r2108 + introduce $any_op autotag
[racktables] / inc / code.php
index ca3ccd65ccefbac7a5bad2c443b6b1a3f2498f3a..867bc66eb23f4dcb8f5a30bfd0631dfa839682bb 100644 (file)
@@ -1,6 +1,6 @@
 <?php
 /*
- * This file implement lexical scanner and syntax analyzer for the RackCode
+ * This file implements lexical scanner and syntax analyzer for the RackCode
  * access configuration language.
  *
  * The language consists of the following lexems:
  * LEX_PREDICATE
  * LEX_BOOLOP
  *
- * Comments last from the first # to the end of the line and are filtered out
- * automatically.
  */
 
 // Complain about martian char.
-function abortLex1 ($state, $text, $pos)
+function lexError1 ($state, $text, $pos)
 {
-       echo "Error! Could not parse the following text (current state is '${state}'): ";
-       echo substr ($text, 0, $pos);
-       echo '<font color = red>-&gt;' . $text{$pos} . '&lt;-</font>';
-       echo substr ($text, $pos + 1);
-       die;
+       $message = "invalid char with code " . ord (mb_substr ($text, $pos, 1)) . ' (';
+       $message .= mb_substr ($text, $pos, 1) . ") at position ${pos} (FSM state is '${state}')";
+       return array ('result' => 'NAK', 'load' => $message);
 }
 
 // Complain about martian keyword.
-function abortLex2 ($state, $word)
+function lexError2 ($word)
 {
-       echo "Error! Could not parse word (current state is '${state}'): '${word}'.";
-       die;
+       return array
+       (
+               'result' => 'NAK',
+               'load' => "could not parse keyword '${word}'"
+       );
 }
 
 // Complain about wrong FSM state.
-function abortLex3 ($state)
+function lexError3 ($state)
 {
-       echo "Error! Lexical scanner final state is still '${state}' after scanning the last char.";
-       die;
+       return array
+       (
+               'result' => 'NAK',
+               'load' => "FSM state is still '${state}' at end of input"
+       );
 }
 
-function abortSynt ($lexname)
+function lexError4 ($s)
 {
-       echo "Error! Unknown lexeme '${lexname}'.";
-       die;
+       return array
+       (
+               'result' => 'NAK',
+               'load' => "Invalid tag or predicate name '${s}'"
+       );
 }
 
 // Produce a list of lexems from the given text. Possible lexems are:
 function getLexemsFromRackCode ($text)
 {
        $ret = array();
-       $textlen = strlen ($text);
+       $textlen = mb_strlen ($text);
        $state = "ESOTSM";
        for ($i = 0; $i < $textlen; $i++) :
-               $char = $text{$i};
+               $char = mb_substr ($text, $i, 1);
                $newstate = $state;
                switch ($state) :
                        case 'ESOTSM':
@@ -71,11 +76,11 @@ function getLexemsFromRackCode ($text)
                                        case ($char == '#'):
                                                $newstate = 'skipping comment';
                                                break;
-                                       case (preg_match ('/^[a-zA-Z]$/', $char)):
-                                               $newstate = 'reading word';
+                                       case (mb_ereg ('[[:alpha:]]', $char) > 0):
+                                               $newstate = 'reading keyword';
                                                $buffer = $char;
                                                break;
-                                       case (preg_match ('/^[ \t\n]$/', $char)):
+                                       case (preg_match ('/^[ \t\n\r]$/', $char)):
                                                // nom-nom...
