====== MAGIK 3 Objectives ======

**This document describes the objectives and (un)implemented tasks (tickets) of the new version of MAGIK tool.** \\

**Creation Date: 2013/05/15.**

**Last Modified:  ~~LASTMOD~~.**

===== Previous MAGIK versions: =====

  * ''MAGIK'' - initial test version.
  * ''MAGIK_WEB/'' - a version presented in November 2012 at CIKM. 
  * ''MAGIK_WEB_AFTER_CIKM/'' - a version developed after November 2012 at CIKM. Novelties:
       * Users
       * Hibernate
       * ...
  * ''MAGIK_WEB_VLDB/'' - a version prepared for the query approximation version of MAGIK
       * no users (guest is a default user) 
       * analytic page
       * several query checked at once for completeness


===== Objectives (2013/05/16) =====
 === Create a top controlling module ===
  * Keeps track about running sessions 
  * Keeps track about opened database connections


 === Implement reasoning that for a given incomplete query ===

       * Generates specializations of max size ''k'' 
       * Generates generalization
     
 === Integrate VLDB and CIKM version ===
       * Integrate completeness statements suggestions and query approximation 
           * Reasoning can be done with one query at time? 

 === Database ===

  * MAGIK 3 connects to database:  ''jdbc:postgresql:/''''/magik-demo.inf.unibz.it:5432/magik3'' (user and pass are as before)
  * Information about schemas and users are kept in the schema named: ''meta_information''

== Schema ==
<fc #FF0000>If we use hibernate then we should save schema objects only!</fc>
<fc #FF0000>Extend schema class with type,dburl,dbusername,dbpassword</fc>
 (hibernate tutorial: http://www.tutorialspoint.com/hibernate/hibernate_overview.htm)

       * Schema table contains information about virtual and remote database schemas 
             * ''schema(schemaname,username,type,dburl,dbusername,dbpassword)''
             * Here ''type in {local,remote}'' 
             * Primary key is ''schema[schemaname,username,type]''. It means that there cannot exists two schemas with the same name and type 
(provided that for a moment we assume a single user in the system ''guest''.
       * Create one table for each FK, FDC, TC, and Query. 
             * ''tcstatement(username, schemaname, type, relation, headterms, condition)''. 
                  *  Primary key:  ''tcstatement[username, schemaname, type, relation]''.
             * ''query(username, schemaname, type, queryname, atoms, outputvars, description, sql, semantics, countexists)''. 
                  *  Primary key:  ''tcstatement[username, schemaname, type, queryname]''.
             * ''foreignkey(username, schemaname, type, sourcerelation, targetrelation, virtual, enforced)''. 
                  *  Primary key:  ''tcstatement[username, schemaname, type, sourcerelation, targetrelation]''.
             * ''finitedomain(username, schemaname, type, relation, argposition, values)''. 
                  * Primary key:  ''finitedomain[username, schemaname, type, relation, argposition]''.
       * **Virtual** schemas: User is allowed to create new virtual schema and they are stored in schema table
             * **Important**: In virtual mode user **is allowed** to alter FKs and FDCs. So FK are listed and they are enforced! 
             * As before we cannot query the database 
                   * There is no comparisons of query answers in analytic window for generalizations and specializations 
       * **Remote** schemas: User is allowed to create new connection to a remote schema and it should posses enough rights to read schema and meta information. 
           * **Important**: In remote mode user** is not allowed** to alter FKs and FDCs, but you should be able to read it from the Postgres Database. <fc #FF0000>NEW</fc>: For reading FDC we should read only ENUM datatypes in Postgres. 
           * Should we do a check if a remote schema has altered? <fc #0000FF>(think later)</fc>
 
== Users ==

       * For now disable user login (allowing possibility for introducing users later). 
       * We have user table 
      
  === TC-statements ===
       * Constants in TC-statements should be represented with using single-quotes, e.g., ''pupil('John', '1', 'b')''
               * i.e.  instead of convention that //constant term starts with a lower-case letter// it should be //constant term are between single quotes//
               * that aslo means that constants can start with upper-case. Therefore, when generating the DLV encoding constants needs to encoded under double-quotes. 
               * Estimate if this is worth persuading.

       * Representing TC-statements in a datalog form instead of the current format (ask Werner?).
               * For example, ''pupil_a(Name,Level,Code) <- pupil_i(Name,Level,Code) , learns_i(Name,'French').''