<?php

/*
 * Ruby's grammar is basically insane. We're not going to aim to correctly
 * highlight all legal Ruby code because we'll be here all year and we'll still
 * get it wrong, but we're going to have a go at getting the standard stuff
 * right as well as:
 *   heredocs
 *   balanced AND NESTED string/regex delimiters
 *   interpolation
 * 
 * disclaimer: I don't actually know Ruby.
 *
 * Problem is that Ruby *appears* to have to disambiguate loads of stuff at
 * runtime, which is frankly a little optimistic for a syntax highlighter.
 * Ruby allows you to omit calling parantheses, so it's not practical (and
 * impossible if the code snippet is incomplete) to figure out operator/operand
 * position. e.g.
 *      x = y %r/z/x
 * is x = y mod r div z div x, unless y is a function, in which case it's:
 *      x = y( /z/x )     where /z/x is a regex
 */


class LuminousRubyScanner extends LuminousScanner {


  // set to true if this is a nested scanner which needs to exit if it
  // encounters a } while nothing else is on the stack, i.e. it is being
  // used to process an interpolated block
  public $interpolation = false;
  protected $curley_braces = 0; // poor man's curly brace stack.

  public $rails = false;

  // operators depend somewhat on whether or not rails is active, else we
  // don't want to consume a '%' if it comes right before a '>', we want
  // to leave that for the rails close-tag detection
  private $operator_regex = null;
  private $string_regex = null;
  private $comment_regex = null;

  // gaaah
  private $numeric = '/
  (?:
    #control codes
    (?:\?(?:\\\[[:alpha:]]-)*[[:alpha:]])
    |
    #hex
    (?:0[xX](?>[0-9A-Fa-f]+)[lL]*)
    |
    # binary
    (?:0[bB][0-1]+)
    |
    #octal
    (?:0[oO0][0-7]+)
    |
    # regular number
    (?:
      (?>[0-9]+)
      (?:
        # fraction
        (?:
          (?:\.?(?>[0-9]+)?
            (?:(?:[eE][\+\-]?)?(?>[0-9]+))?
          )
        )
      )?
    )
    |
    (
      # or only after the point, float x = .1;
      \.(?>[0-9]+)(?:(?:[eE][\+\-]?)?(?>[0-9]+))?
    )
  )
  (?:_+\d+)*
  /x';  

  /// queue of heredoc declarations which will need to be handled as soon as EOL is reached
  /// each element is a tuple: (delimiter(str), identable?, interpolatable?)
  private $heredocs = array();


  public function init() {
    $this->comment_regex =
      $this->rails? "/ \# (?: [^\n%]*+ | %(?!>))* /x"
        : '/#.*/';
    // http://www.zenspider.com/Languages/Ruby/QuickRef.html#23
    $this->operator_regex = '/
      \?   | ;
      | ::? | \*[=\*]? | \/=? | -=? | %=? | ^=? | &&? | \|\|? | \.{2,3}
      | \^=?
      | < (?:=>|<|=)? | >=?
      | =[>~] | ={1,3} 
      | \+=? | ![=~]?
    /x';
//     $this->operator_regex = '/(?: [~!^&*\-+=:;|<>\/?';
//     if ($this->rails) $this->operator_regex .= ']+|%(?!>))+';
//     else $this->operator_regex .= '%]+)';
//     $this->operator_regex .= '/x';
      
    $this->add_identifier_mapping('KEYWORD', array('BEGIN', 'END', 'alias',
      'begin', 'break', 'case', 'class', 'def', 'defined?', 'do',
      'else', 'elsif', 'end', 'ensure', 'for', 'if', 'module', 'next',
      'redo', 'rescue', 'retry', 'return', 'self', 'super', 'then',
      'undef', 'unless', 'until', 'when', 'while', 'yield',
      'false', 'nil', 'self', 'true', '__FILE__', '__LINE__', 'TRUE', 'FALSE',
      'NIL', 'STDIN', 'STDERR', 'ENV', 'ARGF', 'ARGV', 'DATA', 'RUBY_VERSION',
      'RUBY_RELEASE_DATE', 'RUBY_PLATFORM',
      'and', 'in', 'not', 'or',
      'public', 'private', 'protected'
      ));

