13.1.3 Port Buffers and Positions

Some ports—especially those that read from and write to files—are internally buffered:

• An input port is typically block-buffered by default, which means that on any read, the buffer is filled with immediately-available bytes to speed up future reads. Thus, if a file is modified between a pair of reads to the file, the second read can produce stale data. Calling file-position to set an input port’s file position flushes its buffer.

• An output port is typically block-buffered by default, though a terminal output port is line-buffered, and the initial error output port is unbuffered. An output buffer is filled with a sequence of written bytes to be committed as a group, either when the buffer is full (in block mode), when a newline is written (in line mode), when the port is closed via close-output-port, or when a flush is explicitly requested via a procedure like flush-output.

If a port supports buffering, its buffer mode can be changed via file-stream-buffer-mode (even if the port is not a file-stream port).

For an input port, peeking always places peeked bytes into the port’s buffer, even when the port’s buffer mode is 'none; furthermore, on some platforms, testing the port for input (via char-ready? or sync) may be implemented with a peek. If an input port’s buffer mode is 'none, then at most one byte is read for read-bytes-avail!*, read-bytes-avail!, peek-bytes-avail!*, or peek-bytes-avail!; if any bytes are buffered in the port (e.g., to satisfy a previous peek), the procedures may access multiple buffered bytes, but no further bytes are read.

In addition, the initial current output and error ports are automatically flushed when they are terminal ports (see terminal-port?) and when read, read-line, read-bytes, read-string, etc., are performed on the initial standard input port. (More precisely, instead of read, flushing is performed by the default port read handler; see port-read-handler.)

If flushing a file-stream port or TCP port encounters an error when writing, then all buffered bytes in the port are discarded. Consequently, a further attempt to flush or close the port will not fail.

Changed in version 7.4.0.10 of package base: Consistently, discard buffered bytes on error, including in a TCP output port.

If mode is provided, it must be one of 'none, 'line (output only), or 'block, and the port’s buffering is set accordingly. If the port does not support setting the mode, the exn:fail exception is raised.

If mode is not provided, the current mode is returned, or #f is returned if the mode cannot be determined. If port is an input port and mode is 'line, the exn:fail:contract exception is raised.

Calling file-position without a position on a port other than a file-stream port or string port returns the number of bytes that have been read from that port if the position is known (see Counting Positions, Lines, and Columns), otherwise the exn:fail:filesystem exception is raised.

For file-stream ports and string ports, the position-setting variant sets the read/write position to pos relative to the beginning of the file or (byte) string if pos is a number, or to the current end of the file or (byte) string if pos is eof. In position-setting mode, file-position raises the exn:fail:contract exception for port kinds other than file-stream ports and string ports. Furthermore, not all file-stream ports support setting the position; if file-position is called with a position argument on such a file-stream port, the exn:fail:filesystem exception is raised.

When file-position sets the position pos beyond the current size of an output file or (byte) string, the file/string is enlarged to size pos and the new region is filled with 0 bytes; in the case of a file. In the case of a file output port, the file might not be enlarged until more data is written to the file; in that case, beware that writing to a file opened in 'append mode on Unix and Mac OS will reset the file pointer to the end of a file before each write, which defeats file enlargement via file-position. If pos is beyond the end of an input file or (byte) string, then reading thereafter returns eof without changing the port’s position.

When changing the file position for an output port, the port is first flushed if its buffer is not empty. Similarly, setting the position for an input port clears the port’s buffer (even if the new position is the same as the old position). However, although input and output ports produced by open-input-output-file share the file position, setting the position via one port does not flush the other port’s buffer.

The new file size can be either larger or smaller than its current size, but “truncate” in this function’s name reflects that it is normally used to decrease the size of a file, since writing to a file or using file-position can extend a file’s size.