Next: , Previous: Memory Usage, Up: GNU Emacs Internals

E.5 Writing Emacs Primitives

Lisp primitives are Lisp functions implemented in C. The details of interfacing the C function so that Lisp can call it are handled by a few C macros. The only way to really understand how to write new C code is to read the source, but we can explain some things here.

An example of a special form is the definition of or, from eval.c. (An ordinary function would have the same general appearance.)

     DEFUN ("or", For, Sor, 0, UNEVALLED, 0,
       doc: /* Eval args until one of them yields non-nil, then return
     that value.
     The remaining args are not evalled at all.
     If all args return nil, return nil.
     usage: (or CONDITIONS ...)  */)
       (Lisp_Object args)
     {
       register Lisp_Object val = Qnil;
       struct gcpro gcpro1;
     
       GCPRO1 (args);
     
       while (CONSP (args))
         {
           val = eval_sub (XCAR (args));
           if (!NILP (val))
             break;
           args = XCDR (args);
         }
     
       UNGCPRO;
       return val;
     }

Let's start with a precise explanation of the arguments to the DEFUN macro. Here is a template for them:

     DEFUN (lname, fname, sname, min, max, interactive, doc)
lname
This is the name of the Lisp symbol to define as the function name; in the example above, it is or.
fname
This is the C function name for this function. This is the name that is used in C code for calling the function. The name is, by convention, ‘F’ prepended to the Lisp name, with all dashes (‘-’) in the Lisp name changed to underscores. Thus, to call this function from C code, call For.
sname
This is a C variable name to use for a structure that holds the data for the subr object that represents the function in Lisp. This structure conveys the Lisp symbol name to the initialization routine that will create the symbol and store the subr object as its definition. By convention, this name is always fname with ‘F’ replaced with ‘S’.
min
This is the minimum number of arguments that the function requires. The function or allows a minimum of zero arguments.
max
This is the maximum number of arguments that the function accepts, if there is a fixed maximum. Alternatively, it can be UNEVALLED, indicating a special form that receives unevaluated arguments, or MANY, indicating an unlimited number of evaluated arguments (the equivalent of &rest). Both UNEVALLED and MANY are macros. If max is a number, it must be more than min but less than 8.
interactive
This is an interactive specification, a string such as might be used as the argument of interactive in a Lisp function. In the case of or, it is 0 (a null pointer), indicating that or cannot be called interactively. A value of "" indicates a function that should receive no arguments when called interactively. If the value begins with a ‘(’, the string is evaluated as a Lisp form. For examples of the last two forms, see widen and narrow-to-region in editfns.c.
doc
This is the documentation string. It uses C comment syntax rather than C string syntax because comment syntax requires nothing special to include multiple lines. The ‘doc:’ identifies the comment that follows as the documentation string. The ‘/*’ and ‘*/’ delimiters that begin and end the comment are not part of the documentation string.

If the last line of the documentation string begins with the keyword ‘usage:’, the rest of the line is treated as the argument list for documentation purposes. This way, you can use different argument names in the documentation string from the ones used in the C code. ‘usage:’ is required if the function has an unlimited number of arguments.

All the usual rules for documentation strings in Lisp code (see Documentation Tips) apply to C code documentation strings too.

After the call to the DEFUN macro, you must write the argument list for the C function, including the types for the arguments. If the primitive accepts a fixed maximum number of Lisp arguments, there must be one C argument for each Lisp argument, and each argument must be of type Lisp_Object. (Various macros and functions for creating values of type Lisp_Object are declared in the file lisp.h.) If the primitive has no upper limit on the number of Lisp arguments, it must have exactly two C arguments: the first is the number of Lisp arguments, and the second is the address of a block containing their values. These have types int and Lisp_Object * respectively.

Within the function For itself, note the use of the macros GCPRO1 and UNGCPRO. These macros are defined for the sake of the few platforms which do not use Emacs' default stack-marking garbage collector. The GCPRO1 macro “protects” a variable from garbage collection, explicitly informing the garbage collector that that variable and all its contents must be as accessible. GC protection is necessary in any function which can perform Lisp evaluation by calling eval_sub or Feval as a subroutine, either directly or indirectly.

It suffices to ensure that at least one pointer to each object is GC-protected. Thus, a particular local variable can do without protection if it is certain that the object it points to will be preserved by some other pointer (such as another local variable that has a GCPRO). Otherwise, the local variable needs a GCPRO.

