In the case of (auto generated) placeholders I will pass a reference to the original value into DBI->bind_param() , so this will be fine.
It's the case of literals causing me headaches.
The raw concept in short:
- the DB-table is handled as a Perl package with it's proper namespace,
- the columns are constants in this package returning objects blessed into a class SQL::Column
- operators are overloaded for SQL::Column
- operations return objects of class SQL::Operation with nested attributes like {operator => "<", left => bless (['col'], '::Column'), right=> bless (["42"], '::Literal')
- higher syntax expressions like WHERE are just functions imported into that table package, dealing with the nested objects and returning another container object
- In the end I get an abstract syntax tree of nested objects which I can walk and translate to the target SQL dialect
So something like this
query {
package t_table;
WHERE
col < "42"
}
needs to distinguish if the literal was string "42" or 42
Of course I could think about overloading lt too, to have a more Perlish way to get the type (and crosscheck against the table's definition), but the more information I have to catch errors, the better.
I hope my approach is clearer now! :)
| [reply] [d/l] [select] |