Please wait. This can take some minutes ...
Many resources are needed to download a project. Please understand that we have to compensate our server costs. Thank you in advance.
Project price only 1 $
You can buy this project and download/modify it how often you want.
.circumflex-orm.2.3.source-code.dialect.scala Maven / Gradle / Ivy
package ru.circumflex
package orm
import java.sql._
/*!# Dialect
This little thingy does all dirty SQL rendering.
We are orienting the default dialect on the world's most advanced open-source
database, [PostgreSQL][psql].
If you feel that some of the statements do not work with your RDBMS vendor,
trace the exact method and provide it's implementation in your own class.
After that, set the `orm.dialect` configuration parameter accordingly.
[psql]: http://postgresql.org
*/
class Dialect {
def driverClass: String =
throw new ORMException("Missing mandatory configuration parameter 'orm.connection.driver'.")
/*!## JDBC methods */
def prepareStatement(conn: Connection, sql: String): PreparedStatement =
conn.prepareStatement(sql)
/*!## SQL types */
def longType = "BIGINT"
def integerType = "INTEGER"
def numericType(precision: Int, scale: Int): String =
"NUMERIC" + (if (precision == -1) "" else "(" + precision + "," + scale + ")")
def textType = "TEXT"
def varcharType(length: Int): String =
"VARCHAR" + (if (length == -1) "" else "(" + length + ")")
def booleanType = "BOOLEAN"
def dateType = "DATE"
def timeType = "TIME"
def timestampType = "TIMESTAMP"
def xmlType = "XML"
def binaryType = "BYTEA"
/*!## Actions for Foreign Keys */
def fkNoAction = "NO ACTION"
def fkCascade = "CASCADE"
def fkRestrict = "RESTRICT"
def fkSetNull = "SET NULL"
def fkSetDefault = "SET DEFAULT"
/*!## Join Keywords */
def innerJoin = "INNER JOIN"
def leftJoin = "LEFT JOIN"
def rightJoin = "RIGHT JOIN"
def fullJoin = "FULL JOIN"
/*!## Predicates */
def EQ(ex1: String, ex2: String = "?") = ex1 + " = " + ex2
def NE(ex1: String, ex2: String = "?") = ex1 + " <> " + ex2
def GT(ex1: String, ex2: String = "?") = ex1 + " > " + ex2
def GE(ex1: String, ex2: String = "?") = ex1 + " >= " + ex2
def LT(ex1: String, ex2: String = "?") = ex1 + " < " + ex2
def LE(ex1: String, ex2: String = "?") = ex1 + " <= " + ex2
def emptyPredicate = "1 = 1"
def IS_NULL(ex: String) = ex + " IS NULL"
def IS_NOT_NULL(ex: String) = ex + " IS NOT NULL"
def LIKE(ex1: String, ex2: String = "?") = ex1 + " LIKE " + ex2
def ILIKE(ex1: String, ex2: String = "?") = ex1 + " ILIKE " + ex2
def BETWEEN(ex: String, v1: String = "?", v2: String= "?") =
ex + " BETWEEN " + v1 + " AND " + v2
def IN(ex: String) = ex + " IN"
def NOT_IN(ex: String) = ex + " NOT IN"
def parameterizedIn(ex: String, params: Iterable[String]) =
ex + " IN (" + params.mkString(", ") + ")"
def AND = "AND"
def OR = "OR"
def NOT = "NOT"
def ALL = "ALL"
def SOME = "SOME"
def EXISTS = "EXISTS"
def NOT_EXISTS = "NOT EXISTS"
/*!## Functions and others */
def NULL = "NULL"
def DISTINCT = "DISTINCT"
def COUNT(ex: String) = "COUNT(" + ex + ")"
def COUNT_DISTINCT(ex: String) = "COUNT(DISTINCT " + ex + ")"
def MAX(ex: String) = "MAX(" + ex + ")"
def MIN(ex: String) = "MIN(" + ex + ")"
def SUM(ex: String) = "SUM(" + ex + ")"
def AVG(ex: String) = "AVG(" + ex + ")"
/*!## Set operations */
def UNION = "UNION"
def UNION_ALL = "UNION ALL"
def EXCEPT = "EXCEPT"
def EXCEPT_ALL = "EXCEPT ALL"
def INTERSECT = "INTERSECT"
def INTERSECT_ALL = "INTERSECT ALL"
/*!## Order specificators */
def asc = "ASC"
def desc = "DESC"
/*!## Param placeholders */
def placeholder = "?"
def xmlPlaceholder = "XMLPARSE(DOCUMENT ?)"
