I've no doubt that it would be possible to go down this route, but it won't be quite that simple. Two things that spring to mind would be validation of parameter handling and checking of any client-side projections that are run against the data returned from NH. The first could be done, the second would be trickier since any client-side projections are compiled lambdas so pretty opaque. Of course, you could write separate tests just against those lambdas, passing in-memory data to them to check that they operate correctly,but it would certainly add complexity to the tests.
Most of the NH test suite runs against the DB; the issue with the Linq ones was just the way that the data was getting inserted. Now that Fabio's refactored that I think the new test times are quite acceptable. Cheers, Steve On Mon, Jun 21, 2010 at 2:28 PM, Wenig, Stefan <stefan.we...@rubicon.eu>wrote: > Hi > > > > I’m not really following the efforts here, so this might be moot, but here > it is: We’re testing our re-store LINQ provider by translating LINQ queries > to SQL and then comparing actual and expected SQL in the tests. No actual > execution against the DB, except in one or two high-level integration tests. > > > > > You could do the same by emitting the generated HQL AST to an HQL string > and just comparing that to the expected HQL. There’s not really much use in > testing the actual HQL translation from the perspective of a LINQ provider, > is there? > > > > Stefan > > > > *From:* nhibernate-development@googlegroups.com [mailto: > nhibernate-developm...@googlegroups.com] *On Behalf Of *Steve Strong > *Sent:* Monday, June 21, 2010 11:46 AM > *To:* nhibernate-development@googlegroups.com > *Subject:* Re: [nhibernate-development] Re: Linq readonly test overboost > > > > Yeah, sorry about that file - it was built with a code generator, and when > I saw how big it was it I knew it had to be sorted out. But other stuff > always got in the way so it has lived for far longer than it should have... > > On Sat, Jun 19, 2010 at 5:05 AM, Ricardo <rs...@usa.com> wrote: > > I did not test the new version of Linq tests, but the function > CreateNorthwindData with thousands lines of code seems to make visual > studio crawl and consume great amount of memory just to enter the > function. I spitted the function in many parts and the code run much > faster, 5 times or more. I think you could remove the data from C# > code and transfer to a xml file and read the data from a file. > > > On Jun 18, 1:39 am, Fabio Maulo <fabioma...@gmail.com> wrote: > > Hi all. > > I have refactorized the tests of Linq provider to fill DB before the > first > > fixture and drop it after the last fixture > > > NUnit featurehttp://www.nunit.org/index.php?p=setupFixture&r=2.5.5 > > > > > <http://www.nunit.org/index.php?p=setupFixture&r=2.5.5>With NUnit2.5+ > > testrunner the old tests crash (confirmed with Richard). > > > > To run Linq provider integration test for others dialect we have to have > 2 > > files per dialect: > > <dialect-name>LinqReadonlyCreateScript.sql > > <dialect-name>DialectLinqReadonlyDropScript.sql > > > > So far available only for MsSql2008Dialect. > > > > I'm going to update NUnit of our nhibernate\Tools\nunit to the last > version. > > > > -- > > Fabio Maulo > > >
