KDbQuerySchema.h

/* This file is part of the KDE project
   Copyright (C) 2003-2017 Jarosław Staniek <[email protected]>
 
   This library is free software; you can redistribute it and/or
   modify it under the terms of the GNU Library General Public
   License as published by the Free Software Foundation; either
   version 2 of the License, or (at your option) any later version.
 
   This library is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
   Library General Public License for more details.
 
   You should have received a copy of the GNU Library General Public License
   along with this library; see the file COPYING.LIB.  If not, write to
   the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
 * Boston, MA 02110-1301, USA.
*/
 
#ifndef KDB_QUERYSCHEMA_H
#define KDB_QUERYSCHEMA_H
 
#include <QVector>
#include <QHash>
 
#include "KDbFieldList.h"
#include "KDbObject.h"
#include "KDbQueryColumnInfo.h"
#include "KDbToken.h"
 
class KDbConnection;
class KDbOrderByColumn;
class KDbOrderByColumnList;
class KDbQueryAsterisk;
class KDbQuerySchemaFieldsExpanded;
class KDbQuerySchemaParameter;
class KDbQuerySchemaPrivate;
class KDbRelationship;
class KDbTableSchema;
 
//! @short KDbQuerySchema provides information about database query
/*! The query that can be executed using KDb-compatible SQL database engine
 or used as an introspection tool. KDb parser builds KDbQuerySchema objects
 by parsing SQL statements. */
class KDB_EXPORT KDbQuerySchema : public KDbFieldList, public KDbObject
{
public:
    /*! Creates empty query object (without columns). */
    KDbQuerySchema();
 
    /*! Creates query schema object that is equivalent to "SELECT * FROM table"
     sql command. Schema of @a table is used to contruct this query --
     it is defined by just adding all the fields to the query in natural order.
     To avoid problems (e.g. with fields added outside of Kexi using ALTER TABLE)
     we do not use "all-tables query asterisk" (see KDbQueryAsterisk) item to achieve
     this effect.
 
     Properties such as the name and caption of the query are inherited
     from table schema.
 
     We consider that query schema based on @a table is not (a least yet) stored
     in a system table, so query connection is set to @c nullptr
     (even if @a tableSchema's connection is not @c nullptr).
     Id of the created query is set to 0. */
    explicit KDbQuerySchema(KDbTableSchema *tableSchema);
 
    /*! Copy constructor. Creates deep copy of @a querySchema.
     KDbQueryAsterisk objects are deeply copied while only pointers to KDbField objects are copied. */
    KDbQuerySchema(const KDbQuerySchema& querySchema, KDbConnection *conn);
 
    ~KDbQuerySchema() override;
 
    /*! Inserts @a field to the columns list at @a position.
     Inserted field will not be owned by this KDbQuerySchema object,
     but by the corresponding KDbTableSchema.
 
     KDbQueryAsterisk can be also passed as @a field. See the KDbQueryAsterisk class
     description.
 
     @note After inserting a field, corresponding table will be automatically
     added to query's tables list if it is not present there (see tables()).
     KDbField must have its table assigned.
 
     The inserted field will be visible. Use insertInvisibleField(position, field)
     to add an invisible field.
 
     The field is not bound to any particular table within the query.
    */
    bool insertField(int position, KDbField *field) override;
 
    /**
     * @overload bool insertField(int position, KDbField *field)
     * Inserts @a field to the columns list at @a position.
     * @a bindToTable is a table index within the query for which the field should be bound.
     * If @a bindToTable is -1, no particular table will be bound.
     * @see tableBoundToColumn(int columnPosition)
     */
    bool insertField(int position, KDbField *field, int bindToTable);
 
    /**
     * @overload bool insertField(int position, KDbField *field)
     * Inserts @a field to the columns list at @a position.
     * In addition sets field's visibility to @c false. It will not be bound to any table in this query.
     * @since 3.1
     */
    bool insertInvisibleField(int position, KDbField *field);
 
