- how to implement dynamically adjusting views based
  on total number of result records??

- sqlite    with_v3=yes
  perl
  perl-dbi  with_dbd_sqlite=yes
  perl-www
  perl-xml
  perl-util
  apache

- apache.conf:
  Alias /quos/ /u/rse/wrk/ossp/quos/
  <Directory /u/hms/wrk/ossp/quos/>
     Options +ExecCGI
  </Directory>

- prepare: 
  - purpose of all files
  - OpenPKG RDF as example
  - perldoc IO::File
  - perldoc CGI
  - perldoc String::Divert
  - perldoc DBI

  - design for HTML form for boolean expression based query!
    (RT, perldoc DBIx::SearchBuilder)

- UI Query Part (boolean expression to HTML form):

  Boolean Expression:
  expr ::= field op value
         | "!" expr
         | expr "||" expr
         | expr "&&" expr
         | "(" expr ")"
  field ::= /^[A-Z]+$/
  op    ::= "==" | "!=" | "<" | "<=" | ">" | ">=" | "=~" | "!~"
  value ::= /^.*$/

  HTML Form:
  ------------------------------
        ( field op value
        & field op value )
  & ( |   field op value
        & field op value 
    |   ! field op value       
      & ! field op value )     
  ------------------------------
  <:NN V:NN UPDATE CLEAR SUBMIT    
  ------------------------------
       NAME LOAD DELETE    
       NAME SAVE HYPERLINK 
  ------------------------------

- modules:

    * transformation module:
        - format 1: 2d form elements (input after submit of web form)
            lo2   lo1   lo0      fn        fo       fv
           +---+ +---+ +---+  +---------+ +---+ +---------+
      0 ...|   | |   | |   |  |  grp    | |== | | Mail    |
           +---+ +---+ +---+  +---------+ +---+ +---------+
           +---+ +---+ +---+  +---------+ +---+ +---------+
      1 ...|   | |   | ||| |  | desc    | |=~ | | \bMTA\b |
           +---+ +---+ +---+  +---------+ +---+ +---------+
           +---+ +---+ +---+  +---------+ +---+ +---------+
      2 ...|   | |&& | |   |  | class   | |== | | CORE    |
           +---+ +---+ +---+  +---------+ +---+ +---------+
           +---+ +---+ +---+  +---------+ +---+ +---------+
      3 ...|   | |   | ||| |  | class   | |== | | BASE    |
           +---+ +---+ +---+  +---------+ +---+ +---------+
             :    :    :         :          :        :
           
          fn-0="grp"&fo-0="=="&fv-0="Mail"&lo0-1="||"&fn-1="desc"&fo-1="=~"&fv-1="\bMTA\b"
          lo1-2="&&"&fn-2="class"&fo-2="=="&fv-2="CORE"&lo0-3="||"&fn-3="class"&fo-3="=="&fv-3="BASE"

        - format 2a: astract syntax tree for query expression (internal representation)
         
          function-style:
          AND( OR( EQ(grp,"Mail"),
                   RE(desc,"\bMTA\b") ),
               OR( EQ(class,"CORE"),
                   EQ(class,"BASE")   ) )

          in-core-style:
          my $expr = 
              [ "AND", [ "OR", [ "EQ", "grp",   "Mail"   ],
                               [ "RE", "desc",  "\bMTA\b"]  ],
                       [ "OR", [ "EQ", "class", "CORE"   ],
                               [ "EQ", "class", "BASE"   ]  ] ];

          "print $expr->[2]->[1]->[2];" -> "CORE"

          my $format2b = &transform_tree($expr, \&transform_style_perl);
          my $format2c = &transform_tree($expr, \&transform_style_sql);
          sub transform_tree {
              my ($expr, $style) = @_;
              my $str = '';
              my ($op, $a1, $a2) = @{$expr};
              $op = &style($op);
              $a1 = ref($a1) ? &style(&transform_tree($a1, $style)) : &style($a1);
              $a2 = ref($a2) ? &style(&transform_tree($a2, $style)) : &style($a2);
              $str = "$a1 $op $a2";
              return $str;
          }
          sub transform_style_perl {
              my ($token) = @_;
              $token =~ s/^AND$/&&/;
              $token =~ s/^OR$/||/;
              $token =~ s/^EQ$/==/;
              $token =~ s/^RE$/=~/;
              return $token;
          }
          sub transform_style_sql {
              my ($token) = @_;
              $token =~ s/^EQ$/=/;
              $token =~ s/^RE$/MATCH/;
              return $token;
          }

        - format 2b: C-style query expression (input after hyperlink in web page)

          (grp == "Mail" || desc =~ "\bMTA\b") && 
          (class == "CORE || class == "BASE")

        - format 2c: SQL-style query expression (output for database query)

          (quos.q_grp = 'Mail' OR quos.q_desc MATCH? '\bMTA\b') AND 
          (quos.q_class = 'CORE' OR quos.q_class = 'BASE')

    * creation of web formular (render query)
        sub_render_form (cgi-Object | expression, row, column

- Parsing CFG:

  OSSP shiela (with tricks)
  OSSP ac (uses Parse::RecDescent)
  Parse::RecDescent (full featured recursive-descend parsing)

  GENERAL:
  config             ::= directive(s)
  directive          ::= word arg ";"
  arg                ::= "{" config "}"
                       | word
  word               ::= m/"[^"]*"/
                       | m/'[^']*'/
                       | m/[^\s+;]+/

  SPECIAL:
  config             ::= directive(s)
  directive          ::= options
                       | queries
                       | views
  options            ::= "option" "{" options-directive(s) "}" ";"
  options-directive  ::= webpage 
                       | database
  webpage            ::= "webpage" word word ";"
  database           ::= "database" "{" database-directive(s) "}" ";"
  database-directive ::= ...
  ...
  queries            ::= group(s)
  ...
  views              ::= view(s)
  ...
  word               ::= m/"[^"]*"/
                       | m/'[^']*'/
                       | m/[^\s+;]+/

- Parsing RDF:

  my $str = ...; # string representation
  my $rdf = {};  # AST    representation

  $str =~ s/<Repository([^>]+)>(.+?)<\/Repository>/&do_repository($rdf, $1, $2)/sge;
  sub do_repository {
      my ($rdf, $attr, $str) = @_;
      my ($name) = ($attr =~ m/rdf:resource="([^"]+)"/s);
      $rdf->{$name} = {};
      $str =~ s/<rdf:Description([^>]+)>(.+?)<\/rdf:Description>/&do_destription($rdf->{$name}, $1, $2)/sge;
      sub do_description {
          my ($rdf, $attr, $str) = @_;
          $rdf->{...} = ...;
      }
  }

  1. geht nur, wenn die Sprache keine rekursiven Elemente enthält, d.h. ein
     Element kann nicht in sich selber (in beliebiger Tiefe) vorkommen
     (Hinweis: Für Compilerbauer uninteressant)

  2. Blockstruktur der Sprache wird direkt Substitution+Funktionsaufruf umgesetzt

  3. Auf jeder Ebene wird die bereits erkannte Information in einer internen
     Datenstruktur aufgesammelt ($rdf)

- Next Steps:

  - openpkg-rdf2sql.pl:
    . rdf to AST parsing (subst+func)  (1)
    . AST to SQL mapping (INSERT...)   (2)

  - quos.cgi:
    . AST to SQL mapping (WHERE....)   (3)
    . view part
    . result part