Hash values of symbols: why look at the name?

Hash values of symbols: why look at the name?

[Daniel Colascione]

[-- Attachment #1: Type: text/plain, Size: 817 bytes --]

Right now, this form always evaluates to t:

  (eql (sxhash (make-symbol "a")) (sxhash (make-symbol "a")))

This form always evaluates to nil:

  (equal (make-symbol "a") (make-symbol "a"))

The hash of a symbol is the hash of its namel; that seems silly, since
symbol equality is based on identity. We should be able to just use the
symbol address instead, right?


=== modified file 'src/fns.c'
--- src/fns.c	2014-09-15 05:09:44 +0000
+++ src/fns.c	2014-09-16 00:14:52 +0000
@@ -4482,13 +4482,10 @@
       break;

     case Lisp_Misc:
+    case Lisp_Symbol:
       hash = XHASH (obj);
       break;

-    case Lisp_Symbol:
-      obj = SYMBOL_NAME (obj);
-      /* Fall through.  */
-
     case Lisp_String:
       hash = sxhash_string (SSDATA (obj), SBYTES (obj));
       break;



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

Re: Hash values of symbols: why look at the name?

[Stefan Monnier]

> The hash of a symbol is the hash of its namel; that seems silly, since
> symbol equality is based on identity.  We should be able to just use the
> symbol address instead, right?

Sounds right,


        Stefan