    /**
     * @overload bool insertInvisibleField(int position, KDbField *field)
     * Inserts @a field to the columns list at @a position.
     * In addition sets field's visibility to @c false.
     * @a bindToTable is a table index within the query for which the field should be bound.
     * If @a bindToTable is -1, no particular table will be bound.
     * @see tableBoundToColumn(int columnPosition)
     * @since 3.1
     */
    bool insertInvisibleField(int position, KDbField *field, int bindToTable);
 
    /**
     * Appends @a field to the columns list.
     * The field will be visible. Use addInvisibleField(field) to add an invisible field.
     * The field is not bound to any particular table within the query.
     * @see insertField()
     */
    bool addField(KDbField* field);
 
    /*! Appends @a field to the columns list. Also binds to a table at @a bindToTable position.
     * Use bindToTable==-1 if no table should be bound.
     * The field will be visible. Use addInvisibleField(field, bindToTable) to add
     * an invisible field.
     * @see insertField()
     * @see tableBoundToColumn(int columnPosition)
     */
    bool addField(KDbField* field, int bindToTable);
 
    /**
     * @overload bool addField(KDbField* field)
     * Appends @a field to the columns list.
     * The field is not bound to any particular table within the query.
     * In addition sets field's visibility to @c false. It will not be bound to any table in this query.
     * @since 3.1
     */
    bool addInvisibleField(KDbField* field);
 
    /**
     * @overload bool addField(KDbField* field, int bindToTable)
     * Appends @a field to the columns list. Also binds to a table at @a bindToTable position.
     * In addition sets field's visibility to @c false.
     * @see tableBoundToColumn(int columnPosition)
     * @since 3.1
     */
    bool addInvisibleField(KDbField* field, int bindToTable);
 
    /*! Removes field from the columns list. Use with care. */
    bool removeField(KDbField *field) override;
 
    /**
     * Appends a column built on top of @a expr expression.
     * This creates a new KDbField object and adds it to the query schema using addField().
     */
    bool addExpression(const KDbExpression& expr);
 
    /**
     * @overload bool addExpression(const KDbExpression& expr)
     * Appends a column built on top of @a expr expression.
     * In addition sets column's visibility to @c false.
     * @since 3.1
     */
    bool addInvisibleExpression(const KDbExpression& expr);
 
    /*! @return visibility flag for column at @a position.
     By default column is visible. */
    bool isColumnVisible(int position) const;
 
    //! Sets visibility flag for column at @a position to @a visible.
    void setColumnVisible(int position, bool visible);
 
    /*! Appends @a asterisk at the and of columns list. */
    bool addAsterisk(KDbQueryAsterisk *asterisk);
 
    /**
     * @overload bool addAsterisk(KDbQueryAsterisk *asterisk)
     * Appends @a asterisk at the and of columns list.
     * Sets the asterisk as invisible.
     * @since 3.1
     */
    bool addInvisibleAsterisk(KDbQueryAsterisk *asterisk);
 
    /*! Removes all columns and their aliases from the columns list,
     removes all tables and their aliases from the tables list within this query.
     Sets master table information to @c nullptr.
     Does not destroy any objects though. Clears name and all other properties.
     @see KDbFieldList::clear() */
    void clear() override;
 
    /*! @return table that is master to this query.
     All potentially-editable columns within this query belong just to this table.
     This method also can return @c nullptr if there are no tables at all,
     or if previously assigned master table schema has been removed
     with removeTable().
     Every query that has at least one table defined, should have
     assigned a master table.
     If no master table is assigned explicitly, but only one table used in this query,
     a single table is returned here, even if there are table aliases,
     (e.g. "T" table is returned for "SELECT T1.A, T2.B FROM T T1, T T2" statement). */
    KDbTableSchema* masterTable() const;
 
    /*! Sets master table of this query to @a table.
      This table should be also added to query's tables list
      using addTable(). If @a table equals @c nullptr, nothing is performed.
      @see masterTable() */
    void setMasterTable(KDbTableSchema *table);
 
    /*! @return list of tables used in this query.
     It is never @c nullptr. The list also includes master table.
     @see masterTable() */
    QList<KDbTableSchema*>* tables() const;
 
