at 10:16 PM, Steve <[email protected]> wrote: > > > Hi Chaps, > > > Finally, another update on the HQL AST parser work. I've had a few > > other things on my plate the last couple of weeks, but managed to get > > a few hours in on the parser. This post was coming out on Monday, but > > I thought I'd better resync with the NHibernate trunk, and alas it > > broke a couple of things. They are now fixed, so finally I think > > we're nearly there. I've got a little bit of refactoring to do, and > > sort out the error handling, but it's nothing major at all, and > > doesn't stop me form getting onto the LINQ side of things at last. > > There are a few tests that are failing which I could use your opinions > > on whether the new behaviour is acceptable or not - these are listed > > below. > > > If anyone wants to test this, you should just be able to get the HQL > > AST parser from the uNhAddins repository, build it and drop the binary > > into the NHibernate.Tests\bin\Debug-2.0. Finally, add the following to > > your config: > > > <property > > > name="query.factory_class">NHibernate.Hql.Ast.ANTLR.ASTQueryTranslatorFacto > > ry, > > ANTLR-HQL</property> > > > One last thing, there's a change you´ll need to make to a minor change > > to MultiQueryImpl.AggregateQueriesInformation() - change the foreach > > from > > foreach (QueryTranslator translator in queryTranslators) > > to > > foreach (IQueryTranslator translator in queryTranslators) > > > and this line: > > SqlCommandInfo commandInfo = translator.GetQueryStringAndTypes > > (session, queryParameters); > > to > > SqlCommandInfo commandInfo = > > translator.Loader.GetQueryStringAndTypes > > (session, queryParameters); > > > (I've sent a patch to Fabio, so hopefully this'll end up in the trunk > > soon) > > > That's the lot. I'm off to look at LINQ, and if you can let me know > > how to handle the issues below, that would be cool. > > > Cheers, > > > Steve > > > ======================== > > > TestCase 'NHibernate.Test.Hql.HQLFunctions.Cast' failed: > > Expected: 1.29999995f > > But was: 1.3d > > > Z:\NHibernate\Trunk\nhibernate\src\NHibernate.Test\HQL\HQLFunctions.cs > > (598,0): at NHibernate.Test.Hql.HQLFunctions.Cast() > > > The query that is being executed here is "select cast(cast > > (a.BodyWeight as string) as Double) from Animal a". The new parser is > > returning 1.3 as a double, but the test is comparing the return value > > with 1.3F (i.e., a single), and the vagaries of floating point math > > causes failures when doing the comparisons. The old parser returned > > Singles for this test, hence it worked. So the question is, should > > the new parser be changed to return Singles, or should the test be > > updated? > > > ======================== > > > Error: line 1:-1 no viable alternative at input 'is' > > TestCase 'NHibernate.Test.Hql.HQLFunctions.Nullif' > > failed: System.InvalidCastException : Unable to cast object of type > > 'Antlr.Runtime.CommonErrorNode' to type > > 'NHibernate.Hql.Ast.ANTLR.Tree.IASTNode'. > > Z:\uNhAddins\Trunk\ANTLR-HQL\ANTLR-HQL\Output\HqlParser.cs(5698,0): > > at NHibernate.Hql.Ast.ANTLR.HqlParser.relationalExpression() > > [middle of stack deleted for brevity] > > > Z:\NHibernate\Trunk\nhibernate\src\NHibernate.Test\HQL\HQLFunctions.cs > > (428,0): at NHibernate.Test.Hql.HQLFunctions.Nullif() > > > The query here is "from Human h where nullif(h.NickName, '1e1') not is > > null", and the new parser doesn't understand the "not is null" > > construct. The behaviour here matches Hibernate, which also fails > > with a query of this form. I for one don't like the syntax here, so > > would be happy leaving it "broken". The user can obviously re-write > > the query to get the required results (i.e., by using "is not null" > > rather than the sucky "not is null"). Thoughts? BTW, ignore the > > InvalidCastException that shows up in the error above, I've got to > > sort out the error handling bit; it's not been high on the list, since > > it's "just code" - nothing that is going to cause any issues :) > > > ======================== > > > failed: NHibernate.TypeMismatchException : left and right hand sides > > of a binary logic operator were incompatibile > > [NHibernate.DomainModel.A : NHibernate.DomainModel.E] > > TearDown : System.Reflection.TargetInvocationException : Exception has > > been thrown by the target of an invocation. > > ----> NUnit.Framework.AssertionException : Test didn't clean up > > after itself > > > Z:\uNhAddins\Trunk\ANTLR-HQL\ANTLR-HQL\Tree\BinaryLogicOperatorNode.cs > > (85,0): at > > > NHibernate.Hql.Ast.ANTLR.Tree.BinaryLogicOperatorNode.MutateRowValueConstru > > ctorSyntaxesIfNecessary > > (IType lhsType, IType rhsType) > > [middle