stream: Optimize `StringIStream.read_all`
[nit.git] / lib / standard / stream.nit
index 0c7a1a3..ddc01fe 100644 (file)
@@ -16,8 +16,6 @@ import error
 
 in "C" `{
        #include <unistd.h>
-       #include <poll.h>
-       #include <errno.h>
        #include <string.h>
        #include <signal.h>
 `}
@@ -63,16 +61,103 @@ abstract class IStream
        end
 
        # Read a string until the end of the line.
+       #
+       # The line terminator '\n' and '\r\n', if any, is removed in each line.
+       #
+       # ~~~
+       # var txt = "Hello\n\nWorld\n"
+       # var i = new StringIStream(txt)
+       # assert i.read_line == "Hello"
+       # assert i.read_line == ""
+       # assert i.read_line == "World"
+       # assert i.eof
+       # ~~~
+       #
+       # Only LINE FEED (`\n`), CARRIAGE RETURN & LINE FEED (`\r\n`), and
+       # the end or file (EOF) is considered to delimit the end of lines.
+       # CARRIAGE RETURN (`\r`) alone is not used for the end of line.
+       #
+       # ~~~
+       # var txt2 = "Hello\r\n\n\rWorld"
+       # var i2 = new StringIStream(txt2)
+       # assert i2.read_line == "Hello"
+       # assert i2.read_line == ""
+       # assert i2.read_line == "\rWorld"
+       # assert i2.eof
+       # ~~~
+       #
+       # NOTE: Use `append_line_to` if the line terminator needs to be preserved.
        fun read_line: String
        do
                if last_error != null then return ""
-               assert not eof
+               if eof then return ""
                var s = new FlatBuffer
                append_line_to(s)
-               return s.to_s
+               return s.to_s.chomp
        end
 
+       # Read all the lines until the eof.
+       #
+       # The line terminator '\n' and `\r\n` is removed in each line,
+       #
+       # ~~~
+       # var txt = "Hello\n\nWorld\n"
+       # var i = new StringIStream(txt)
+       # assert i.read_lines == ["Hello", "", "World"]
+       # ~~~
+       #
+       # This method is more efficient that splitting
+       # the result of `read_all`.
+       #
+       # NOTE: SEE `read_line` for details.
+       fun read_lines: Array[String]
+       do
+               var res = new Array[String]
+               while not eof do
+                       res.add read_line
+               end
+               return res
+       end
+
+       # Return an iterator that read each line.
+       #
+       # The line terminator '\n' and `\r\n` is removed in each line,
+       # The line are read with `read_line`. See this method for details.
+       #
+       # ~~~
+       # var txt = "Hello\n\nWorld\n"
+       # var i = new StringIStream(txt)
+       # assert i.each_line.to_a == ["Hello", "", "World"]
+       # ~~~
+       #
+       # Unlike `read_lines` that read all lines at the call, `each_line` is lazy.
+       # Therefore, the stream should no be closed until the end of the stream.
+       #
+       # ~~~
+       # i = new StringIStream(txt)
+       # var el = i.each_line
+       #
+       # assert el.item == "Hello"
+       # el.next
+       # assert el.item == ""
+       # el.next
+       #
+       # i.close
+       #
+       # assert not el.is_ok
+       # # closed before "world" is read
+       # ~~~
+       fun each_line: LineIterator do return new LineIterator(self)
+
        # Read all the stream until the eof.
+       #
+       # The content of the file is returned verbatim.
+       #
+       # ~~~
+       # var txt = "Hello\n\nWorld\n"
+       # var i = new StringIStream(txt)
+       # assert i.read_all == txt
+       # ~~~
        fun read_all: String
        do
                if last_error != null then return ""
@@ -85,6 +170,38 @@ abstract class IStream
        end
 