    /*! Appends @a table schema as one of tables used in a query.
     If @a alias is not empty, it will be assigned to this table
     using setTableAlias(position, alias). */
    void addTable(KDbTableSchema *table, const QString& alias = QString());
 
    /*! Removes @a table schema from this query.
     This does not destroy @a table object but only takes it out of the list.
     If this table was master for the query, master table information is also
     invalidated. */
    void removeTable(KDbTableSchema *table);
 
    /*! @return table with name @a tableName or 0 if this query has no such table. */
    KDbTableSchema* table(const QString& tableName) const;
 
    /*! @return @c true if the query uses @a table. */
    bool contains(KDbTableSchema *table) const;
 
    /*! Convenience function.
     @return table field by searching through all tables in this query.
     The field does not need to be included on the list of query columns.
     Similarly, query aliases are not taken into account.
 
     @a fieldOrTableAndFieldName string may contain table name and field name
     with '.' character between them, e.g. "mytable.myfield".
     This is recommended way to avoid ambiguity.
     0 is returned if the query has no such
     table defined of the table has no such field defined.
     If you do not provide a table name, the first field found is returned.
 
     KDbQuerySchema::table("mytable")->field("myfield") could be
     alternative for findTableField("mytable.myfield") but it can crash
     if "mytable" is not defined in the query.
 
     @see KDb::splitToTableAndFieldParts()
    */
    KDbField* findTableField(const QString &fieldOrTableAndFieldName) const;
 
    /*! @return alias of a column at @a position or empty string if there is no alias for this column
     or if there is no such column within the query defined. If the column is an expression and has
     no alias defined, a new unique alias will be generated automatically on this call.
    */
    QString columnAlias(int position) const;
 
    /*! @return number of column aliases */
    int columnAliasesCount() const;
 
    /*! Provided for convenience.
     @return @c true if a column at @a position has non empty alias defined
     within the query.
     If there is no alias for this column,
     or if there is no such column in the query defined, @c false is returned. */
    bool hasColumnAlias(int position) const;
 
    /*! Sets @a alias for a column at @a position, within the query.
     Passing empty string to @a alias clears alias for a given column. */
    bool setColumnAlias(int position, const QString& alias);
 
    /*! @return a table position (within FROM section),
     that is bound to column at @a columnPosition (within SELECT section).
     This information can be used to find if there is alias defined for
     a table that is referenced by a given column.
 
     For example, for "SELECT t2.id FROM table1 t1, table2 t2" query statement,
     columnBoundToTable(0) returns 1, what means that table at position 1
     (within FROM section) is bound to column at position 0, so we can
     now call tableAlias(1) to see if we have used alias for this column (t2.id)
     or just a table name (table2.id).
 
     These checks are performed e.g. by KDbConnection::selectStatement()
     to construct a statement string maximally identical to originally
     defined query statement.
 
     -1 is returned if:
      - @a columnPosition is out of range (i.e. < 0 or >= fieldCount())
      - a column at @a columnPosition is not bound to any table (i.e.
        no database field is used for this column,
        e.g. "1" constant for "SELECT 1 from table" query statement)
    */
    int tableBoundToColumn(int columnPosition) const;
 
    /*! @return number of table aliases */
    int tableAliasesCount() const;
 
    /*! @return alias of a table at @a position (within FROM section)
     or null string if there is no alias for this table
     or if there is no such table within the query defined. */
    QString tableAlias(int position) const;
 
    /*! @return alias of a table @a tableName (within FROM section)
     or empty value if there is no alias for this table
     or if there is no such table within the query defined. */
    QString tableAlias(const QString& tableName) const;
 
    /*! @return alias of a table @a tableName (within FROM section).
    If there is no alias for this table, its name is returned.
    Empty value is returned if there is no such table within the query defined. */
    QString tableAliasOrName(const QString& tableName) const;
 
    /*! @return table position (within FROM section) that has attached
     alias @a name.
     If there is no such alias, -1 is returned.
     Only first table's position attached for this alias is returned.
     It is not especially bad, since aliases rarely can be duplicated,
     what leads to ambiguity.
     Duplicated aliases are only allowed for trivial queries that have
     no database fields used within their columns,
     e.g. "SELECT 1 from table1 t, table2 t" is ok
     but "SELECT t.id from table1 t, table2 t" is not.
    */
    int tablePositionForAlias(const QString& name) const;
 