    // http://www.tutorialspoint.com/ruby/ruby_builtin_functions.htm
    // don't know how reliable that is... doesn't look incredibly inspiring
    $this->add_identifier_mapping('FUNCTION', array('abord', 'Array',
    'at_exit', 'autoload', 'binding', 'block_given?', 'callcc', 'caller',
    'catch', 'chomp', 'chomp!', 'chop', 'chop!', 'eval', 'exec', 'exit',
    'exit!', 'fail', 'Float', 'fork', 'format', 'gets', 'global_variables',
    'gsub', 'gsub!', 'Integer', 'lambda', 'proc', 'load', 'local_variables',
    'loop', 'open', 'p', 'print', 'printf', 'proc', 'puts', 'raise', 'fail',
    'rand', 'readlines', 'require', 'scan', 'select', 'set_trace_func',
    'sleep', 'split', 'sprintf', 'srand', 'String', 'syscall', 'system',
    'sub', ',sub!', 'test', 'throw', 'trace_var', 'trap', 'untrace_var',
    'abs', 'ceil', 'coerce', 'divmod', 'floor', 'integer?', 'modulo',
    'nonzero?', 'remainder', 'round', 'truncate', 'zero?', 'chr', 'size',
    'step', 'times', 'to_f', 'to_int', 'to_i', 'finite?', 'infinite?',
    'nan?', 'atan2', 'cos', 'exp', 'frexp', 'ldexp', 'log', 'log10', 'sin',
    'sqrt', 'tan'));


    // this can break a bit with Ruby's whacky syntax
    $this->remove_filter('pcre');
    // don't want this.
    $this->remove_filter('comment-to-doc');

    $this->add_filter('REGEX', create_function('$tok',
      'return LuminousFilters::pcre($tok, (isset($tok[1][0]) && $tok[1][0] === "/"));'));
  }

  
  protected function is_regex() {
    /*
     * Annoyingly I don't really know exactly what rules Ruby uses for
     * disambiguating regular expressions. There might be some incorrect
     * assumptions in here.
     */
    
    if ($this->check('%/=\s%'))
      return false;
    $following_space = (bool)$this->check("%/[ \t]%");
    $space = false;
    for($i=count($this->tokens)-1; $i>=0; $i--) {
      $tok = $this->tokens[$i];
      if ($tok[0] === 'COMMENT') continue;
      elseif ($tok[0] === 'OPERATOR') return true;
      elseif($tok[0] === 'STRING') return true;
      elseif ($tok[1] === '(' || $tok[1] === ',' || $tok[1] === '{' ||
          $tok[1] === '[') {
        // this is definitely an operand
        return true;
      }
      elseif($tok[0] === null) {
        $space = true;
        continue;
      }
      elseif($tok[0] === 'NUMERIC') {
        // this is definitely an operator
        return false;
      }
      elseif ($tok[0] === 'IDENT'
        || $tok[0] === 'CONSTANT'
        || $tok[0] === 'VALUE' // aka :symbols
      ) {
        // this could be an operator or operand
        // Kate's syntax engine seems to operate on the following basis:
        if ($space && $following_space) return false;
        return $space;
      }
      return false;
    }
    return true; // no preceding tokens, presumably a code fragment.
  }


  protected function interpolate() {
    $interpolation_scanner = new LuminousRubyScanner();
    $interpolation_scanner->string($this->string());
    $interpolation_scanner->pos($this->pos());
    $interpolation_scanner->interpolation = true;
    $interpolation_scanner->init();
    $interpolation_scanner->main();
    $this->record($interpolation_scanner->tagged(), 'INTERPOLATION', true);
    $this->pos($interpolation_scanner->pos());
  }


