williamyaoh/named-read-macros

Make read macros more Lispy. Attach read macros to symbols.

NAMED-READ-MACROS

Copyright (c) 2017-2021 William Yao

Description

Read macros are pretty handy when Lispy sexps just don't cut it. Unfortunately, having to hunt around in the dispatch table or in the read macro table to avoid read macro collisions doesn't leave a whole bunch of usable characters. So instead, NAMED-READ-MACROS lets you attach a read macro to any symbol, and call it like any other function or macro.

(named-readtables:in-readtable named-read-macros:readtable)

(named-read-macros:define escapify
  (with-output-to-string (output)
    (loop for char = (read-char *standard-input* nil nil t)
          while char
          do (write-char char output))))

(escapify
  Any characters I put in here will be put into the output string, even
  things that would normally require escapes, like backslashes!

  \ \ \ \
end-escapify)
 => "Any characters I put in here will be put into the output string, even
       things that would normally require escapes, like backslashes!

       \\ \\ \\ \\"

You can hijack the Lisp reader without having to carefully avoid read macro conflicts! And since NAMED-READ-MACROS attaches read macros to symbols, you don't have to fuss about with readtables either, and you can export and import read macros the same way you do with functions and macros and all the other symbol-based stuff.

Usage

For any file which needs to use your named read macros, make sure to either switch your readtable to NAMED-READ-MACROS:READTABLE, or fuse a read table you're using with NAMED-READ-MACROS:READTABLE-MIXIN.

• READTABLE

  Contains the standard readtable, but with the macro character for #\( rewritten to check for a named read macro in the head position.

• READTABLE-MIXIN

  Just contains the rewritten macro for #\(.

Then define a named read macro with NAMED-READ-MACROS:DEFINE.

• DEFINE NAME &BODY BODY

  Creates a named read macro, which executes BODY in a context where *STANDARD-INPUT* is bound to a stream containing just the contents which the read macro was called with, and associates this macro with NAME.

  Just like a normal macro, BODY should return a Lisp form to then get evaluated where the named read macro was called.

  Because NAMED-READ-MACROS hijacks the Lisp reader, we can't rely on matching parentheses to know when the newly-defined read macro ends, so we look for the sequence of characters "END-${NAME}", immediately followed by a close parenthesis, in order to know when the read macro ends. Case is checked against NAME by transforming the ending string according to the case of the current readtable; if (SYMBOL-NAME NAME) matches the transformed ending string exactly, we've found the end tag. In particular, this means that using DEFINE with a pipe-enclosed symbol with lowercase characters will make such a read macro impossible to end under the standard readtable (though why one would define such a macro is another question entirely!)

  Leading whitespace after opening the read macro will not be passed to BODY, but trailing whitespace before the ending tag will.

  Note that DEFINE has no compile-time effects by default; the rationale is that doing so would also require any functions used by a read macro defined by DEFINE to be available at compile-time. Since this is not the usual, DEFINE explicitly has no compile-time effects to avoid this problem. If you want a named read macro to be available at compile-time, wrap DEFINE and any necessary functions in an explicit EVAL-WHEN.

Going forward

An interesting usage would be to write an LALR(1) parser generator for Common Lisp, but instead of generating a function or a standalone executable, it instead generates read macros. Any parsers generated could then be embedded directly within Lisp code, allowing for very complex read macros to be easily written, and compile directly to Lisp forms.

Another use would be something like Perl's __DATA__ directive, embedding files and data directly in Lisp code.