                                                break;
                                        case ($char == '{'):
@@ -85,13 +90,13 @@ function getLexemsFromRackCode ($text)
                                                $newstate = 'reading predicate 1';
                                                break;
                                        default:
-                                               abortLex1 ($state, $text, $i);
+                                               return lexError1 ($state, $text, $i);
                                }
                                break;
-                       case 'reading word':
+                       case 'reading keyword':
                                switch (TRUE)
                                {
-                                       case (preg_match ('/^[a-zA-Z]$/', $char)):
+                                       case (mb_ereg ('[[:alpha:]]', $char) > 0):
                                                $buffer .= $char;
                                                break;
                                        case (preg_match ('/^[ \t\n]$/', $char)):
@@ -117,72 +122,78 @@ function getLexemsFromRackCode ($text)
                                                                $ret[] = array ('type' => 'LEX_BOOLCONST', 'load' => $buffer);
                                                                break;
                                                        default:
-                                                               abortLex2 ($state, $buffer);
+                                                               return lexError2 ($buffer);
                                                }
                                                $newstate = 'ESOTSM';
                                                break;
                                        default:
-                                               abortLex1 ($state, $text, $i);
+                                               return lexError1 ($state, $text, $i);
                                }
                                break;
                        case 'reading tag 1':
                                switch (TRUE)
                                {
-                                       case (preg_match ('/^[ \t\n]$/', $char)):
+                                       case (preg_match ('/^[ \t\n\r]$/', $char)):
                                                // nom-nom...
                                                break;
-                                       case (preg_match ('/^[a-zA-Z\$]$/', $char)):
+                                       case (mb_ereg ('[[:alnum:]\$]', $char) > 0):
                                                $buffer = $char;
                                                $newstate = 'reading tag 2';
                                                break;
                                        default:
-                                               abortLex1 ($state, $text, $i);
+                                               return lexError1 ($state, $text, $i);
                                }
                                break;
                        case 'reading tag 2':
                                switch (TRUE)
                                {
                                        case ($char == '}'):
-                                               $ret[] = array ('type' => 'LEX_TAG', 'load' => rtrim ($buffer));
+                                               $buffer = rtrim ($buffer);
+                                               if (!validTagName ($buffer, TRUE))
+                                                       return lexError4 ($buffer);
+                                               $ret[] = array ('type' => 'LEX_TAG', 'load' => $buffer);
                                                $newstate = 'ESOTSM';
                                                break;
-                                       case (preg_match ('/^[a-zA-Z0-9 _-]$/', $char)):
+                                       case (mb_ereg ('[[:alnum:]\. _~-]', $char) > 0):
                                                $buffer .= $char;
                                                break;
                                        default:
-                                               abortLex1 ($state, $text, $i);
+                                               return lexError1 ($state, $text, $i);
                                }
                                break;
                        case 'reading predicate 1':
                                switch (TRUE)
                                {
-                                       case (preg_match ('/^[ \t\n]$/', $char)):
+                                       case (preg_match ('/^[ \t\n\r]$/', $char)):
                                                // nom-nom...
                                                break;
-                                       case (preg_match ('/^[a-zA-Z]$/', $char)):
+                                       case (mb_ereg ('[[:alnum:]]', $char) > 0):
                                                $buffer = $char;
                                                $newstate = 'reading predicate 2';
                                                break;
                                        default:
-                                               abortLex1 ($state, $text, $i);
+                                               return lexError1 ($state, $text, $i);
                                }
                                break;
                        case 'reading predicate 2':
                                switch (TRUE)
                                {
                                        case ($char == ']'):
-                                               $ret[] = array ('type' => 'LEX_PREDICATE', 'load' => rtrim ($buffer));
+                                               $buffer = rtrim ($buffer);
+                                               if (!validTagName ($buffer))
+                                                       return lexError4 ($buffer);
+                                               $ret[] = array ('type' => 'LEX_PREDICATE', 'load' => $buffer);
                                                $newstate = 'ESOTSM';
                                                break;
-                                       case (preg_match ('/^[a-zA-Z0-9 _-]$/', $char)):
+                                       case (mb_ereg ('[[:alnum:]\. _~-]', $char) > 0):
                                                $buffer .= $char;
                                                break;
                                        default:
-                                               abortLex1 ($state, $text, $i);
+                                               return lexError1 ($state, $text, $i);
                                }
                                break;
                        case 'skipping comment':
-                               switch ($text{$i})
+                               switch ($char)
                                {
                                        case "\n":
                                                $newstate = 'ESOTSM';
@@ -195,42 +206,25 @@ function getLexemsFromRackCode ($text)
                endswitch;
                $state = $newstate;
        endfor;
-       if ($state != 'ESOTSM')
-               abortLex3 ($state);
-       return $ret;
-}
-
-function syntReduce_BOOLCONST (&$stack)
-{
-       if (count ($stack) < 1)
-               return FALSE;
-       $top = array_pop ($stack);
-       if ($top['type'] == 'LEX_BOOLCONST')
-       {
-               $s = array
-               (
-                       'type' => 'SYNT_EXPRESSION',
-                       'kids' => array ($top)
-               );
-               array_push ($stack, $s);
-               return TRUE;
-       }
-       // No luck, push it back.