    /*! @return position (within the FROM section) of table @a tableName.
     -1 is returned if there's no such table declared in the FROM section.
     @see tablePositions()
    */
    int tablePosition(const QString& tableName) const;
 
    /*! @return a list of all occurrences of table @a tableName (within the FROM section).
     E.g. for "SELECT * FROM table t, table t2" tablePositions("table") returns {0, 1} list.
     Empty list is returned if there's no table @a tableName used in the FROM section at all.
     @see tablePosition() */
    QList<int> tablePositions(const QString& tableName) const;
 
    /*! Provided for convenience.
     @return @c true if a table at @a position (within FROM section of the query)
     has non empty alias defined.
     If there is no alias for this table,
     or if there is no such table in the query defined, @c false is returned. */
    bool hasTableAlias(int position) const;
 
    /*! Provided for convenience.
     @return @c true if non empty table alias @a name is defined for a table. */
    bool hasTableAlias(const QString &name) const;
 
    /*! @return column position that has defined alias @a name.
     If there is no such alias, -1 is returned. */
    int columnPositionForAlias(const QString& name) const;
 
    /*! Provided for convenience.
     @return @c true if non empty alias @a name is defined for any column. */
    bool hasColumnAlias(const QString &name) const;
 
    /*! Sets @a alias for a table at @a position (within FROM section
     of the query).
     Passing empty sting to @a alias clears alias for a given table
     (only for specified @a position). */
    bool setTableAlias(int position, const QString& alias);
 
    /*! @return a list of relationships defined for this query.
     It is never @c nullptr.*/
    QList<KDbRelationship*>* relationships() const;
 
    /*! Appends a new relationship defined by @a field1 and @a field2.
     Both fields should belong to two different tables of this query.
     This is convenience function useful for a typical cases.
     It automatically creates KDbRelationship object for this query.
     If one of the fields are primary keys, it will be detected
     and appropriate master-detail relation will be established.
     This functiuon does nothing if the arguments are invalid. */
    KDbRelationship* addRelationship(KDbField *field1, KDbField *field2);
 
    /*! @return list of KDbQueryAsterisk objects defined for this query.
     It is never @c nullptr. */
    KDbField::List* asterisks() const;
 
    //! Mode for field() and columnInfo()
    //! @since 3.1
    enum class ExpandMode {
        Unexpanded, //!< All fields are returned even if duplicated
        Expanded    //!< Expanded list of the query fields is computed so queries with asterisks
                    //!< are processed well
    };
 
    /*! @return field for @a identifier or @c nullptr if no field for this name
     was found within the query. fieldsExpanded() method is used
     to lookup expanded list of the query fields, so queries with asterisks
     are processed well.
     If a field has alias defined, name is not taken into account,
     but only its alias. If a field has no alias:
     - field's name is checked
     - field's table and field's name are checked in a form of "tablename.fieldname",
       so you can provide @a identifier in this form to avoid ambiguity.
 
     If there are more than one fields with the same name equal to @a identifier,
     first-found is returned (checking is performed from first to last query field).
     Structures needed to compute result of this method are cached,
     so only first usage costs o(n) - another usages cost o(1).
 
     Example:
     Let query be defined by "SELECT T.B AS X, T.* FROM T" statement and let T
     be table containing fields A, B, C.
     Expanded list of columns for the query is: T.B AS X, T.A, T.B, T.C.
     - Calling field("B") will return a pointer to third query column (not the first,
       because it is covered by "X" alias). Additionally, calling field("X")
       will return the same pointer.
     - Calling field("T.A") will return the same pointer as field("A").
 