  // handles the heredoc array. Call at eol/bol when the heredoc queue is
  // not empty
  protected function do_heredoc() {
  
    assert (!empty($this->heredocs));
    
    $start = $this->pos();
    
    for($i=0; $i<count($this->heredocs) ; ) {
      $top = $this->heredocs[$i];
      list($ident, $identable, $interpolatable) = $top;
      $searches = array(
        sprintf('/^%s%s\\b/m', $identable? "[ \t]*" : '',
          preg_quote($ident, '/'))
      );
      if ($interpolatable)
        $searches[] = '/\#\{/';
      list($next, $matches) = $this->get_next($searches);
      if ($next === -1) {
        // no match for end delim, run to EOS
        $this->record(substr($this->string(), $start), 'HEREDOC');
        $this->terminate();
        break;
      }
      assert($matches !== null);
      if ($matches[0] === '#{') { // interpolation, break heredoc and do that.
        $this->pos($next);
        $this->record(substr($this->string(), $start, $this->pos()-$start), 'HEREDOC');
        $this->record($matches[0], 'DELIMITER');
        $this->pos_shift(strlen($matches[0]));
        $this->interpolate();        
        if ($this->peek() === '}')
          $this->record($this->get(), 'DELIMITER');
        $start = $this->pos();
      }
      else {
        // 
        $this->pos($next);
        $this->record(substr($this->string(), $start, $this->pos()-$start), 'HEREDOC');
        $this->record($matches[0], 'DELIMITER');
        $this->pos($next + strlen($matches[0]));
        $start = $this->pos();
        $i++;
      }
      // subscanner might have consumed all the string, in which case there's 
      // no point continuing
      if ($this->eos()) break;
      
    }
    // we may or may not have technically addressed all the heredocs in the
    // queue, but we do want to clear them out now
    $this->heredocs = array();
  }


  private function record_string_range($from, $to, $type, $split) {
    if ($to === $from) return;
    $substr = substr($this->string(), $from, $to-$from);
    if ($split) {
      foreach(preg_split('/(\s+)/', $substr, -1, PREG_SPLIT_DELIM_CAPTURE) as $s) {
        $type_ = preg_match('/^\s+$/', $s)? null : $type;
        $this->record($s, $type_);
      }
    } else {
      $this->record($substr, $type);
    }
  }

  // handles string types (inc regexes), which may have nestable delimiters or
  // interpolation.
  // strdata is defined in the big ugly block in main()
  // TODO: proper docs
  protected function do_string($str_data) {
    list($type, $open_delimiter, $close_delimiter, $pos, $interpolation,
      $fancy_delim, $split) = $str_data;
    $balanced = $open_delimiter !== $close_delimiter;
    $template = '/(?<!\\\\)((?:\\\\\\\\)*)(%s)/';
    $patterns = array();
    $patterns['term'] = sprintf($template, preg_quote($close_delimiter, '/'));
    if ($balanced) {
      // for nesting balanced delims
      $patterns['nest'] = sprintf($template, preg_quote($open_delimiter, '/'));
    }
    if ($interpolation) {
      $patterns['interp'] = sprintf($template, preg_quote('#{', '/'));
    }
    $nesting_level = 0;
    $break = false;
    while (!$break) {
      list($name, $index, $matches) = $this->get_next_named($patterns);
      
      if ($name === null) {
        // special case, no matches, record the rest of the string and break
        // immediately
        $this->record_string_range($pos, strlen($this->string()), $type, $split);
        $this->terminate();
        break;
      }
      elseif ($name === 'nest') {
        // nestable opener
        $nesting_level++;
        $this->pos( $index + strlen($matches[0]) );
      }
      elseif($name === 'term') {
        // terminator, may be nested
        if ($nesting_level === 0) {
          // wasn't nested, real terminator. 
          if ($fancy_delim) {
            // matches[1] is either empty or a sequence of backslashes
            $this->record_string_range($pos, $index+strlen($matches[1]), $type, $split);
            $this->record($matches[2], 'DELIMITER');
          } else {
            $this->record_string_range($pos, $index+strlen($matches[0]), $type, $split);
          }
          $break = true;
        }
        else {
          // pop a nesting level
          $nesting_level--;
        }
        $this->pos( $index + strlen($matches[0]) );

