A SQL NULL means a column has no value. It is not the same as the zero value of the column’s Go destination. If an employee’s department is NULL, scanning it into a string fails rather than producing an empty string. section 10.4.4
A nullable department
A query reads an employee’s department and age. With `department` declared as `string`, `rows.Scan(&department, &age)` can fail with `converting NULL to string is unsupported`. Declaring `department` as `*string` and scanning into `&department` instead makes it nil when the column is NULL. section 10.4.4
var (
department sql.NullString #1
age int
)
for rows.Next() {
err := rows.Scan(&department, &age)
// ...
}
Alternatively, scan the department into `sql.NullString`. Check `Valid` before using `String`: `Valid` reports whether the column was not NULL. section 10.4.4
Choose a pointer when a nil destination expresses absence clearly; choose an `sql.NullXXX` type when an explicit validity field makes that intent clearer. Neither approach turns NULL into an ordinary string value. section 10.4.4
The same distinction matters outside SQL. In a Go configuration struct, an omitted integer field becomes 0, so a plain `int` cannot distinguish a missing port from one deliberately set to 0; a `*int` can. Do not assume nil always means a missing database value, though: a nil channel also has its own blocking behavior. section 2.11.1 section 9.6