     This method is also a product of inheritance from KDbFieldList.
     */
    const KDbField *field(KDbConnection *conn, const QString &identifier,
                          ExpandMode mode = ExpandMode::Expanded) const;
 
    /**
     * @overload
     */
    KDbField *field(KDbConnection *conn, const QString &identifier,
                    ExpandMode mode = ExpandMode::Expanded);
 
    /*! @return field id or @c nullptr if there is no such a field. */
    KDbField* field(int id) override;
 
    using KDbFieldList::field;
 
    /*! @overload KDbField* field(int id) */
    const KDbField* field(int id) const override;
 
    /*! Like KDbQuerySchema::field(const QString& name) but returns not only KDbField
     object for @a identifier but entire KDbQueryColumnInfo object.
     @a identifier can be:
     - a fieldname
     - an aliasname
     - a tablename.fieldname
     - a tablename.aliasname
     Note that if there are two occurrrences of the same name,
     only the first is accessible using this method. For instance,
     calling columnInfo("name") for "SELECT t1.name, t2.name FROM t1, t2" statement
     will only return the column related to t1.name and not t2.name, so you'll need to
     explicitly specify "t2.name" as the identifier to get the second column. */
    KDbQueryColumnInfo *columnInfo(KDbConnection *conn, const QString &identifier,
                                   ExpandMode mode = ExpandMode::Expanded) const;
 
    //! Mode for fieldsExpanded() and visibleFieldsExpanded()
    //! @since 3.1
    enum class FieldsExpandedMode {
        Default,                      //!< All fields are returned even if duplicated
        Unique,                       //!< Unique list of fields is returned
        WithInternalFields,           //!< Like Default but internal fields (for lookup) are appended
        WithInternalFieldsAndRecordId //!< Like WithInternalFields but record ID (big int type) field
                                      //!< is appended after internal fields
    };
 
    /*! @return fully expanded list of fields.
     KDbQuerySchema::fields() returns vector of fields used for the query columns,
     but in a case when there are asterisks defined for the query,
     it does not expand KDbQueryAsterisk objects to field lists but return every
     asterisk as-is.
     This could be inconvenient when you need just a fully expanded list of fields,
     so this method does the work for you.
 
     If @a options is Unique, each field is returned in the vector only once
     (first found field is selected).
     Note however, that the same field can be returned more than once if it has attached
     a different alias.
     For example, let t be TABLE( a, b ) and let query be defined
     by "SELECT *, a AS alfa FROM t" statement. Both fieldsExpanded(Default)
     and fieldsExpanded(Unique) will return [ a, b, a (alfa) ] list.
     On the other hand, for query defined by "SELECT *, a FROM t" statement,
     fieldsExpanded(Default) will return [ a, b, a ] list while
     fieldsExpanded(Unique) will return [ a, b ] list.
 
     If @a options is WithInternalFields or WithInternalFieldsAndRecordID,
     additional internal fields are also appended to the vector.
 
     If @a options is WithInternalFieldsAndRecordId,
     one fake BigInteger column is appended to make space for Record ID column used
     by KDbCursor implementations. For example, let city_id in TABLE persons(surname, city_id)
     reference cities.id in TABLE cities(id, name) and let query q be defined
     by "SELECT * FROM persons" statement. We want to display persons' city names instead of city_id's.
     To do this, cities.name has to be retrieved as well, so the following statement should be used:
     "SELECT * FROM persons, cities.name LEFT OUTER JOIN cities ON persons.city_id=cities.id".
     Thus, calling fieldsExpanded(WithInternalFieldsAndRecordId) will return 4 elements instead of 2:
     persons.surname, persons.city_id, cities.name, {ROWID}. The {ROWID} item is the placeholder
     used for fetching ROWID by KDb cursors.
 
     By default, all fields are returned in the vector even
     if there are multiple occurrences of one or more (options == Default).
 
