In this chapter an example is provided to illustrate the main features of the stream
package. In this example a stream class is defined to provide a wrapper for file-stream
which uses the Unicode Line Separator instead of the usual ASCII CR/LF combination to mark the end of lines in the file. Methods are then defined, specializing on the user defined stream class to ensure that it handles reading from and writing to a file correctly.
Streams can be capable of input or output (or both), and may deal with characters or with binary elements. The stream
package provides a number of stream classes with different capabilities from which user defined streams can inherit. In our example the stream must be capable of input and output, and must read characters. The following code defines our stream class appropriately:
(defclass unicode-ls-stream
(stream:fundamental-character-input-stream
stream:fundamental-character-output-stream)
((file-stream :initform nil
:initarg :file-stream
:accessor ls-stream-file-stream)))
The new class, unicode-ls-stream
, has fundamental-character-input-stream and fundamental-character-output-stream as its superclasses, which means it inherits the relevant default character I/O methods. We shall be overriding some of these with more relevant and efficient implementations later.
Note that we have also provided a file-stream slot. When making an instance of unicode-ls-stream
we can create an instance of a Common Lisp file stream in this slot. This allows us to use the Common Lisp file stream functionality for reading from and writing to a file.
We know that the stream will read from a file using file-stream
functionality and that the stream element type will be character
. The following defines a method on stream-element-type to return the correct element type.
(defmethod stream-element-type ((stream unicode-ls-stream))
'character)
Streams can be defined for input only, output only, or both. In our example, the unicode-ls-stream
class needs to be able to read from a file and write to a file, and we therefore defined it to inherit from an input and an output stream class. We could have defined disjoint classes instead, one inheriting from fundamental-character-input-stream and the other from fundamental-character-output-stream. This would have allowed us to rely on the default methods for the direction predicates.
However, given that we have defined one bi-directional stream class, we must define our own methods for the direction predicates. To allow this, the Common Lisp predicates input-stream-p and output-stream-p
are implemented as generic functions.
(defmethod input-stream-p ((stream unicode-ls-stream))
(input-stream-p (ls-stream-file-stream stream)))
(defmethod output-stream-p ((stream unicode-ls-stream))
(output-stream-p (ls-stream-file-stream stream)))
The above code allows us to "trampoline" the correct direction predicate functionality from file-stream
, using the ls-stream-file-stream
accessor we defined previously.
The following method for stream-read-char reads a character from the stream. If the character read is a #\Line-Separator
, then the method returns #\Newline
, otherwise the character read is returned. stream-read-char returns :eof
at the end of the file.
(defmethod stream:stream-read-char ((stream unicode-ls-stream))
(let ((char (read-char (ls-stream-file-stream stream)
nil :eof)))
(if (eql char #\Line-Separator)
#\Newline
char)))
There is no need to define a new method for stream-read-line as the default method uses stream-read-char repeatedly to read a line, and our implementation of stream-read-char ensures that this will work.
We also need to make sure that if a #\Newline
is unread, it is unread as a #\Line-Separator
. The following method for
stream-unread-char
uses the Common Lisp file stream function unread-char
to achieve this.
(defmethod stream:stream-unread-char ((stream unicode-ls-stream)
char)
(unread-char (if (eql char #\Newline) #\Line-Separator char)
(ls-stream-file-stream stream)))
Finally, although the default methods for
stream-listen
and
stream-clear-input
would work for our stream, it is faster to use the functions provided by file-stream
, again using our accessor ls-stream-file-stream
.
(defmethod stream:stream-listen ((stream unicode-ls-stream))
(listen (ls-stream-file-stream stream)))
(defmethod stream:stream-clear-input ((stream unicode-ls-stream))
(clear-input (ls-stream-file-stream stream)))
The following method for
stream-write-char
uses write-char
to write a character to the stream. If the character written to unicode-ls-stream
is a #\Newline
, then the method writes a #\Line-Separator
to the file stream.
