[GRASS-dev] adding 'desc' to dbf sql driver

[Moritz Lennert]

Daniel Calvelo wrote:
> Hi Moritz,
> 
> Looks good to me. In the yac.y patch, you don't need to duplicate the
> code for ORDER BY NAME and ORDER BY NAME ASC.

How do I take care of the case where neither asc nor desc are given in 
the sql ? This should default to asc, but this:

+y_order_asc:
+               NAME | NAME ASC                         { 
sqpOrderColumn( $1, SORT_ASC ); }

does not work, i.e. results of 'order by' without asc are not sorted.

So how does this need to be formulated ?


> For cmp_row_desc, you
> could just return( - cmp_row_asc) with the same arguments. Glynn, a
> cursory look?

Did that and it seems to work.

> 
> FYI, the text file in lib/db/sqlp/test/test allows for some testing of
> the parser. That may help you find any hidden bugs in the sqlp part of
> your patch. 


I get

Input row: -->>INSERT INTO pok ( c1, c2,c3) VALUES ( 'abc', +32, +56.7 );
<<--
Input statement: -->>INSERT INTO pok ( c1, c2,c3) VALUES ( 'abc', +32, 
+56.7 )<<--
Error: statement was not parsed successfully.

Don't know how this would be related to my changes.

(BTW

INSERT INTO pok ( c1, c2,c3) VALUES ( 'abc', -32, -56.7 );

does not cause any problems.)

If I remove the above insert statement and add:

SELECT c1,c3 FROM pok order by c1;

it runs perfectly.

However, when I insert

SELECT c1,c3 FROM pok order by c1 ASC;
or
SELECT c1,c3 FROM pok order by c1 DESC;

I get a segmentation fault:

Input row: -->>SELECT c1,c3 FROM pok order by c1 ASC;
<<--
Input statement: -->>SELECT c1,c3 FROM pok order by c1 ASC<<--
********** SQL PARSER RESULT **********
INPUT: SELECT c1,c3 FROM pok order by c1 ASC
COMMAND: SELECT
TABLE: pok
COLUMN  1: c1
COLUMN  2: c3
./test2: line 15: 32474 Done                    echo "SELECT c1,c3 FROM 
pok where c3 <> 34 and c5 = 2.3;
SELECT c1,c3 FROM pok order by c1 ASC;
INSERT INTO pok  VALUES ( 'abc', 32, 56.7 );
INSERT INTO pok ( c1, c2,c3) VALUES ( 'abc', 32, 56.7 );
INSERT INTO pok ( c1, c2,c3) VALUES ( 'abc', -32, -56.7 );
upDAte pok SET c2 = '5', c3 = 1  WHERE c1 = '1' AND c5 = 6;
delete FROM pok WHERE c1 = '1' and c2=3 AND c5 <= 4.35;
CREATE TABLE pok ( c1 INT, c2 VARCHAR (5), c3 DOUBLE );
DROP TABLE pok;
ALTER TABLE pok ADD COLUMN id int;
ALTER TABLE pok ADD popis varchar(10);
UPDATE pok SET c2 = NULL, c1=c2, c3=(-c3+12)/c1 where c1 NOT NULL;
update pok set c1=c2,c2=c1,c3=NULL where c1+2>1/cat+0.5 and not (c1=1 or 
c2=2);"     32475 Segmentation fault      | ./sqlptest


I'm a bit lost here. What is print.c used for ?

> You might also wish to update print.c to accomodate for
> asc/desc (it's a one-line change near the end).

Did that, but not sure if correctly. Should I check for the existance of 
orderDir before using it, i.e. should it rather be

if ( sqlpStmt->command == SQLP_SELECT ) {
     if ( sqlpStmt->orderDir )
        {
         fprintf( stderr, "ORDER BY: %s %s\n", sqlpStmt->orderCol, 
sqlpStmt->orderDir );
        } else {
         fprintf( stderr, "ORDER BY: %s\n", sqlpStmt->orderCol );
        }

or is this not necessary ?


> 
> And of course, after testing, this has to be documented :)
> 

Obviously... :-)

Moritz