     Note: You should assign the resulted vector in your space - it will be shared
     and implicity copied on any modification.
     This method's result is cached by KDbQuerySchema object.
    @todo js: UPDATE CACHE!
    */
    inline KDbQueryColumnInfo::Vector fieldsExpanded(
            KDbConnection *conn, FieldsExpandedMode mode = FieldsExpandedMode::Default) const
    {
        return fieldsExpandedInternal(conn, mode, false);
    }
 
    /*! Like fieldsExpanded() but returns only visible fields. */
    inline KDbQueryColumnInfo::Vector visibleFieldsExpanded(
            KDbConnection *conn, FieldsExpandedMode options = FieldsExpandedMode::Default) const
    {
        return fieldsExpandedInternal(conn, options, true);
    }
 
    /*! @return list of internal fields used for lookup columns. */
    KDbQueryColumnInfo::Vector internalFields(KDbConnection *conn) const;
 
    /*! @return info for expanded of internal field at index @a index.
     The returned field can be either logical or internal (for lookup),
     the latter case is @c true if @a index &gt;= fieldsExpanded().count().
     Equivalent of KDbQuerySchema::fieldsExpanded(WithInternalFields).at(index). */
    KDbQueryColumnInfo* expandedOrInternalField(KDbConnection *conn, int index) const;
 
    //! Mode for columnsOrder()
    //! @since 3.1
    enum class ColumnsOrderMode {
        UnexpandedList,                 //!< A map for unexpanded list is created
        UnexpandedListWithoutAsterisks, //!< A map for unexpanded list is created, with asterisks skipped
        ExpandedList                    //!< A map for expanded list is created
    };
 
    /*! @return a hash for fast lookup of query columns' order.
     - If @a options is UnexpandedList, each KDbQueryColumnInfo pointer is mapped to the index
       within (unexpanded) list of fields, i.e. "*" or "table.*" asterisks are considered
       to be single items.
     - If @a options is UnexpandedListWithoutAsterisks, each KDbQueryColumnInfo pointer
       is mapped to the index within (unexpanded) list of columns that come from asterisks
       like "*" or "table.*" are not included in the map at all.
     - If @a options is ExpandedList (the default) this method provides is exactly opposite
       information compared to vector returned by fieldsExpanded().
 
     This method's result is cached by the KDbQuerySchema object.
     Note: indices of internal fields (see internalFields()) are also returned
     here - in this case the index is counted as a sum of size(e) + i (where "e" is
     the list of expanded fields and i is the column index within internal fields list).
     This feature is used eg. at the end of KDbConnection::updateRecord() where need indices of
     fields (including internal) to update all the values in memory.
 
     Example use: let t be table (int id, name text, surname text) and q be query
     defined by a statement "select * from t".
 
     - columnsOrder(ExpandedList) will return the following map: KDbQueryColumnInfo(id)->0,
       KDbQueryColumnInfo(name)->1, KDbQueryColumnInfo(surname)->2.
     - columnsOrder(UnexpandedList) will return the following map: KDbQueryColumnInfo(id)->0,
       KDbQueryColumnInfo(name)->0, KDbQueryColumnInfo(surname)->0 because the column
       list is not expanded. This way you can use the returned index to get KDbField*
       pointer using field(int) method of KDbFieldList superclass.
     - columnsOrder(UnexpandedListWithoutAsterisks) will return the following map:
       KDbQueryColumnInfo(id)->0,
    */
    QHash<KDbQueryColumnInfo *, int> columnsOrder(KDbConnection *conn,
                                                  ColumnsOrderMode mode = ColumnsOrderMode::ExpandedList) const;
 
    /*! @return table describing order of primary key (PKEY) fields within the query.
     Indexing is performed against vector returned by fieldsExpanded().
     It is usable for e.g. Connection::updateRecord(), when we need
     to locate each primary key's field in a constant time.
 
     Returned vector is owned and cached by KDbQuerySchema object. When you assign it,
     it is implicity shared. Its size is equal to number of primary key
     fields defined for master table (masterTable()->primaryKey()->fieldCount()).
 
     Each element of the returned vector:
     - can belong to [0..fieldsExpanded().count()-1] if there is such
       primary key's field in the fieldsExpanded() list.
     - can be equal to -1 if there is no such primary key's field
       in the fieldsExpanded() list.
 
     If there are more than one primary key's field included in the query,
     only first-found column (oin the fieldsExpanded() list) for each pkey's field is included.
 