of stack deleted for brevity] > > Z:\NHibernate\Trunk\nhibernate\src\NHibernate.Test\Legacy > > \ABCProxyTest.cs(292,0): at > > NHibernate.Test.Legacy.ABCProxyTest.OneToOne() > > > The query here is "from E e, A a where e.Reverse = a.Forward and a > > = ?". I think the new parser has this correct (I won't explain the > > domain model here, it's probably easier if you just look at the > > code). To back up my claim, Hibernate shows the same behaviour for > > this test :) Take a look at the test and associated model, and let me > > know what you think should happen. > > > ======================== > > > The following 4 are all from the Legacy test fixtures, and they all > > fail in Hibernate in the same way. Since they are under legacy, I've > > not spent much time looking at them. Are they things we need to > > support in the new parser? > > > TestCase 'NHibernate.Test.Legacy.FooBarTest.CollectionsInSelect' > > failed: System.NullReferenceException : Object reference not set to an > > instance of an object. > > TearDown : System.Reflection.TargetInvocationException : Exception has > > been thrown by the target of an invocation. > > ----> NUnit.Framework.AssertionException : Test didn't clean up > > after itself. session closed: False database cleaned: True connection > > closed: True > > Z:\uNhAddins\Trunk\ANTLR-HQL\ANTLR-HQL\Tree\SelectExpressionList.cs > > (30,0): at > > NHibernate.Hql.Ast.ANTLR.Tree.SelectExpressionList.CollectSelectExpressions > > () > > [middle of stack deleted for brevity] > > Z:\NHibernate\Trunk\nhibernate\src\NHibernate.Test\Legacy > > \FooBarTest.cs(2471,0): at > > NHibernate.Test.Legacy.FooBarTest.CollectionsInSelect() > > Query was "select count(*) from Bar as bar where '1' in (from g in > > bar.Component.Glarch.ProxyArray.elements where g.Name='foo')" > > > TestCase 'NHibernate.Test.Legacy.FooBarTest.Query' > > failed: NHibernate.ADOException : could not execute query > > [SQL: select bar0_.[foo_id] as foo1_517_, bar0_.Version as > > Version517_, bar0_.foo as foo517_, bar0_.long_ as long5_517_, bar0_. > > [@@##integer_ *] as column6_517_, bar0_.float_ as float7_517_, > > bar0_.X as X517_, bar0_.date_ as date9_517_, bar0_.timestamp_ as > > timestamp10_517_, bar0_.boolean_ as boolean11_517_, bar0_.bool_ as > > bool12_517_, bar0_.null_ as null13_517_, bar0_.short_ as short14_517_, > > bar0_.char_ as char15_517_, bar0_.zero_ as zero16_517_, bar0_.int_ as > > int17_517_, bar0_.string_ as string18_517_, bar0_.byte_ as > > byte19_517_, bar0_.YesNo as YesNo517_, bar0_.[stat...@###] as > > status21_517_, bar0_.[localeayzabc1...@#$] as localea22_517_, > > bar0_.first_name as first23_517_, bar0_.surname as surname517_, > > bar0_.Dependent as Dependent517_, bar0_.count_ as count26_517_, > > bar0_.name_ as name27_517_, bar0_.g__ as g28_517_, bar0_.cmpnt_null_ > > as cmpnt29_517_, bar0_.subcount as subcount517_, bar0_.subname as > > subname517_, bar0_.fee_sub as fee32_517_, bar0_.null_cmpnt_ as > > null33_517_, bar0_.the_time as the34_517_, bar0_.Baz as Baz517_, > > bar0_.bar_String as bar36_517_, bar0_.name_name as name37_517_, > > bar0_.bar_count as bar38_517_, bar0_.Name as Name517_, bar0_.clazz as > > clazz517_, bar0_.[gen_id] as gen41_517_, 2 * bar0_.int_ as formula37_ > > from [foos] bar0_, StringDateMap stringdate2_ where bar0_. > > [$foo_subclass^]='B' and bar0_.Baz=baz1_.baz_id and > > baz1_.baz_id=stringdate2_.id_ and stringdate2_.map_key = 'now' and > > (stringdate2_.map_value is not null)] > > ----> System.Data.SqlClient.SqlException : The multi-part identifier > > "baz1_.baz_id" could not be bound. > > The multi-part identifier "baz1_.baz_id" could not be bound. > > TearDown : System.Reflection.TargetInvocationException : Exception has > > been thrown by the target of an invocation. > > ----> NUnit.Framework.AssertionException : Test didn't clean up > > after itself. session closed: False database cleaned: False connection > > closed: True > > Z:\NHibernate\Trunk\nhibernate\src\NHibernate\Loader\Loader.cs > > (1661,0): at NHibernate.Loader.Loader.DoList(ISessionImplementor > > session, QueryParameters queryParameters) > > [middle of stack deleted for brevity] > > Z:\NHibernate\Trunk\nhibernate\src\NHibernate.Test\Legacy > > \FooBarTest.cs(855,0): at NHibernate.Test.Legacy.FooBarTest.Query() > > Query was "select bar from Bar bar where bar.Baz.StringDateMap['now'] > > is not null" > > > TestCase 'NHibernate.Test.Legacy.MasterDetailTest.CollectionQuery' > > failed: System.NullReferenceException : Object reference not set to an > > instance of an object. > > TearDown : System.Reflection.TargetInvocationException : Exception has > > been thrown by the target of an invocation. > > ----> > > ... > > read more »
