Re: [Help-glpk] Database connectivity in MathProg

[Andrew Makhorin]

> You proposed to execute DROP, CREATE, INSERT in the case of table OUT. After
> DROP the table no longer exists in the data base. Hence the structure must
> be given by MathProg.

> Otherwise the table should not be DROPped. Instead only existing records
> should be deleted.

I think the latter case should be implemented, i.e. writing to the output
table should include DELETE and INSERT

> In a SQL table with key columns for any key combination of keys there can
> be only one record in the table.

AFAIK, this is not necessary.

> Hence it is necessary to fill the keys.

> In Your syntax
> table t { ... } OUT "..." : expr ~ A, expr ~ B, ... ;
> { ... } defines the keys for which the expr shall be written. It is
> unclear to me where You specify to which columns the keys are written.

> This is what is defined in the AMPL syntax as
> { ... } -> [k ~ K, ...]

Since tables are not deleted, information about key columns will be
kept. Let, for example, there is a table created as:

CREATE TABLE FOO
(  I INT,
   J CHAR(8),
   P INT,
   Q CHAR(20),
   PRIMARY KEY(I, J)
);

Then the output table in MathProg might look like:

table mytab{i in S1, j in S2: i != j} OUT "..." "FOO":
   i+1 ~ I, j ~ J, a[i,j+1]+123 ~ P, b[i-1,j] ~ Q;

which assumes deleting all existing records from table FOO and then
inserting new records:

INSERT INTO FOO(I, J, P, Q) VALUES(...)