     Returns empty vector if there is no master table or no master table's pkey.
     @see example for pkeyFieldCount().
    @todo js: UPDATE CACHE!
    */
    QVector<int> pkeyFieldsOrder(KDbConnection *conn) const;
 
    /*! @return number of master table's primary key fields included in this query.
     This method is useful to quickly check whether the vector returned by pkeyFieldsOrder()
     if filled completely.
 
     User e.g. in KDbConnection::updateRecord() to check if entire primary
     key information is specified.
 
     Examples: let table T has (ID1 INTEGER, ID2 INTEGER, A INTEGER) fields,
     and let (ID1, ID2) is T's primary key.
     -# The query defined by "SELECT * FROM T" statement contains all T's
        primary key's fields as T is the master table, and thus pkeyFieldCount()
        will return 2 (both primary key's fields are in the fieldsExpanded() list),
        and pkeyFieldsOrder() will return vector {0, 1}.
     -# The query defined by "SELECT A, ID2 FROM T" statement, and thus pkeyFieldCount()
        will return 1 (only one primary key's field is in the fieldsExpanded() list),
        and pkeyFieldsOrder() will return vector {-1, 1}, as second primary key's field
        is at position #1 and first field is not specified at all within the query.
    */
    int pkeyFieldCount(KDbConnection *conn);
 
    /*! @return a list of field infos for all auto-incremented fields
     from master table of this query. This result is cached for efficiency.
     fieldsExpanded() is used for that.
    */
    KDbQueryColumnInfo::List* autoIncrementFields(KDbConnection *conn) const;
 
    /*! @return a preset statement (if any). */
    KDbEscapedString statement() const;
 
    /*! Forces a raw SQL statement @a sql for the query. This means that no statement is composed
     * from KDbQuerySchema's content. */
    void setStatement(const KDbEscapedString& sql);
 
    /*! @return a string that is a result of concatenating all column names
     for @a infolist, with "," between each one.
     This is usable e.g. as argument like "field1,field2"
     for "INSERT INTO (xxx) ..". The result of this method is effectively cached,
     and it is invalidated when set of fields changes (e.g. using clear()
     or addField()).
 
     This method is similar to KDbFieldList::sqlFieldsList() it just uses
     KDbQueryColumnInfo::List instead of KDbField::List.
 
     @a escapingType can be used to alter default escaping type.
     If @a conn is not provided for DriverEscaping, no escaping is performed. */
    static KDbEscapedString
    sqlColumnsList(const KDbQueryColumnInfo::List &infolist, KDbConnection *conn = nullptr,
                   KDb::IdentifierEscapingType escapingType = KDb::DriverEscaping);
 
    /*! @return cached list of autoincrement fields created using sqlColumnsList()
        on a list returned by autoIncrementFields(). The field names are escaped using
        driver escaping. */
    KDbEscapedString autoIncrementSqlFieldsList(KDbConnection *conn) const;
 
    /**
     * @brief Sets a WHERE expression @a exp.
     *
     * Previously set WHERE expression will be removed. A null expression
     * (KDbExpression()) can be passed to remove existing WHERE expresssion.
     * @return @c false if @a expr is not a valid WHERE expression. validate() is called
     * to check this. On failure the WHERE expression for this query is cleared. In this
     * case a string pointed by @a errorMessage (if provided) is set to a general error
     * message and a string pointed by @a errorDescription (if provided) is set to a
     * detailed description of the error.
     */
    bool setWhereExpression(const KDbExpression &expr, QString *errorMessage = nullptr,
                            QString *errorDescription = nullptr);
 
    /*! @return WHERE expression or 0 if this query has no WHERE expression */
    KDbExpression whereExpression() const;
 
    /**
     * @brief Appends a part to WHERE expression.
     *
     * Simplifies creating of WHERE expression if used instead of setWhereExpression().
     * @return @c false if the newly constructed WHERE expression is not valid.
     * validate() is called to check this. On failure the WHERE expression for this query
     * is left unchanged. In this case a string pointed by @a errorMessage (if provided)
     * is set to a general error message and a string pointed by @a errorDescription
     * (if provided) is set to a detailed description of the error.
     */
    bool addToWhereExpression(KDbField *field, const QVariant &value,
                              KDbToken relation = '=', QString *errorMessage = nullptr,
                              QString *errorDescription = nullptr);
 
    /*! Sets a list of columns for ORDER BY section of the query.
     Each name on the list must be a field or alias present within the query
     and must not be covered by aliases. If one or more names cannot be found
     within the query, the method will have no effect.
     Any previous ORDER BY settings will be removed.
 