        # Read a string until the end of the line and append it to `s`.
+       #
+       # Unlike `read_line` and other related methods,
+       # the line terminator '\n', if any, is preserved in each line.
+       # Use the method `Text::chomp` to safely remove it.
+       #
+       # ~~~
+       # var txt = "Hello\n\nWorld\n"
+       # var i = new StringIStream(txt)
+       # var b = new FlatBuffer
+       # i.append_line_to(b)
+       # assert b == "Hello\n"
+       # i.append_line_to(b)
+       # assert b == "Hello\n\n"
+       # i.append_line_to(b)
+       # assert b == txt
+       # assert i.eof
+       # ~~~
+       #
+       # If `\n` is not present at the end of the result, it means that
+       # a non-eol terminated last line was returned.
+       #
+       # ~~~
+       # var i2 = new StringIStream("hello")
+       # assert not i2.eof
+       # var b2 = new FlatBuffer
+       # i2.append_line_to(b2)
+       # assert b2 == "hello"
+       # assert i2.eof
+       # ~~~
+       #
+       # NOTE: The single character LINE FEED (`\n`) delimits the end of lines.
+       # Therefore CARRIAGE RETURN & LINE FEED (`\r\n`) is also recognized.
        fun append_line_to(s: Buffer)
        do
                if last_error != null then return
@@ -105,6 +222,53 @@ abstract class IStream
        fun eof: Bool is abstract
 end
 
+# Iterator returned by `IStream::each_line`.
+# See the aforementioned method for details.
+class LineIterator
+       super Iterator[String]
+
+       # The original stream
+       var stream: IStream
+
+       redef fun is_ok
+       do
+               var res = not stream.eof
+               if not res and close_on_finish then stream.close
+               return res
+       end
+
+       redef fun item
+       do
+               var line = self.line
+               if line == null then
+                       line = stream.read_line
+               end
+               self.line = line
+               return line
+       end
+
+       # The last line read (cache)
+       private var line: nullable String = null
+
+       redef fun next
+       do
+               # force the read
+               if line == null then item
+               # drop the line
+               line = null
+       end
+
+       # Close the stream when the stream is at the EOF.
+       #
+       # Default is false.
+       var close_on_finish = false is writable
+
+       redef fun finish
+       do
+               if close_on_finish then stream.close
+       end
+end
+
 # IStream capable of declaring if readable without blocking
 abstract class PollableIStream
        super IStream
@@ -158,12 +322,9 @@ abstract class BufferedIStream
        super IStream
        redef fun read_char
        do
-               if last_error != null then return 0
-               if eof then last_error = new IOError("Stream has reached eof")
-               if _buffer_pos >= _buffer.length then
-                       fill_buffer
-               end
-               if _buffer_pos >= _buffer.length then
+               if last_error != null then return -1
+               if eof then
+                       last_error = new IOError("Stream has reached eof")
                        return -1
                end
                var c = _buffer.chars[_buffer_pos]
@@ -176,7 +337,6 @@ abstract class BufferedIStream
                if last_error != null then return ""
                if _buffer.length == _buffer_pos then
                        if not eof then
-                               fill_buffer
                                return read(i)
                        end
                        return ""
@@ -214,6 +374,15 @@ abstract class BufferedIStream
                        var i = _buffer_pos
                        while i < _buffer.length and _buffer.chars[i] != '\n' do i += 1
 
+                       var eol
+                       if i < _buffer.length then
+                               assert _buffer.chars[i] == '\n'
+                               i += 1
+                               eol = true
+                       else
+                               eol = false
+                       end
+
                        # if there is something to append
                        if i > _buffer_pos then
                                # Enlarge the string (if needed)
@@ -225,25 +394,30 @@ abstract class BufferedIStream
                                        s.add(_buffer.chars[j])
                                        j += 1
                                end
+                               _buffer_pos = i
+                       else
+                               assert end_reached
+                               return
                        end
 
-                       if i < _buffer.length then
-                               # so \n is in _buffer[i]
-                               _buffer_pos = i + 1 # skip \n
+                       if eol then
+                               # so \n is found
                                return
                        else
                                # so \n is not found
-                               _buffer_pos = i
-                               if end_reached then
-                                       return
-                               else
-                                       fill_buffer
-                               end
+                               if end_reached then return
+                               fill_buffer
                        end
                end
        end
 
-       redef fun eof do return _buffer_pos >= _buffer.length and end_reached
+       redef fun eof
+       do
+               if _buffer_pos < _buffer.length then return false
+               if end_reached then return true
+               fill_buffer
+               return _buffer_pos >= _buffer.length and end_reached
+       end
 
        # The buffer
        private var buffer: nullable FlatBuffer = null
@@ -265,136 +439,12 @@ abstract class BufferedIStream
        end
 end
 
+# An Input/Output Stream
 abstract class IOStream
        super IStream
        super OStream
 end
 