(defmethod stream:stream-write-char ((stream unicode-ls-stream)
char)
(write-char (if (eql char #\Newline)
#\Line-Separator
char)
(ls-stream-file-stream stream)))
The default method for
stream-write-string
calls
stream-write-char
repeatedly to write a string to the stream. However, the following is a more efficient implementation for our stream.
(defmethod stream:stream-write-string ((stream unicode-ls-stream)
string &optional (start 0)
(end (length string)))
(loop with i = start
until (>= i end)
do (let* ((newline (position #\Newline
string :start i :end end))
(this-end (or newline end)))
(write-string string (ls-stream-file-stream stream)
:start i :end this-end)
(incf i this-end)
(when newline
(stream:stream-terpri stream)
(incf i)))
finally (return string)))
We do not need to define our own method for
stream-terpri
, as the default uses
stream-write-char
, and therefore works appropriately
To be useful, the
stream-line-column
and
stream-start-line-p
generic functions need to know the number of characters preceding a #\Line-Separator
. However, since the LispWorks file stream records line position only by #\Newline
characters, this information is not available. Hence we define the two generic functions to return nil
:
(defmethod stream:stream-line-column
((stream unicode-ls-stream))
nil)
(defmethod stream:stream-start-line-p
((stream unicode-ls-stream))
nil)
Finally, the methods for
stream-force-output
,
stream-finish-output
and
stream-clear-output
are "trampolined" from the standard force-output
, finish-output
and clear-output
functions.
(defmethod stream:stream-force-output ((stream
unicode-ls-stream))
(force-output (ls-stream-file-stream stream)))
(defmethod stream:stream-finish-output ((stream
unicode-ls-stream))
(finish-output (ls-stream-file-stream stream)))
(defmethod stream:stream-clear-output ((stream
unicode-ls-stream))
(clear-output (ls-stream-file-stream stream)))
Now that the stream class has been defined, and all the methods relevant to it have been set up, we can create an instance of our user defined stream to test it. The following function takes a filename and optionally a stream direction as its arguments and makes an instance of unicode-ls-stream
. It ensures that the file-stream
slot of the stream contains a Common Lisp file-stream
capable of reading from or writing to a file given by the filename argument.
(defun open-unicode-ls-file (filename &key (direction :input))
(make-instance 'unicode-ls-stream :file-stream
(open filename
:direction direction
:external-format :unicode
:element-type 'character)))
The following macro uses open-unicode-ls-stream
in a similar manner to the Common Lisp macro with-open-file
:
(defmacro with-open-unicode-ls-file ((var filename
&key (direction :input))
&body body)
`(let ((,var (open-unicode-ls-file ,filename
:direction ,direction)))
(unwind-protect
(progn ,@body)
(close ,var))))
We now have the required functions and macros to test our user defined stream. The following code uses config.sys
as a source of input to an instance of our stream, and outputs it to the file unicode-ls.out
, changing all occurrences of #\Newline
to #\Line-Separator
in the process.
(with-open-unicode-ls-file (ss "C:\\unicode-ls.out"
:direction :output)
(write-line "-*- Encoding: Unicode; -*-" ss)
(with-open-file (ii "C:\\config.sys") ; Don't edit this file!
(loop with line = nil
while (setf line (read-line ii nil nil))
do (write-line line ss))))
After running the above code, if your load the file C:\unicode-ls.out
into an editor (for example, a LispWorks editor), you can see the line separator used instead of CR/LF. Most editors do not yet recognize the Unicode Line Separator character yet. In some editors it appears as a blank glyph, whereas in the LispWorks editor it appears as <2028>
. In LispWorks you can use Alt+X What Cursor Position
or Ctrl+X =
to identify the unprintable characters.
You can also use the follow code to print out the contents of the new file line by line.
(with-open-unicode-ls-file (ss "C:\\unicode-ls.out")
(loop while (when-let (line (read-line ss nil nil))
(write-line line))))
LispWorks User Guide and Reference Manual - 13 Feb 2015