     Note that this information is cleared whenever you call methods that
     modify list of columns (KDbQueryColumnInfo), i.e. insertField(),
     addField(), removeField(), addExpression(), etc.
     (because KDbOrderByColumn items can point to a KDbQueryColumnInfo that's removed by these
     methods), so you should use setOrderByColumnList() method after the query
     is completely built. */
    void setOrderByColumnList(const KDbOrderByColumnList& list);
 
    /*! @return a list of columns listed in ORDER BY section of the query.
     Read notes for @ref setOrderByColumnList(). */
    KDbOrderByColumnList* orderByColumnList();
 
    /*! @see orderByColumnList() */
    const KDbOrderByColumnList* orderByColumnList() const;
 
    /*! @return query schema parameters. These are taked from the WHERE section
     (a tree of expression items). */
    QList<KDbQuerySchemaParameter> parameters(KDbConnection *conn) const;
 
    //! @return @c true if this query is valid
    /*! Detailed validation is performed in the same way as parsing of query statements
     * by the KDbParser.
     * Example :Let the query be "SELECT <fields> FROM <tables> WHERE <whereExpression>".
     * First each field from <fields> (@see fields()) is validated using
     * KDbField::expression().validate(). Then the <whereExpression> (@see
     * whereExpression())
     * is validated using KDbExpression::validate().
     *
     * On error a string pointed by @a errorMessage (if provided) is set to a general
     * error message and a string pointed by @a errorDescription (if provided) is set to a
     * detailed description of the error.
     */
    //! @todo add tests
    bool validate(QString *errorMessage = nullptr, QString *errorDescription = nullptr);
 
protected:
    KDbQuerySchemaFieldsExpanded *computeFieldsExpanded(KDbConnection *conn) const;
 
    //! Used by fieldsExpanded(KDbConnection*, FieldsExpandedMode)
    //! and visibleFieldsExpanded(KDbConnection*, FieldsExpandedMode).
    KDbQueryColumnInfo::Vector fieldsExpandedInternal(KDbConnection *conn,
                                                      FieldsExpandedMode mode,
                                                      bool onlyVisible) const;
 
    /** Internal method used by all insert*Field methods.
     * The new column can also be explicitly bound to a specific position on tables list.
     * @a bindToTable is a table index within the query for which the field should be bound.
     * If @a bindToTable is -1, no particular table will be bound.
     * @see tableBoundToColumn(int columnPosition)
     */
    bool insertFieldInternal(int position, KDbField *field, int bindToTable, bool visible);
 
    /**
     * Internal method used by add*Asterisk() methods.
     * Appends @a asterisk at the and of columns list, sets visibility.
     */
    bool addAsteriskInternal(KDbQueryAsterisk *asterisk, bool visible);
 
    /** Internal method used by all add*Expression methods.
     * Appends expression @a expr at the and of columns list, sets visibility.
     */
    bool addExpressionInternal(const KDbExpression& expr, bool visible);
 
    /** Internal method used by a query parser.
     */
    void setWhereExpressionInternal(const KDbExpression &expr);
 
    friend class KDbQuerySchemaPrivate;
 
    Q_DISABLE_COPY(KDbQuerySchema)
    KDbQuerySchemaPrivate * const d;
};
 
//! A pair (connection, table-or-schema) for QDebug operator<<
//! @since 3.1
typedef std::tuple<KDbConnection*, const KDbQuerySchema&> KDbConnectionAndQuerySchema;
 
//! Sends connection and query schema information @a connectionAndSchema to debug output @a dbg.
//! @since 3.1
KDB_EXPORT QDebug operator<<(QDebug dbg,
                             const KDbConnectionAndQuerySchema &connectionAndSchema);
 
#endif