SQL persistent database interface

Author(s): I. Caballero, D. Cabeza, J.M. Gómez, M. Hermenegildo, J. F. Morales, and M. Carro, clip@dia.fi.upm.es, http://www.clip.dia.fi.upm.es/, The CLIP Group, Facultad de Informática, Universidad Politécnica de Madrid.

Version: 1.10#7 (2006/4/26, 19:22:13 CEST)

Version of last change: 1.9#113 (2003/11/27, 20:56:6 CET)

The purpose of this library is to implement an instance of the generic concept of persistent predicates, where external relational databases are used for storage (see the documentation of the persdb library and [CHGT98,Par97] for details). To this end, this library exports SQL persistent versions of the assertz_fact/1, retract_fact/1 and retractall_fact/1 builtin predicates. Persistent predicates also allow concurrent updates from several programs, since each update is atomic.

The notion of persistence provides a very natural and transparent way to access database relations from a Prolog program. Stub definitions are provided for such predicates which access the database when the predicate is called (using the db_client library). A Prolog to SQL translator is used to generate the required SQL code dynamically (see library pl2sql).

This library also provides facilities for reflecting more complex views of the database relations as Prolog predicates. Such views can be constructed as conjunctions, disjunctions, projections, etc. of database relations. Also, SQL-like aggregation operations are supported.

• Implementation of the Database Interface
• Example(s)
• Usage and interface (persdbrt_mysql)
• Documentation on exports (persdbrt_mysql)
• Documentation on multifiles (persdbrt_mysql)
• Documentation on internals (persdbrt_mysql)
• Known bugs and planned improvements (persdbrt_mysql)

Implementation of the Database Interface

The architecture of the low-level implementation of the database interface was defined with two goals in mind:

• to simplify the communication between the Prolog system and the relational database engines as much as possible, and
• to give as much flexibility as possible to the overall system. This includes simultaneous access to several databases, allowing both the databases and clients to reside on the same physical machine or different machines, and allowing the clients to reside in Win95/NT or Unix machines.

In order to allow the flexibility mentioned above, a client-sever architecture was chosen. At the server side, a MySQL server connects to the databases using the MySQL. At the client side, a MySQL client interface connects to this server. The server daemon (mysqld) should be running in the server machine; check your MySQL documentation on how to do that.

After the connection is established a client can send commands to the mediator server which will pass them to the corresponding database server, and then the data will traverse in the opposite direction. These messages include logging on and off from the database, sending SQL queries, and receiving the responses.

The low level implementation of the current library is accomplished by providing abstraction levels over the MySQL interface library. These layers of abstraction implement the persistent predicate view, build the appropriate commands for the database using a translator of Prolog goals to SQL commands, issue such commands using the mediator send/receive procedures, parse the responses, and present such responses to the Prolog engine via backtracking.

Example(s)

Usage and interface (persdbrt_mysql)

Documentation on exports (persdbrt_mysql)

PREDICATE: init_sql_persdb/0:

Usage:

• Description: Internal predicate, used to transform predicates statically declared as persistent (see sql_persistent/3) into real persistent predicates.

PREDICATE: dbassertz_fact/1:

Usage: dbassertz_fact(+Fact)

• Description: Persistent extension of assertz_fact/1: the current instance of Fact is interpreted as a fact (i.e., a relation tuple) and is added to the end of the definition of the corresponding predicate. If any integrity constraint violation is done (database stored predicates), an error will be displayed. The predicate concerned must be statically ( sql_persistent/3) or dinamically ( make_sql_persistent/3) declared. Any uninstantiated variables in the Fact will be replaced by new, private variables. Note: assertion of facts with uninstantiated variables not implemented at this time.
• Call and exit should be compatible with: +Fact is a fact (a term whose main functor is not ':-'/2). (persdbrt_mysql:fact/1)

PREDICATE: dbretract_fact/1:

Usage: dbretract_fact(+Fact)

• Description: Persistent extension of retract_fact/1: deletes on backtracking all the facts which unify with Fact. The predicate concerned must be statically ( sql_persistent/3) or dinamically ( make_sql_persistent/3) declared.
• Call and exit should be compatible with: +Fact is a fact (a term whose main functor is not ':-'/2). (persdbrt_mysql:fact/1)

PREDICATE: dbcurrent_fact/1:

Usage: dbcurrent_fact(+Fact)

• Description: Persistent extension of current_fact/1: the fact Fact exists in the current database. The predicate concerned must be declared sql_persistent/3. Provides on backtracking all the facts (tuples) which unify with Fact.
• Call and exit should be compatible with: +Fact is a fact (a term whose main functor is not ':-'/2). (persdbrt_mysql:fact/1)

PREDICATE: dbretractall_fact/1:

Usage: dbretractall_fact(+Fact)

• Description: Persistent extension of retractall_fact/1: when called deletes all the facts which unify with Fact. The predicate concerned must be statically ( sql_persistent/3) or dinamically ( make_sql_persistent/3) declared.
• Call and exit should be compatible with: +Fact is a fact (a term whose main functor is not ':-'/2). (persdbrt_mysql:fact/1)

PREDICATE: make_sql_persistent/3:

Meta-predicate with arguments: make_sql_persistent(addmodule,?,?).

