[Mondrian] Re: Regression tests MonetDB

Stefan de Konink stefan at konink.de
Tue Feb 17 14:08:47 EST 2009


Julian Hyde wrote:
>> We hope you could take a peak at the output. One feature we currently do
>> not support yet is SELECT * FROM VALUES(...) as t;
> 
> Don't worry, neither does Oracle.

It is on the feature request page now.

> Do you have a way to evaluate an 
> expression so that it always returns a single row? Oracle has the Dual 
> table, MySQL has 'select expression'. Etc.

Are you talking about limit 1?

>> But that doesn't
>> explain the Tests run: 1992,  Failures: 297,  Errors: 146.
>>
>> Next to this people here consider:
>>
>> as "Name (Key)"
>>
>>
>> Invalid identifiers (that should only contain alphanumeric characters);
>> but I found someone willing to hack it in.
>>
> 
> Glad you found a hack, but every database I know allows you to have 
> spaces, mixed case, and punctuation in identifers, so you should add 
> that feature. Standard SQL uses double-quote to quote such identifiers.

If you can give the page number of ISO 9075 standard that defines 
especially *this*, I am sure it will come in more soon ;) But there is a 
fundamental problem (in kernel) with this currently that should be fixed 
first.

>> For the rest, some tips will be needed how to fix it. If you want I can
>> help you to setup MonetDB yourself.
>>
> 
> You/we will need to create a MonetDB dialect, and get it to pass 
> DialectTest. The rest will follow.

I think it is not about dialects anymore; but really about why do we not 
pass a test and why is the regression test producing such strange 
queries. I really have examples of them UPPER( a_number ) for example, 
that don't match in any possible way.

> Give me the patched binaries and I will try to run the tests myself.

I suggest you checkout the CVS HEAD tomorrow. Tonight a significant 
patch will not yet be propagated.

> Ps haven't looked at yr logs yet. Only so much you can do on an iPhone 
> in a paediatrician's office. :)

:D!

> Pls pls cc mondrian at Pentaho on msgs - I want to keep this in a public 
> forum.

I am subscribed now too :)



Stefan





More information about the Mondrian mailing list