Case sensitivity in ODBC ?? - Mailing list pgsql-interfaces

Dear pgsql-interface list,

I have lived that SQL database were not to be case-sensitive. However, 
it seems that, at least through the ODBC interface, PostgreSQL 6.5.3 
is :

Config : Linux RedHat 6.1, PostgreSQL 6.5.3 from the PostgreSQL ftp 
site's RPM, unixODBC1.7 (including their PostgreSQL driver). Works 
great from various interfaces (including StarOffice ...).

I'm working from the R statistical language/package through a beta 
ODBC interface. Note that the interpreted R language is case-sensitive ...

Logical Setup

R user interface ---> RODBC library ---> unixODBC ---> PostgreSQL                            ^
 |
 
I'm trying to enhance that---+

This interface mostly works, but has some odd behaviour :

> sqlQuery(chan1,"create table Test1 (id int4 not null primary key, val 
varchar(255))") /* This creates the table test1, correctly */

> sqlColumns(chan1,"test1") /* Sanity check */ TABLE_QUALIFIER TABLE_OWNER TABLE_NAME COLUMN_NAME DATA_TYPE 
TYPE_NAME
1              NA          NA      test1          id         4      
int4
2              NA          NA      test1         val        12   
varchar PRECISION LENGTH SCALE RADIX NULLABLE REMARKS DISPLAY_SIZE 
FIELD_TYPE
1        10      4     0    10        0      NA           11         
23
2       254    254    NA    NA        1      NA          254       
1043
/* This is the expected answer */

> sqlTables(chan1) TABLE_QUALIFIER TABLE_OWNER  TABLE_NAME TABLE_TYPE REMARKS
1              NA          NA   pga_forms      TABLE      NA
2              NA          NA  pga_layout      TABLE      NA
3              NA          NA pga_queries      TABLE      NA
4              NA          NA pga_reports      TABLE      NA
5              NA          NA  pga_schema      TABLE      NA
6              NA          NA pga_scripts      TABLE      NA
7              NA          NA       test1      TABLE      NA
/* This also is the expected answer. Note however that the name of the 
table is lowercased */


> sqlColumns(chan1,"Test1") /* Same sanity check, with the original name 
*/
Error in sqlColumns(chan1, "Test1") : Test1 :table not found on 
channel 0
/* This is unexpected if SQL, PostgreSQL and ODBC are case-insensitive 
*/

Furthermore : debugging shows that the initial request for table 
creation seds the name with its capital, as shown in the next example 
:

> sqlSave(chan1,USArrests,rownames="State") /* The sqlSave function 
creates the necessary table if it does not exists */
[1] "CREATE TABLE USArrests  (State varchar(255)  ,Murder varchar(255) 
,Assault varchar(255)  ,UrbanPop varchar(255)  ,Rape varchar(255)  )"
/* This is a debugging output showing the exact request sent to 
PosstgreSQL, minus the ending semicolumn. Yes the types are 
ridiculous, and that's what I'm trying to fix ... */
Error in sqlColumns(channel, tablename) : USArrests :table not found 
on channel 0
/* This is an (unexpected) error message*/
> sqlTables(chan1) TABLE_QUALIFIER TABLE_OWNER  TABLE_NAME TABLE_TYPE REMARKS
1              NA          NA   pga_forms      TABLE      NA
2              NA          NA  pga_layout      TABLE      NA
3              NA          NA pga_queries      TABLE      NA
4              NA          NA pga_reports      TABLE      NA
5              NA          NA  pga_schema      TABLE      NA
6              NA          NA pga_scripts      TABLE      NA
7              NA          NA   usarrests      TABLE      NA
/* A table « usarrests » has been created, but querying « USArrests » 
does not work. However :*/

> sqlColumns(chan1,"usarrests") TABLE_QUALIFIER TABLE_OWNER TABLE_NAME COLUMN_NAME DATA_TYPE 
TYPE_NAME
1              NA          NA  usarrests       state        12   
varchar
2              NA          NA  usarrests      murder        12   
varchar
3              NA          NA  usarrests     assault        12   
varchar
4              NA          NA  usarrests    urbanpop        12   
varchar
5              NA          NA  usarrests        rape        12   
varchar PRECISION LENGTH SCALE RADIX NULLABLE REMARKS DISPLAY_SIZE 
FIELD_TYPE
1       254    254    NA    NA        1      NA          254       
1043
2       254    254    NA    NA        1      NA          254       
1043
3       254    254    NA    NA        1      NA          254       
1043
4       254    254    NA    NA        1      NA          254       
1043
5       254    254    NA    NA        1      NA          254       
1043
>  
/* Querying « usarrests » does ! Note that column names have been 
lowercased as well ... */

This seems to me proof enough that the behaviour of PostgreSQL (or 
ODBC driver) related to case sensitivity is not coherent.

Could some kind soul shed some light on this ?

Thanks in advance,
                    Emmanuel Charpentier