/*!## Features Compliance */
def supportsSchema: Boolean = true
def supportsDropConstraints: Boolean = true
/*!## Commons */
def quoteLiteral(expr: String) = "'" + expr.replace("'", "''") + "'"
def escapeParameter(value: Any): String = value match {
case Some(v) => escapeParameter(v)
case None | null => "NULL"
case v => quoteLiteral(v.toString)
}
def relationQualifiedName(relation: Relation[_, _]) =
if (supportsSchema) relation.schema.name + "." + relation.relationName
else relation.relationName
def alias(expression: String, alias: String) =
expression + " AS " + alias
def qualifyColumn(vh: ValueHolder[_, _], tableAlias: String) =
tableAlias + "." + vh.name
def on(expression: Expression) = "ON (" + expression.toInlineSql + ")"
def not(expression: String) = "NOT (" + expression + ")"
def subquery(expression: String, subquery: SQLQuery[_]) =
expression + " ( " + subquery.toSql + " )"
/*!## Data Definition Language */
def constraintDefinition(constraint: Constraint) =
"CONSTRAINT " + constraint.constraintName + " " + constraint.sqlDefinition
def alterTable(rel: Relation[_, _], action: String) =
"ALTER TABLE " + rel.qualifiedName + " " + action
def alterTableAddConstraint(constraint: Constraint) =
alterTable(constraint.relation, "ADD " + constraintDefinition(constraint));
def alterTableDropConstraint(constraint: Constraint) =
alterTable(constraint.relation, "DROP CONSTRAINT " + constraint.constraintName);
def createSchema(schema: Schema) = "CREATE SCHEMA " + schema.name
def dropSchema(schema: Schema) = "DROP SCHEMA " + schema.name + " CASCADE"
def createTable[PK, R <: Record[PK, R]](table: Table[PK, R]) =
"CREATE TABLE " + table.qualifiedName + " (" +
table.fields.map(_.toSql).mkString(", ") +
", PRIMARY KEY (" + table.PRIMARY_KEY.name + "))"
def dropTable[PK, R <: Record[PK, R]](table: Table[PK, R]) =
"DROP TABLE " + table.qualifiedName
def createView[PK, R <: Record[PK, R]](view: View[PK, R]) =
"CREATE VIEW " + view.qualifiedName + " (" +
view.fields.map(_.name).mkString(", ") + ") AS " +
view.query.toInlineSql
def dropView[PK, R <: Record[PK, R]](view: View[PK, R]) =
"DROP VIEW " + view.qualifiedName
def createIndex(idx: Index): String = {
var result = "CREATE "
if (idx.isUnique) result += "UNIQUE "
result += "INDEX " + idx.name + " ON " + idx.relation.qualifiedName +
" USING " + idx.usingClause + " (" + idx.expression + ")"
if (idx.whereClause != EmptyPredicate)
result += " WHERE " + idx.whereClause.toInlineSql
result
}
def dropIndex(idx: Index) =
"DROP INDEX " + idx.relation.schema.name + "." + idx.name
def columnDefinition[R <: Record[_, R]](field: Field[_, R]): String = {
var result = field.name + " " + field.sqlType
if (field.isNotNull) result += " NOT NULL"
if (field.isUnique) result += " UNIQUE"
result += defaultExpression(field)
result
}
def compositeFieldName(names: String*): String = names.mkString(", ")
def initializeRelation[R <: Record[_, R]](relation: Relation[_, R]) {}
def initializeField[R <: Record[_, R]](field: Field[_, R]) = field match {
case f: AutoIncrementable[_, _]
if (f.isAutoIncrement && !field.record.relation.isInstanceOf[View[_, R]]) => {
val seqName = sequenceName(f)
val seq = new SchemaObject {
val objectName = "SEQUENCE " + seqName
val sqlDrop = "DROP SEQUENCE " + seqName
val sqlCreate = "CREATE SEQUENCE " + seqName
}
f.record.relation.addPreAux(seq)
}
case _ =>
}
def defaultExpression[R <: Record[_, R]](field: Field[_, R]): String =
field match {
case a: AutoIncrementable[_, _] if (a.isAutoIncrement) =>