-       array_push ($stack, $top);
-       return FALSE;
+       if ($state != 'ESOTSM' and $state != 'skipping comment')
+               return lexError3 ($state);
+       return array ('result' => 'ACK', 'load' => $ret);
 }
 
 // Parse the given lexems stream into a list of RackCode sentences. Each such
-// sentence is a syntax tree, suitable for tag sequence evaluation. It will
-// contain all of the input lexems framed into a parse tree built from the
-// following nodes:
-// SYNT_NOT (1 argument, SYNT_EXPR)
-// SYNT_BOOLOP (2 arguments of type SYNT_EXPR)
-// SYNT_EXPR (1 argument, different types)
-// SYNT_DEF (2 arguments: subject and definition)
+// sentence is a syntax tree, suitable for tag sequence evaluation. The final
+// parse tree may contain the following nodes:
+// LEX_TAG
+// LEX_PREDICATE
+// LEX_BOOLCONST
+// SYNT_NOT (1 argument, holding SYNT_EXPR)
+// SYNT_BOOLOP (2 arguments, each holding SYNT_EXPR)
+// SYNT_DEFINITION (2 arguments: term and definition)
 // SYNT_GRANT (2 arguments: decision and condition)
-// SYNT_CODESENTENCE (either a grant or a definition)
 // SYNT_CODETEXT (sequence of sentences)
+//
+// After parsing the input successfully a list of SYNT_GRANT and SYNT_DEFINITION
+// trees is returned.
 function getSentencesFromLexems ($lexems)
 {
        $stack = array(); // subject to array_push() and array_pop()
@@ -259,7 +253,6 @@ function getSentencesFromLexems ($lexems)
                        )
                        {
                                // shift!
-                               echo 'shi[f]t!]';
                                array_push ($stack, $stacktop);
                                array_push ($stack, $lexems[$done++]);
                                continue;
@@ -338,7 +331,7 @@ function getSentencesFromLexems ($lexems)
                                                'load' => array
                                                (
                                                        'type' => 'SYNT_NOTEXPR',
-                                                       'arg' => $stacktop
+                                                       'load' => $stacktop['load']
                                                )
                                        )
                                );
@@ -383,8 +376,8 @@ function getSentencesFromLexems ($lexems)
                                                (
                                                        'type' => 'SYNT_BOOLOP',
                                                        'subtype' => $stacksecondtop['load'],
-                                                       'left' => $stackthirdtop,
-                                                       'right' => $stacktop
+                                                       'left' => $stackthirdtop['load'],
+                                                       'right' => $stacktop['load']
                                                )
                                        )
                                );
@@ -406,7 +399,7 @@ function getSentencesFromLexems ($lexems)
                                        (
                                                'type' => 'SYNT_GRANT',
                                                'decision' => $stacksecondtop['load'],
-                                               'condition' => $stacktop
+                                               'condition' => $stacktop['load']
                                        )
                                );
                                continue;
@@ -426,34 +419,15 @@ function getSentencesFromLexems ($lexems)
                                        array
                                        (
                                                'type' => 'SYNT_DEFINITION',
-                                               'term' => $stacksecondtop,
-                                               'definition' => $stacktop
+                                               'term' => $stacksecondtop['load'],
+                                               'definition' => $stacktop['load']
                                        )
                                );
                                continue;
                        }
                        if
                        (
-                               $stacktop['type'] == 'SYNT_GRANT' or
-                               $stacktop['type'] == 'SYNT_DEFINITION'
-                       )
-                       {
-                               // reduce!