Usage: make_sql_persistent(PrologPredTypes, TableAttributes, Keyword)

• Description: Dynamic version of the sql_persistent/3 declaration.
• The following properties should hold upon exit: PrologPredTypes is a structure describing a Prolog predicate name with its types. (persdbrt_mysql:prologPredTypes/1) TableAttributes is a structure describing a table name and some attributes. (persdbrt_mysql:tableAttributes/1) Keyword is the name of a persistent storage location. (persdbrt_mysql:persLocId/1)

PREDICATE: dbfindall/4:

Meta-predicate with arguments: dbfindall(?,?,goal,?).

Usage: dbfindall(+DBId, +Pattern, +ComplexGoal, -Results)

• Description: Similar to findall/3, but Goal is executed in database DBId. Certain restrictions and extensions apply to both Pattern and ComplexGoal stemming from the Prolog to SQL translation involved (see the corresponding type definitions for details).
• Call and exit should be compatible with: +DBId a unique identifier of a database session connection. (mysql_client:dbconnection/1) +Pattern is a database projection term. (pl2sql:projterm/1) +ComplexGoal is a database query goal. (pl2sql:querybody/1) -Results is a list. (basic_props:list/1)

PREDICATE: dbcall/2:

Usage: dbcall(+DBId, +ComplexGoal)

• Description: Internal predicate, used by the transformed versions of the persistent predicates. Not meant to be called directly by users. It is exported by the library so that it can be used by the transformed versions of the persistent predicates in the modules in which they reside. Sends ComplexGoal to database DBId for evaluation. ComplexGoal must be a call to a persistent predicate which resides in database DBId.
• Call and exit should be compatible with: +DBId a unique identifier of a database session connection. (mysql_client:dbconnection/1) +ComplexGoal is a database query goal. (pl2sql:querybody/1)

PREDICATE: sql_query/3:

Usage: sql_query(+DBId, +SQLString, AnswerTableTerm)

• Description: ResultTerm is the response from database DBId to the SQL query in SQLString to database DBId. AnswerTableTerm can express a set of tuples, an error answer or a 'ok' response (see answertableterm/1 for details). At the moment, sql_query/3 log in and out for each query. This should be changed to log in only the first time and log out on exit and/or via a timer in the standard way.
• Call and exit should be compatible with: +DBId a unique identifier of a database session connection. (mysql_client:dbconnection/1) +SQLString is a string containing SQL code. (pl2sql:sqlstring/1) AnswerTableTerm is a response from the ODBC database interface. (persdbrt_mysql:answertableterm/1)

PREDICATE: sql_get_tables/2:

Usage 1: sql_get_tables(+Location, -Tables)

• Description: Tables contains the tables available in Location.
• Call and exit should be compatible with: +Location is a structure describing a database. (persdbrt_mysql:database_desc/1) -Tables is a list of atms. (basic_props:list/2)

Usage 2: sql_get_tables(+DbConnection, -Tables)

• Description: Tables contains the tables available in DbConnection.
• Call and exit should be compatible with: +DbConnection a unique identifier of a database session connection. (mysql_client:dbconnection/1) -Tables is a list of atms. (basic_props:list/2)

PREDICATE: sql_table_types/3:

Usage 1: sql_table_types(+Location, +Table, -AttrTypes)

• Description: AttrTypes are the attributes and types of Table in Location.
• Call and exit should be compatible with: +Location is a structure describing a database. (persdbrt_mysql:database_desc/1) +Table is an atom. (basic_props:atm/1) -AttrTypes is a list. (basic_props:list/1)

Usage 2: sql_table_types(+DbConnection, +Table, -AttrTypes)

• Description: AttrTypes are the attributes and types of Table in DbConnection.
• Call and exit should be compatible with: +DbConnection a unique identifier of a database session connection. (mysql_client:dbconnection/1) +Table is an atom. (basic_props:atm/1) -AttrTypes is a list. (basic_props:list/1)

REGTYPE: socketname/1:

Usage: socketname(IPP)

• Description: IPP is a structure describing a complete TCP/IP port address.

REGTYPE: dbname/1:

Usage: dbname(DBId)

• Description: DBId is the identifier of an database.

REGTYPE: user/1:

Usage: user(User)

• Description: User is a user name in the database.

REGTYPE: passwd/1:

Usage: passwd(Passwd)

• Description: Passwd is the password for the user name in the database.

REGTYPE: projterm/1:

Usage: projterm(DBProjTerm)

• Description: DBProjTerm is a database projection term.

REGTYPE: querybody/1:

Usage: querybody(DBGoal)

• Description: DBGoal is a database query goal.

(UNDOC_REEXPORT): sqltype/1:

Imported from sqltypes (see the corresponding documentation for details).

