Fixed spec 1.9: String#lines

kernel/common/string18.rb

@@ -567,4 +567,112 @@ def <<(other)
     append(other)
   end
   alias_method :concat, :<<
+
+  # Splits <i>self</i> using the supplied parameter as the record separator
+  # (<code>$/</code> by default), passing each substring in turn to the supplied
+  # block. If a zero-length record separator is supplied, the string is split on
+  # <code>\n</code> characters, except that multiple successive newlines are
+  # appended together.
+  #
+  #   print "Example one\n"
+  #   "hello\nworld".each { |s| p s }
+  #   print "Example two\n"
+  #   "hello\nworld".each('l') { |s| p s }
+  #   print "Example three\n"
+  #   "hello\n\n\nworld".each('') { |s| p s }
+  #
+  # <em>produces:</em>
+  #
+  #   Example one
+  #   "hello\n"
+  #   "world"
+  #   Example two
+  #   "hel"
+  #   "l"
+  #   "o\nworl"
+  #   "d"
+  #   Example three
+  #   "hello\n\n\n"
+  #   "world"
+  def each_line(sep=$/)
+    return to_enum(:each_line, sep) unless block_given?
+
+    # weird edge case.
+    if sep.nil?
+      yield self
+      return self
+    end
+
+    sep = StringValue(sep)
+
+    pos = 0
+
+    size = @num_bytes
+    orig_data = @data
+
+    # If the separator is empty, we're actually in paragraph mode. This
+    # is used so infrequently, we'll handle it completely separately from
+    # normal line breaking.
+    if sep.empty?
+      sep = "\n\n"
+      pat_size = 2
+
+      while pos < size
+        nxt = find_string(sep, pos)
+        break unless nxt
+
+        while @data[nxt] == 10 and nxt < @num_bytes
+          nxt += 1
+        end
+
+        match_size = nxt - pos
+
+        # string ends with \n's
+        break if pos == @num_bytes
+
+        str = substring(pos, match_size)
+        yield str unless str.empty?
+
+        # detect mutation within the block
+        if !@data.equal?(orig_data) or @num_bytes != size
+          raise RuntimeError, "string modified while iterating"
+        end
+
+        pos = nxt
+      end
+
+      # No more separates, but we need to grab the last part still.
+      fin = substring(pos, @num_bytes - pos)
+      yield fin if fin and !fin.empty?
+
+    else
+
+      # This is the normal case.
+      pat_size = sep.size
+
+      while pos < size
+        nxt = find_string(sep, pos)
+        break unless nxt
+
+        match_size = nxt - pos
+        str = substring(pos, match_size + pat_size)
+        yield str unless str.empty?
+
+        # detect mutation within the block
+        if !@data.equal?(orig_data) or @num_bytes != size
+          raise RuntimeError, "string modified while iterating"
+        end
+
+        pos = nxt + pat_size
+      end
+
+      # No more separates, but we need to grab the last part still.
+      fin = substring(pos, @num_bytes - pos)
+      yield fin unless fin.empty?
+    end
+
+    self
+  end
+
+  alias_method :lines, :each_line
 end

kernel/common/string19.rb

@@ -645,4 +645,107 @@ def chr
     end
   end
 
+  # Splits <i>self</i> using the supplied parameter as the record separator
+  # (<code>$/</code> by default), passing each substring in turn to the supplied
+  # block. If a zero-length record separator is supplied, the string is split on
+  # <code>\n</code> characters, except that multiple successive newlines are
+  # appended together.
+  #
+  #   print "Example one\n"
+  #   "hello\nworld".each { |s| p s }
+  #   print "Example two\n"
+  #   "hello\nworld".each('l') { |s| p s }
+  #   print "Example three\n"
+  #   "hello\n\n\nworld".each('') { |s| p s }
+  #
+  # <em>produces:</em>
+  #
+  #   Example one
+  #   "hello\n"
+  #   "world"
+  #   Example two
+  #   "hel"
+  #   "l"
+  #   "o\nworl"
+  #   "d"
+  #   Example three
+  #   "hello\n\n\n"
+  #   "world"
+  def each_line(sep=$/)
+    return to_enum(:each_line, sep) unless block_given?
+
+    # weird edge case.
+    if sep.nil?
+      yield self
+      return self
+    end
+
+    sep = StringValue(sep)
+
+    pos = 0
+
+    size = @num_bytes
+    orig_data = @data
+
+    # If the separator is empty, we're actually in paragraph mode. This
+    # is used so infrequently, we'll handle it completely separately from
+    # normal line breaking.
+    if sep.empty?
+      sep = "\n\n"
+      pat_size = 2
+
+      while pos < size
+        nxt = find_string(sep, pos)
+        break unless nxt
+
+        while @data[nxt] == 10 and nxt < @num_bytes
+          nxt += 1
+        end
+
+        match_size = nxt - pos
+
+        # string ends with \n's
+        break if pos == @num_bytes
+
+        str = substring(pos, match_size)
+        yield str unless str.empty?
+
+        # detect mutation within the block
+        if !@data.equal?(orig_data) or @num_bytes != size
+          raise RuntimeError, "string modified while iterating"
+        end
+
+        pos = nxt
+      end
+
+      # No more separates, but we need to grab the last part still.
+      fin = substring(pos, @num_bytes - pos)
+      yield fin if fin and !fin.empty?
+
+    else
+
+      # This is the normal case.
+      pat_size = sep.size
+      unmodified_self = clone
+
+      while pos < size
+        nxt = unmodified_self.find_string(sep, pos)
+        break unless nxt
+
+        match_size = nxt - pos
+        str = unmodified_self.substring(pos, match_size + pat_size)
+        yield str unless str.empty?
+
+        pos = nxt + pat_size
+      end
+
+      # No more separates, but we need to grab the last part still.
+      fin = unmodified_self.substring(pos, @num_bytes - pos)
+      yield fin unless fin.empty?
+    end
+
+    self
+  end
+
+  alias_method :lines, :each_line
 end

spec/tags/19/ruby/core/string/lines_tags.txt

@@ -1 +0,0 @@
-fails:String#lines does not care if the string is modified while substituting