Case-sensitivity problems with columns (DbAttributes)

[John Huss <jo...@gmail.com>]

I've run into problems a few times caused by case-mismatch issues with
columns (DbAttributes).

The primary one being when I have SQLTemplate queries that return actual
DataObjects.  If the DbAttribute is declared with mixed case (myTablePK)
then my query just returns a list of nulls because it can't find the pk
field to create an ObjectId.  I'm using postgres mostly which is not
case-sensitive and returns the columns in lowercase it seems.

What do you think about changing some of this behavior?

Like forcing or providing a switch to ensure DbAttribute names are
lowercase?

Or doing more intelligent matching that ignores case as much as possible?
I.E. if there are no duplicate columns when ignoring case, then ignore case?

John

Re: Case-sensitivity problems with columns (DbAttributes)

[John Huss <jo...@gmail.com>]

On Mon, Jun 15, 2015 at 10:16 AM Andrus Adamchik <an...@objectstyle.org>
wrote:

> Hi John,
>
> IIRC you can force PostgreSQL (or any other DB really) to treat column
> names as case-sensitive by including them in double quotes when creating
> and querying the tables. In Cayenne this means setting "Quote SQL
> Identifiers" to "on" in the DataMap and using explicit double quotes in
> SQLTemplate.
>
> Not going to help with "SELECT * FROM", but should address all other cases.
>

I definitely don't want my DB to be case sensitive, or to make my all my
SQL statements much longer by adding quotes everywhere.


>
> > Like forcing or providing a switch to ensure DbAttribute names are
> > lowercase?
>
> If the above doesn't help, you can probably decorate DataMapLoader service
> to conditionally convert loaded DataMap to lower case DbAttributes (if you
> can access DataSource and determine DB type prior to project loading), or
> do that in the EntityResolver after ServerRuntime startup (less clean, but
> same end result).
>

I've addressed it in my project by doing this. If that's going to be the
recommended solution it should be part of the library.  The code to switch
everything to lowercase is a bit involved. I wonder if it would be
reasonable to detect the DB and do this automatically?

The specific problem I ran into is this: in ObjectResolver.createObjectId
there is this:

            Object val = dataRow.get(key);
            // this is possible when processing left outer joint prefetches
            if (val == null) {
                return null;
            }

Since the case doesn't match for the PK  null is returned.  Making this
fallback to a case-insensitive lookup would help a bit.  However, all the
columns have this problem, not just the PKs.  The difference the PK columns
are more likely to be defined in mixed case because they are generated into
java source as the PK_COLUMN constant and mixed case is necessary for a
good name to be generated.


>
> Andrus
>
>
> > On Jun 13, 2015, at 12:13 AM, John Huss <jo...@gmail.com> wrote:
> >
> > I've run into problems a few times caused by case-mismatch issues with
> > columns (DbAttributes).
> >
> > The primary one being when I have SQLTemplate queries that return actual
> > DataObjects.  If the DbAttribute is declared with mixed case (myTablePK)
> > then my query just returns a list of nulls because it can't find the pk
> > field to create an ObjectId.  I'm using postgres mostly which is not
> > case-sensitive and returns the columns in lowercase it seems.
> >
> > What do you think about changing some of this behavior?
> >
> > Like forcing or providing a switch to ensure DbAttribute names are
> > lowercase?
> >
> > Or doing more intelligent matching that ignores case as much as possible?
> > I.E. if there are no duplicate columns when ignoring case, then ignore
> case?
> >
> > John
>
>

Re: Case-sensitivity problems with columns (DbAttributes)

[Andrus Adamchik <an...@objectstyle.org>]

Hi John,

IIRC you can force PostgreSQL (or any other DB really) to treat column names as case-sensitive by including them in double quotes when creating and querying the tables. In Cayenne this means setting "Quote SQL Identifiers" to "on" in the DataMap and using explicit double quotes in SQLTemplate.

Not going to help with "SELECT * FROM", but should address all other cases.

> Like forcing or providing a switch to ensure DbAttribute names are
> lowercase?

If the above doesn't help, you can probably decorate DataMapLoader service to conditionally convert loaded DataMap to lower case DbAttributes (if you can access DataSource and determine DB type prior to project loading), or do that in the EntityResolver after ServerRuntime startup (less clean, but same end result).

Andrus


> On Jun 13, 2015, at 12:13 AM, John Huss <jo...@gmail.com> wrote:
> 
> I've run into problems a few times caused by case-mismatch issues with
> columns (DbAttributes).
> 
> The primary one being when I have SQLTemplate queries that return actual
> DataObjects.  If the DbAttribute is declared with mixed case (myTablePK)
> then my query just returns a list of nulls because it can't find the pk
> field to create an ObjectId.  I'm using postgres mostly which is not
> case-sensitive and returns the columns in lowercase it seems.
> 
> What do you think about changing some of this behavior?
> 
> Like forcing or providing a switch to ensure DbAttribute names are
> lowercase?
> 
> Or doing more intelligent matching that ignores case as much as possible?
> I.E. if there are no duplicate columns when ignoring case, then ignore case?
> 
> John