The macro GCPRO1 protects just one local variable. If you want to protect two variables, use GCPRO2 instead; repeating GCPRO1 will not work. Macros GCPRO3, GCPRO4, GCPRO5, and GCPRO6 also exist. All these macros implicitly use local variables such as gcpro1; you must declare these explicitly, with type struct gcpro. Thus, if you use GCPRO2, you must declare gcpro1 and gcpro2.

UNGCPRO cancels the protection of the variables that are protected in the current function. It is necessary to do this explicitly.

You must not use C initializers for static or global variables unless the variables are never written once Emacs is dumped. These variables with initializers are allocated in an area of memory that becomes read-only (on certain operating systems) as a result of dumping Emacs. See Pure Storage.

Defining the C function is not enough to make a Lisp primitive available; you must also create the Lisp symbol for the primitive and store a suitable subr object in its function cell. The code looks like this:

     defsubr (&sname);

Here sname is the name you used as the third argument to DEFUN.

If you add a new primitive to a file that already has Lisp primitives defined in it, find the function (near the end of the file) named syms_of_something, and add the call to defsubr there. If the file doesn't have this function, or if you create a new file, add to it a syms_of_filename (e.g., syms_of_myfile). Then find the spot in emacs.c where all of these functions are called, and add a call to syms_of_filename there.

The function syms_of_filename is also the place to define any C variables that are to be visible as Lisp variables. DEFVAR_LISP makes a C variable of type Lisp_Object visible in Lisp. DEFVAR_INT makes a C variable of type int visible in Lisp with a value that is always an integer. DEFVAR_BOOL makes a C variable of type int visible in Lisp with a value that is either t or nil. Note that variables defined with DEFVAR_BOOL are automatically added to the list byte-boolean-vars used by the byte compiler.

If you want to make a Lisp variables that is defined in C behave like one declared with defcustom, add an appropriate entry to cus-start.el.

If you define a file-scope C variable of type Lisp_Object, you must protect it from garbage-collection by calling staticpro in syms_of_filename, like this:

     staticpro (&variable);

Here is another example function, with more complicated arguments. This comes from the code in window.c, and it demonstrates the use of macros and functions to manipulate Lisp objects.

     DEFUN ("coordinates-in-window-p", Fcoordinates_in_window_p,
       Scoordinates_in_window_p, 2, 2, 0,
       doc: /* Return non-nil if COORDINATES are in WINDOW.
       ...
       or `right-margin' is returned.  */)
       (register Lisp_Object coordinates, Lisp_Object window)
     {
       struct window *w;
       struct frame *f;
       int x, y;
       Lisp_Object lx, ly;
     
       CHECK_LIVE_WINDOW (window);
       w = XWINDOW (window);
       f = XFRAME (w->frame);
       CHECK_CONS (coordinates);
       lx = Fcar (coordinates);
       ly = Fcdr (coordinates);
       CHECK_NUMBER_OR_FLOAT (lx);
       CHECK_NUMBER_OR_FLOAT (ly);
       x = FRAME_PIXEL_X_FROM_CANON_X (f, lx) + FRAME_INTERNAL_BORDER_WIDTH(f);
       y = FRAME_PIXEL_Y_FROM_CANON_Y (f, ly) + FRAME_INTERNAL_BORDER_WIDTH(f);
     
       switch (coordinates_in_window (w, x, y))
         {
         case ON_NOTHING:            /* NOT in window at all. */
           return Qnil;
     
         ...
     
         case ON_MODE_LINE:          /* In mode line of window. */
           return Qmode_line;
     
         ...
     
         case ON_SCROLL_BAR:         /* On scroll-bar of window.  */
           /* Historically we are supposed to return nil in this case.  */
           return Qnil;
     
         default:
           abort ();
         }
     }

Note that C code cannot call functions by name unless they are defined in C. The way to call a function written in Lisp is to use Ffuncall, which embodies the Lisp function funcall. Since the Lisp function funcall accepts an unlimited number of arguments, in C it takes two: the number of Lisp-level arguments, and a one-dimensional array containing their values. The first Lisp-level argument is the Lisp function to call, and the rest are the arguments to pass to it. Since Ffuncall can call the evaluator, you must protect pointers from garbage collection around the call to Ffuncall.

The C functions call0, call1, call2, and so on, provide handy ways to call a Lisp function conveniently with a fixed number of arguments. They work by calling Ffuncall.

eval.c is a very good file to look through for examples; lisp.h contains the definitions for some important macros and functions.

If you define a function which is side-effect free, update the code in byte-opt.el that binds side-effect-free-fns and side-effect-and-error-free-fns so that the compiler optimizer knows about it.