-                               array_pop ($stack);
-                               array_push
-                               (
-                                       $stack,
-                                       array
-                                       (
-                                               'type' => 'SYNT_CODESENTENCE',
-                                               'load' => $stacktop
-                                       )
-                               );
-                               continue;
-                       }
-                       if
-                       (
-                               $stacktop['type'] == 'SYNT_CODESENTENCE' and
+                               ($stacktop['type'] == 'SYNT_GRANT' or $stacktop['type'] == 'SYNT_DEFINITION') and
                                $stacksecondtop['type'] == 'SYNT_CODETEXT'
                        )
                        {
@@ -470,12 +444,12 @@ function getSentencesFromLexems ($lexems)
                        }
                        if
                        (
-                               $stacktop['type'] == 'SYNT_CODESENTENCE'
+                               $stacktop['type'] == 'SYNT_GRANT' or
+                               $stacktop['type'] == 'SYNT_DEFINITION'
                        )
                        {
                                // reduce!
                                array_pop ($stack);
-                               $stacksecondtop['log'][] = $stacktop;
                                array_push
                                (
                                        $stack,
@@ -500,9 +474,166 @@ function getSentencesFromLexems ($lexems)
                }
                // The moment of truth.
                if (count ($stack) == 1 and $stack[0]['type'] == 'SYNT_CODETEXT')
-                       return $stack[0];
-               return NULL;
+                       return array ('result' => 'ACK', 'load' => $stack[0]['load']);
+               return array ('result' => 'NAK', 'load' => 'Syntax error!');
+       }
+}
+
+function eval_expression ($expr, $tagchain, $ptable)
+{
+       switch ($expr['type'])
+       {
+               case 'LEX_TAG': // Return true, if given tag is present on the tag chain.
+                       foreach ($tagchain as $tagInfo)
+                               if ($expr['load'] == $tagInfo['tag'])
+                                       return TRUE;
+                       return FALSE;
+               case 'LEX_PREDICATE': // Find given predicate in the symbol table and evaluate it.
+                       $pname = $expr['load'];
+                       if (!isset ($ptable[$pname]))
+                       {
+                               showError ("Predicate '${pname}' is referenced before declaration", __FUNCTION__);
+                               return;
+                       }
+                       return eval_expression ($ptable[$pname], $tagchain, $ptable);
+               case 'LEX_BOOLCONST': // Evaluate a boolean constant.
+                       switch ($expr['load'])
+                       {
+                               case 'true':
+                                       return TRUE;
+                               case 'false':
+                                       return FALSE;
+                               default:
+                                       showError ("Could not parse a boolean constant with value '${expr['load']}'", __FUNCTION__);
+                                       return; // should failure be harder?