      }
      elseif($name === 'interp') {
        // interpolation - temporarily break string highlighting, then
        // do interpolation, then resume.
        $this->record_string_range($pos, $index + strlen($matches[1]), $type, $split);
        $this->record($matches[2], 'DELIMITER');
        $this->pos( $index + strlen($matches[0]) );
        
        $this->interpolate();
        if (($c = $this->peek()) === '}')
          $this->record($this->get(), 'DELIMITER');
        $pos = $this->pos();
      }
      else {
        assert(0);
      }
      if ($break) break;
    }
    if ($type === 'REGEX' && $this->scan('/[iomx]+/'))
      $this->record($this->match(), 'KEYWORD');
  }
  

  public function main() {

    while (!$this->eos()) {

      if ($this->bol() && !empty($this->heredocs)) {
        $this->do_heredoc();
      }
      
      if ($this->interpolation) {
        $c = $this->peek();
        if ($c === '{') $this->curley_braces++;
        elseif($c === '}') {
          $this->curley_braces--;
          if ($this->curley_braces <= 0) { break;}
        }
      }
      if ($this->rails && $this->check('/-?%>/')) {
        break;
      }

      $c = $this->peek();
      if ($c === '=' && $this->scan('/^=begin .*? (^=end|\\z)/msx')) {
        $this->record($this->match(), 'DOCCOMMENT');
      }
      elseif($c === '#' && $this->scan($this->comment_regex))
        $this->record($this->match(), 'COMMENT');
      
      elseif($this->scan($this->numeric) !== null) {
        $this->record($this->match(), 'NUMERIC');
      }
      elseif( $c === '$' && $this->scan('/\\$
  (?:
    (?:[!@`\'\+1~=\/\\\,;\._0\*\$\?:"&<>])
    |
    (?: -[0adFiIlpvw])
    |
    (?:DEBUG|FILENAME|LOAD_PATH|stderr|stdin|stdout|VERBOSE)
  )/x') || $this->scan('/(\\$|@@?)\w+/')) {
        $this->record($this->match(), 'VARIABLE');
      }
      elseif($this->scan('/:\w+/')) {
        $this->record($this->match(), 'VALUE');
      }
      elseif ( $c === '<' && $this->scan('/(<<(-?))([\'"`]?)([A-Z_]\w*)(\\3)/i')) {
        $m = $this->match_groups();
        $this->record($m[0], 'DELIMITER');
        $hdoc = array($m[4], $m[2] === '-', $m[3] !== "'");
        $this->heredocs[] = $hdoc;
      }

      // TODO: "% hello " is I think a valid string, using whitespace as
      // delimiters. We're going to disallow this for now because
      // we're not disambiguating between that and modulus
      elseif (($c === '"' || $c === "'" || $c === '`' || $c === '%') &&
        $this->scan('/[\'"`]|%( [qQrswWx](?![[:alnum:]]|$) | (?![[:alnum:]\s]|$))/xm')
        || ($c === '/' && $this->is_regex())
        )
      {
        $interpolation = false;
        $type = 'STRING';
        $delimiter;
        $pos;
        $fancy_delim = false;
        $split = false;

        if ($c === '/') {
          $interpolation = true;
          $type = 'REGEX';
          $delimiter = $c;
          $pos = $this->pos();
          $this->get();
        } else {
          $pos = $this->match_pos();
          $delimiter = $this->match();
          if ($delimiter === '"') {
            $interpolation = true;
          } elseif($delimiter === "'") {}
          elseif($delimiter === '`') {
            $type = 'FUNCTION';
          }
          else {
            $delimiter = $this->get();
            $m1 = $this->match_group(1);
            if ($m1 === 'Q' || $m1 === 'r' || $m1 === 'W' || $m1 === 'x')
              $interpolation = true;
            if ($m1 === 'w' || $m1 === 'W')
              $split = true;
            if ($m1 === 'x') $type = 'FUNCTION';
            elseif($m1 === 'r') $type = 'REGEX';