Documentation on multifiles (persdbrt_mysql)

PREDICATE: sql_persistent_location/2:

Relates names of locations (the Keywords) with descriptions of such locations (Locations).

The predicate is multifile.

The predicate is of type dynamic.

Usage: sql_persistent_location(Keyword, DBLocation)

• Description: In this usage, DBLocation is a relational database, in which case the predicate is stored as tuples in the database.
• The following properties should hold upon exit: Keyword is the name of a persistent storage location. (persdbrt_mysql:persLocId/1) DBLocation is a structure describing a database. (persdbrt_mysql:database_desc/1)

Documentation on internals (persdbrt_mysql)

REGTYPE: tuple/1:

tuple(T) :-
        list(T,atm).
tuple(T) :-
        list(T,atm).

Usage: tuple(T)

• Description: T is a tuple of values from the ODBC database interface.

REGTYPE: dbconnection/1:

Usage: dbconnection(H)

• Description: H a unique identifier of a database session connection.

DECLARATION: sql_persistent/3:

Usage: :- sql_persistent(PrologPredTypes, TableAttributes, Keyword).

• Description: Declares the predicate corresponding to the main functor of PrologPredTypes as SQL persistent. Keyword is the name of a location where the persistent storage for the predicate is kept, which in this case must be an external relational database. The description of this database is given through the sql_persistent_location predicate, which must contain a fact in which the first argument unifies with Keyword. TableAttributes provides the table name and attributes in the database corresponding respectively to the predicate name and arguments of the (virtual) Prolog predicate. Although a predicate may be persistent, other usual clauses can be defined in the source code. When querying a persistent predicate with non-persistent clauses, persistent and non-persisten clauses will be evaluated in turn; the order of evaluation is the usual Prolog order, considering that persistent clauses are defined in the program point where the sql_persistent/3 declaration is. Example:

  :- sql_persistent(product( integer,    integer, string, string ),
                product( quantity,   id,      name,   size   ),
                radiowebdb).
  
  sql_persistent_location(radiowebdb,
     db('SQL Anywhere 5.0 Sample', user, pass,
        'r2d5.dia.fi.upm.es':2020)).
  

• The following properties should hold upon exit: PrologPredTypes is a structure describing a Prolog predicate name with its types. (persdbrt_mysql:prologPredTypes/1) TableAttributes is a structure describing a table name and some attributes. (persdbrt_mysql:tableAttributes/1) Keyword is the name of a persistent storage location. (persdbrt_mysql:persLocId/1)

PREDICATE: db_query/4:

Usage: db_query(+DBId, +ProjTerm, +Goal, ResultTerm)

• Description: ResultTerm contains all the tuples which are the response from database DBId to the Prolog query Goal, projected onto ProjTerm. Uses pl2sqlstring/3 for the Prolog to SQL translation and sql_query/3 for posing the actual query.
• Call and exit should be compatible with: +DBId a unique identifier of a database session connection. (mysql_client:dbconnection/1) +ProjTerm is a database projection term. (pl2sql:projterm/1) +Goal is a database query goal. (pl2sql:querybody/1) ResultTerm is a tuple of values from the ODBC database interface. (persdbrt_mysql:tuple/1)

PREDICATE: db_query_one_tuple/4:

Usage: db_query_one_tuple(+DBId, +ProjTerm, +Goal, ResultTerm)

• Description: ResultTerm is one of the tuples which are the response from database DBId to the Prolog query Goal, projected onto ProjTerm. Uses pl2sqlstring/3 for the Prolog to SQL translation and sql_query_one_tuple/3 for posing the actual query. After last tuple has been reached, a null tuple is unified with ResultTerm, and the connection to the database finishes.
• Call and exit should be compatible with: +DBId a unique identifier of a database session connection. (mysql_client:dbconnection/1) +ProjTerm is a database projection term. (pl2sql:projterm/1) +Goal is a database query goal. (pl2sql:querybody/1) ResultTerm is a predicate containing a tuple. (persdbrt_mysql:answertupleterm/1)

PREDICATE: sql_query_one_tuple/3:

Usage: sql_query_one_tuple(+DBId, +SQLString, ResultTuple)

• Description: ResultTuple contains an element from the set of tuples which represents the response in DBId to the SQL query SQLString. If the connection is kept, succesive calls return consecutive tuples, until the last tuple is reached. Then a null tuple is unified with ResultTuple and the connection is finished (calls to mysql_disconnect/1).
• Call and exit should be compatible with: +DBId a unique identifier of a database session connection. (mysql_client:dbconnection/1) +SQLString is a string containing SQL code. (pl2sql:sqlstring/1) ResultTuple is a tuple of values from the ODBC database interface. (persdbrt_mysql:tuple/1)

Known bugs and planned improvements (persdbrt_mysql)

• At least in the shell, reloading a file after changing the definition of a persistent predicate does not eliminate the old definition...
• Functionality missing: some questions need to be debugged.
• Warning: still using kludgey string2term and still using some non-uniquified temp files.
• Needs to be unified with the file-based library.

Go to the first, previous, next, last section, table of contents.