+                       }
+               case 'SYNT_NOTEXPR':
+                       return !eval_expression ($expr['load'], $tagchain, $ptable);
+               case 'SYNT_BOOLOP':
+                       $leftresult = eval_expression ($expr['left'], $tagchain, $ptable);
+                       switch ($expr['subtype'])
+                       {
+                               case 'or':
+                                       if ($leftresult)
+                                               return TRUE; // early success
+                                       return eval_expression ($expr['right'], $tagchain, $ptable);
+                               case 'and':
+                                       if (!$leftresult)
+                                               return FALSE; // early failure
+                                       return eval_expression ($expr['right'], $tagchain, $ptable);
+                               default:
+                                       showError ("Cannot evaluate unknown boolean operation '${boolop['subtype']}'");
+                                       return;
+                       }
+               default:
+                       showError ("Evaluation error, cannot process expression type '${expr['type']}'", __FUNCTION__);
+                       break;
+       }
+}
+
+function gotClearanceForTagChain ($tagchain)
+{
+       global $rackCode;
+       $ptable = array();
+       foreach ($rackCode as $sentence)
+       {
+               switch ($sentence['type'])
+               {
+                       case 'SYNT_DEFINITION':
+                               $ptable[$sentence['term']] = $sentence['definition'];
+                               break;
+                       case 'SYNT_GRANT':
+                               if (eval_expression ($sentence['condition'], $tagchain, $ptable))
+                                       switch ($sentence['decision'])
+                                       {
+                                               case 'allow':
+                                                       return TRUE;
+                                               case 'deny':
+                                                       return FALSE;
+                                               default:
+                                                       showError ("Condition match for unknown grant decision '${sentence['decision']}'", __FUNCTION__);
+                                                       break;
+                                       }
+                               break;
+                       default:
+                               showError ("Can't process sentence of unknown type '${sentence['type']}'", __FUNCTION__);
+                               break;
+               }
        }
+       return FALSE;
+}
+
+function getRackCode ($text)
+{
+       if (!mb_strlen ($text))
+               return array ('result' => 'NAK', 'load' => 'The RackCode text was found empty in ' . __FUNCTION__);
+       $text = str_replace ("\r", '', $text) . "\n";
+       $lex = getLexemsFromRackCode ($text);
+       if ($lex['result'] != 'ACK')
+               return $lex;
+       $synt = getSentencesFromLexems ($lex['load']);
+       if ($synt['result'] != 'ACK')
+               return $synt;
+       // An empty sentence list is semantically valid, yet senseless,
+       // so checking intermediate result once more won't hurt.
+       if (!count ($synt['load']))
+               return array ('result' => 'NAK', 'load' => 'Empty parse tree found in ' . __FUNCTION__);
+       return semanticFilter ($synt['load']);
+}
+
+// Return NULL, if the given expression can be evaluated against the given
+// predicate list. Return the name of the first show stopper otherwise.
+function firstUnrefPredicate ($plist, $expr)
+{
+       switch ($expr['type'])
+       {
+               case 'LEX_BOOLCONST':
+               case 'LEX_TAG':
+                       return NULL;
+               case 'LEX_PREDICATE':
+                       return in_array ($expr['load'], $plist) ? NULL : $expr['load'];
+               case 'SYNT_NOTEXPR':
+                       return firstUnrefPredicate ($plist, $expr['load']);
+               case 'SYNT_BOOLOP':
+                       if (($tmp = firstUnrefPredicate ($plist, $expr['left'])) !== NULL)
+                               return $tmp;
+                       if (($tmp = firstUnrefPredicate ($plist, $expr['right'])) !== NULL)
+                               return $tmp;
+                       return NULL;
+               default:
+                       return NULL;
+       }
+}
+
+function semanticFilter ($code)
+{
+       $predicatelist = array();
+       foreach ($code as $sentence)
+               switch ($sentence['type'])
+               {
+                       case 'SYNT_DEFINITION':
+                               $up = firstUnrefPredicate ($predicatelist, $sentence['definition']);
+                               if ($up !== NULL)
+                                       return array
+                                       (
+                                               'result' => 'NAK',
+                                               'load' => "definition [${sentence['term']}] uses unknown predicate [${up}]"
+                                       );
+                               $predicatelist[] = $sentence['term'];
+                               break;
+                       case 'SYNT_GRANT':
+                               $up = firstUnrefPredicate ($predicatelist, $sentence['condition']);
+                               if ($up !== NULL)
+                                       return array
+                                       (
+                                               'result' => 'NAK',
+                                               'load' => "grant sentence uses unknown predicate [${up}]"
+                                       );
+                               break;
+                       default:
+                               return array ('result' => 'NAK', 'load' => 'unknown sentence type');
+               }
+       return array ('result' => 'ACK', 'load' => $code);
 }
 
 ?>