            $fancy_delim = true;
            $this->record($this->match() . $delimiter, 'DELIMITER');
            $pos = $this->pos();
          }
        }
        $data = array($type, $delimiter, LuminousUtils::balance_delimiter($delimiter),
          $pos, $interpolation, $fancy_delim, $split);
        $this->do_string($data);
      }
      elseif( (ctype_alpha($c) || $c === '_') &&
        ($m = $this->scan('/[_a-zA-Z]\w*[!?]?/')) !== null) {          
        $this->record($m, ctype_upper($m[0])? 'CONSTANT' : 'IDENT');
        if ($m === '__END__') {
          if (!$this->interpolation) {
            $this->record($this->rest(), null);
            $this->terminate();
          }
          break;
        }
      }
      elseif($this->scan($this->operator_regex))
        $this->record($this->match(), 'OPERATOR');
      elseif($this->scan("/[ \t]+/")) $this->record($this->match(), null);
      else {
        $this->record($this->get(), null);
      }
    }
    // In case not everything was popped
    if (isset($this->state_[0])) {
      $this->record(
        substr($this->string(), $this->state_[0][3],
          $this->pos() - $this->state_[0][3]),
          $this->state_[0][0]
      );
      $this->terminate();
    }
  }


  public static function guess_language($src, $info) {
    if (strpos($info['shebang'], 'ruby') !== false) return 1.0;
    elseif($info['shebang']) return 0;
    $p = 0;
    if (strpos($src, 'nil')) $p += 0.05;
    if (strpos($src, '.nil?')) $p += 0.02;
    if (strpos($src, '.empty?')) $p += 0.02;
    // interpolation
    if (strpos($src, '#{$')) $p += 0.02;
    // @ and $ vars
    if (preg_match('/@[a-zA-Z_]/', $src) && preg_match('/\\$[a-zA-Z_]/', $src))
      $p += 0.02;
    // symbols
    if (preg_match('/:[a-zA-Z_]/', $src)) $p += 0.01;
    // func def - no args
    if (preg_match("/^\s*+def\s++[a-zA-Z_]\w*+[ \t]*+[\n\r]/m", $src)) 
      $p += 0.1;
    // {|x[,y[,z...]]| is a very ruby-like construct
    if (preg_match('/ \\{ \\| 
      \s*+ [a-zA-Z_]\w*+ \s*+ 
        (,\s*+[a-zA-Z_]\w*+\s*+)*+
      \\|/x', $src))
      $p += 0.15;
    // so is 'do |x|'
    if (preg_match("/\\bdo\s*+\\|[^\\|\r\n]++\\|/", $src)) 
      $p += 0.05;

    // class defs with inheritance has quite distinct syntax
    // class x < y
    if (preg_match(
      "/^ \s* class \s+ \w+ \s* < \s* \w+(::\w+)* [\t ]*+ [\r\n] /mx",
      $src))
      $p += 0.1;

    $num_lines = $info['num_lines'];
    // let's say if 5% of lines are hash commented that's a good thing
    if (substr_count($src, '#') > $num_lines/20) $p += 0.05;
    // =~ /regex/
    if (preg_match('%=~\s++/%', $src)) $p += 0.02;

    if (preg_match('/unless\s+[^\?]++\?/', $src)) $p += 0.05;

    if (preg_match('/^(\s*+)def\s+.*^\1end\s/ms', $src)) $p += 0.05;

    if (preg_match('/\.to_\w+(?=\s|$)/', $src)) $p += 0.01;


    return $p;
  }
}