chore!: Change H2 Oracle longType and longAutoincType from NUMBER(19) to BIGINT and add CHECK constraint in Oracle

[joc-a]

Description

Detailed description:

• What:
  The SQL type of H2 Oracle Long type was changed. A CHECK constraint was also added for Oracle.
• Why:

1. H2 enforces the range on its own when BIGINT is used.
2. It is a preparation step for making detecting column type change for migration easier

---

Type of Change

Please mark the relevant options with an "X":

• Bug fix
• New feature
• Documentation update
• Improvement

Updates/remove existing public API methods:

• Is breaking change

Affected databases:

• MariaDB
• Mysql5
• Mysql8
• Oracle
• Postgres
• SqlServer
• H2
• SQLite

Checklist

• Unit tests are in place
• The build is green (including the Detekt check)
• All public methods affected by my PR has up to date API docs
• Documentation for my change is up to date

---

Related Issues

[obabichevjb]

I don't know if it's ok or not, but to me it looks a bit suspicious at the current moment.

We actually modify the table object, so user will have changes in table object if he tries to create statement for that table.

The method createStatement() does not look like the one that should modify table object, so some users may be surprised

[joc-a]

@obabichevjb I agree. I didn't like this solution, to be frank, but I didn't have any other ideas. If you have any suggestions on how to do this in a cleaner way, please let me know.

[obabichevjb]

It's a good question, I'm not sure about what is the behavior we want to achieve.

If we want to guarantee that only long values would be stored in the database base and exception would be thrown otherwise, it could be problematic (like in the PR)

I tried to check SQLite, and I found out that if the column type INTEGER is used, then the column will store only Kotlin's Long values. But it would not complain if the user tried to save the value beyond Long.MIN_VALUE/Long.MAX_VALUE, it will just "cut" the values to the range.

[joc-a]

I've decided to simply exclude SQLite from this expected behaviour simply because the raw SQL does not have that behaviour to begin with, and it is a hassle for Exposed to try to do it. If a user wants to enforce it, they can simply add their own custom check constraint like this:

val long = long("long_column").check { column ->
    fun typeOf(value: String) = object : ExpressionWithColumnType<String>() {
        override fun toQueryBuilder(queryBuilder: QueryBuilder) = queryBuilder { append("typeof($value)") }
        override val columnType: IColumnType<String> = TextColumnType()
    }
    val typeCondition = Expression.build { typeOf(column.name) eq stringLiteral("integer") }
    if (column.columnType.nullable) {
        column.isNull() or typeCondition
    } else {
        typeCondition
    }
}

[joc-a]

@e5l What do you think?

[e5l]

I agree with @obabichevjb about modifying tables in createStatement, so I'm good if we're not doing that.

The one major thing to consider is a strict mode (like https://dev.mysql.com/doc/refman/8.4/en/sql-mode.html) or strict tables (https://www.sqlite.org/stricttables.html). We should mention these in the docs and check if we're working as expected if those are enabled.

[e5l]

left a comment