db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Amit Handa <Amit.Ha...@Sun.COM>
Subject Re: (DERBY-13) Quoted names with embedded period mishandled in from
Date Thu, 06 Jan 2005 09:42:09 GMT
Daniel John Debrunner wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> I think we need to reset on Derby-13, and also look at Derby-18 at the
> same time. I was confused over the exposed correlation name, thinking it
> had a schema, but it doesn't.
> 
> A table in a list either has an exposed table name or an exposed
> correlation name. The correlation name is defined by the AS clause and
> can not have a schema.
> 
> Thus with
> 
> set current schema A
> select ... FROM T1, B.T2, T3 AS TC
> 
> there should be three exposed names
> 
> A.T1 as an exposed <table name>
> B.T2 as an exposed <table name>
> TC as an exposed <correlation name>
> 
> The SQL standard rules are then (for table references in the same scope)
> 
> 1) No duplicates can exist among the exposed <table name>s
> 2) No duplicates can exist among the exposed <correlation name>s
> 3) Any <correlation name> must not be the same as the unqualifed part of
> any exposed <table name>
> 
> 
> One option in the code might be to have
> 
> // Returns the exposed table name,
> // returns null if the table has a correlation name
> TableName getExposedTableName()
> 
> // Returns the exposed correlation name,
> // returns null if the table does not have a correlation name
> String getExposedCorrelationName()
> 
> Thus to keep the two concepts separate (since they are), rather than
> trying to shoehorn them into a single value (as the current code does).
> Then the duplicate checking would have to be smarter.
> 

I get what you are saying. Note that we already have the above two methods
in a subclass of FromTable(FromVTI.java, getExposedName() and getExposedTableName() )
I am wondering was it intended to be that way i.e. to introduce the (above) methods
in a subclass and not in base class. Is it by design ?
And if we go by what is suggested we may be breaking something else.
The subclasses will now over ride from FromTable.java insted fo FromVTI.java.

But we can always give a shot at the proposed change.
Just my thoughts after looking at the issue.

> I think this would also enable the dubious null schema handling in
> TableName to be removed, as TableNames should always have a schema.
> 
> Dan.
> 
> 
> 
> 
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.2.5 (MingW32)
> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
> 
> iD8DBQFB3E7SIv0S4qsbfuQRAkxJAKCZUIFXnm/+zsqO1MUnqqfMEyOPNQCg5gTa
> C/ri7UZzL80BtsaKM2gi59Y=
> =9aTO
> -----END PGP SIGNATURE-----
> 






Mime
View raw message