" DEFAULT NEXTVAL('" + sequenceName(field) + "')"
case _ =>
field.defaultExpression.map(" DEFAULT " + _).getOrElse("")
}
def sequenceName[R <: Record[_, R]](vh: ValueHolder[_, R]) =
vh.record.relation.schema.name + "." +
vh.record.relation.relationName + "_" + vh.name + "_seq"
def uniqueKeyDefinition(uniq: UniqueKey) =
"UNIQUE (" + uniq.columns.map(_.name).mkString(", ") + ")"
def foreignKeyDefinition(fk: ForeignKey) =
"FOREIGN KEY (" + fk.childColumns.map(_.name).mkString(", ") +
") REFERENCES " + fk.parentRelation.qualifiedName + " (" +
fk.parentColumns.map(_.name).mkString(", ") + ") " +
"ON DELETE " + fk.onDelete.toSql + " " +
"ON UPDATE " + fk.onUpdate.toSql
def checkConstraintDefinition(check: CheckConstraint) =
"CHECK (" + check.expression + ")"
/*!## Structured Query Language */
def join(j: JoinNode[_, _, _, _]): String = joinInternal(j, null)
protected def joinInternal(node: RelationNode[_, _], on: String): String = {
var result = ""
node match {
case j: JoinNode[_, _, _, _] =>
result += joinInternal(j.left, on) +
" " + j.joinType.toSql + " " +
joinInternal(j.right, j.sqlOn)
case _ =>
result += node.toSql
if (on != null) result += " " + on
}
result
}
def select(q: Select[_]): String = {
var result = "SELECT "
if (q.isDistinct)
result += "DISTINCT "
result += q.projections.map(_.toSql).mkString(", ")
if (q.fromClause.size > 0)
result += " FROM " + q.fromClause.map(_.toSql).mkString(", ")
if (q.whereClause != EmptyPredicate)
result += " WHERE " + q.whereClause.toSql
if (q.groupByClause != "")
result += " GROUP BY " + q.groupByClause
if (q.havingClause != EmptyPredicate)
result += " HAVING " + q.havingClause.toSql
q.setOps.foreach {
case (op: SetOperation, subq: SQLQuery[_]) =>
result += " " + op.toSql + " ( " + subq.toSql + " )"
case _ =>
}
if (q.orderByClause.size > 0)
result += " ORDER BY " + q.orderByClause.map(_.toSql).mkString(", ")
if (q.limit > -1)
result += " LIMIT " + q.limit
if (q.offset > 0)
result += " OFFSET " + q.offset
result
}
def identityLastIdPredicate[PK, R <: Record[PK, R]](node: RelationNode[PK, R]): Predicate =
new SimpleExpression(node.alias + "." + node.relation.PRIMARY_KEY.name + " = LASTVAL()", Nil)
def identityLastIdQuery[PK, R <: Record[PK, R]](node: RelationNode[PK, R]): SQLQuery[PK] =
new Select(expr[PK]("LASTVAL()"))
def sequenceNextValQuery[PK, R <: Record[PK, R]](node: RelationNode[PK, R]): SQLQuery[PK] =
new Select(expr[PK]("NEXTVAL('" + sequenceName(node.relation.PRIMARY_KEY) + "')"))
/*!## Data Manipulation Language */
def insert[PK, R <: Record[PK, R]](dml: Insert[PK, R]): String = {
var result = "INSERT INTO " + dml.relation.qualifiedName
if (dml.fields.size > 0)
result += " (" + dml.fields.map(_.name).mkString(", ") +
") VALUES (" + dml.fields.map(_.placeholder).mkString(", ") + ")"
else result += " DEFAULT VALUES"
result
}
def insertSelect[PK, R <: Record[PK, R]](dml: InsertSelect[PK, R]) =
"INSERT INTO " + dml.relation.qualifiedName + " (" +
dml.relation.fields.map(_.name).mkString(", ") + ") " + dml.query.toSql
def update[PK, R <: Record[PK, R]](dml: Update[PK, R]): String = {
var result = "UPDATE " + dml.node.toSql + " SET " +
dml.setClause.map(f => f._1.name + " = " + f._1.placeholder).mkString(", ")
if (dml.whereClause != EmptyPredicate) result += " WHERE " + dml.whereClause.toSql
result
}
def delete[PK, R <: Record[PK, R]](dml: Delete[PK, R]): String = {
var result = "DELETE FROM " + dml.node.toSql
if (dml.whereClause != EmptyPredicate) result += " WHERE " + dml.whereClause.toSql
result
}
}