-##############################################################"
-
-abstract class FDStream
-       super IOS
-       # File description
-       var fd: Int
-
-       redef fun close do native_close(fd)
-
-       private fun native_close(i: Int): Int is extern "stream_FDStream_FDStream_native_close_1"
-       private fun native_read_char(i: Int): Int is extern "stream_FDStream_FDStream_native_read_char_1"
-       private fun native_read(i: Int, buf: NativeString, len: Int): Int is extern "stream_FDStream_FDStream_native_read_3"
-       private fun native_write(i: Int, buf: NativeString, len: Int): Int is extern "stream_FDStream_FDStream_native_write_3"
-       private fun native_write_char(i: Int, c: Char): Int is extern "stream_FDStream_FDStream_native_write_char_2"
-end
-
-class FDIStream
-       super FDStream
-       super IStream
-       redef var eof: Bool = false
-       
-       redef fun read_char
-       do
-               var nb = native_read_char(fd)
-               if nb == -1 then eof = true
-               return nb
-       end
-end
-
-class FDOStream
-       super FDStream
-       super OStream
-       redef var is_writable = true
-
-       redef fun write(s)
-       do
-               var nb = native_write(fd, s.to_cstring, s.length)
-               if nb < s.length then is_writable = false
-       end
-end
-
-class FDIOStream
-       super FDIStream
-       super FDOStream
-       super IOStream
-end
-
-redef interface Object
-       # returns first available stream to read or write to
-       # return null on interruption (possibly a signal)
-       protected fun poll( streams : Sequence[FDStream] ) : nullable FDStream
-       do
-               var in_fds = new Array[Int]
-               var out_fds = new Array[Int]
-               var fd_to_stream = new HashMap[Int,FDStream]
-               for s in streams do
-                       var fd = s.fd
-                       if s isa FDIStream then in_fds.add( fd )
-                       if s isa FDOStream then out_fds.add( fd )
-
-                       fd_to_stream[fd] = s
-               end
-
-               var polled_fd = intern_poll( in_fds, out_fds )
-
-               if polled_fd == null then
-                       return null
-               else
-                       return fd_to_stream[polled_fd]
-               end
-       end
-
-       private fun intern_poll(in_fds: Array[Int], out_fds: Array[Int]) : nullable Int is extern import Array[Int].length, Array[Int].[], Int.as(nullable Int) `{
-               int in_len, out_len, total_len;
-               struct pollfd *c_fds;
-               sigset_t sigmask;
-               int i;
-               int first_polled_fd = -1;
-               int result;
-
-               in_len = Array_of_Int_length( in_fds );
-               out_len = Array_of_Int_length( out_fds );
-               total_len = in_len + out_len;
-               c_fds = malloc( sizeof(struct pollfd) * total_len );
-
-               /* input streams */
-               for ( i=0; i<in_len; i ++ ) {
-                       int fd;
-                       fd = Array_of_Int__index( in_fds, i );
-
-                       c_fds[i].fd = fd;
-                       c_fds[i].events = POLLIN;
-               }
-
-               /* output streams */
-               for ( i=0; i<out_len; i ++ ) {
-                       int fd;
-                       fd = Array_of_Int__index( out_fds, i );
-
-                       c_fds[i].fd = fd;
-                       c_fds[i].events = POLLOUT;
-               }
-
-               /* poll all fds, unlimited timeout */
-               result = poll( c_fds, total_len, -1 );
-
-               if ( result > 0 ) {
-                       /* analyse results */
-                       for ( i=0; i<total_len; i++ )
-                               if ( c_fds[i].revents & c_fds[i].events || /* awaited event */
-                                        c_fds[i].revents & POLLHUP ) /* closed */
-                               {
-                                       first_polled_fd = c_fds[i].fd;
-                                       break;
-                               }
-
-                       return Int_as_nullable( first_polled_fd );
-               }
-               else if ( result < 0 )
-                       fprintf( stderr, "Error in Stream:poll: %s\n", strerror( errno ) );
-
-               return null_Int();
-       `}
-end
-
 # Stream to a String.
 #
 # Mainly used for compatibility with OStream type and tests.
@@ -410,7 +460,9 @@ class StringOStream
                content.add(str.to_s)
        end
 
+       # Is the stream closed?
        protected var closed = false
+
        redef fun close do closed = true
 end
 
@@ -441,5 +493,12 @@ class StringIStream
                source = ""
        end
 
+       redef fun read_all do
+               var c = cursor
+               cursor = source.length
+               if c == 0 then return source
+               return source.substring_from(c)
+       end
+
        redef fun eof do return cursor >= source.length
 end