[Ecls-list] APROPOS issues, proposed enhancement

[Matthew Mondor]

I noticed that the native APROPOS (not SLIME's) returns duplicate
symbols, it seems to happen notably for symbols of the COMMON-LISP
package:

> (apropos '#:most-positive)
MOST-POSITIVE-LONG-FLOAT  Constant: 1.7976931348623157d308
MOST-POSITIVE-FIXNUM  Constant: 2305843009213693951
MOST-POSITIVE-DOUBLE-FLOAT  Constant: 1.7976931348623157d308
MOST-POSITIVE-SHORT-FLOAT  Constant: 3.4028235e38
MOST-POSITIVE-SINGLE-FLOAT  Constant: 3.4028235e38
MOST-POSITIVE-LONG-FLOAT  Constant: 1.7976931348623157d308
MOST-POSITIVE-FIXNUM  Constant: 2305843009213693951
MOST-POSITIVE-DOUBLE-FLOAT  Constant: 1.7976931348623157d308
MOST-POSITIVE-SHORT-FLOAT  Constant: 3.4028235e38
MOST-POSITIVE-SINGLE-FLOAT  Constant: 3.4028235e38

This is of course allowed by the standard (actually an APROPOS-LIST
issue, although it can be annoying and some implementations avoid
this, such as SBCL), but if we specify a package designator to tell
APROPOS to only return symbols from C, for instance, it seems that we
are getting unintended symbols from the COMMON-LISP package, and way
more duplicates of them:

> (apropos '#:push :c)
C::C1STACK-PUSH  Function
C::C1STACK-PUSH-VALUES  Function
C::PUSH-VARS  Function
C::C2STACK-PUSH-VALUES  Function
C::P1STACK-PUSH-VALUES  Function
C::COMPILER-PUSH-EVENTS  Function
C::STACK-PUSH
C::STACK-PUSH-VALUES
VECTOR-PUSH-EXTEND  Function
PUSHNEW  Macro
VECTOR-PUSH  Function
PUSH  Macro
VECTOR-PUSH-EXTEND  Function
PUSHNEW  Macro
VECTOR-PUSH  Function
PUSH  Macro
VECTOR-PUSH-EXTEND  Function
PUSHNEW  Macro
VECTOR-PUSH  Function
PUSH  Macro
VECTOR-PUSH-EXTEND  Function
PUSHNEW  Macro
VECTOR-PUSH  Function
PUSH  Macro
[...] more duplicates from COMMON-LISP package [...]

Is this part a bug?


Sample SBCL output:

* (apropos '#:most-positive)
MOST-POSITIVE-DOUBLE-FLOAT (bound)
MOST-POSITIVE-FIXNUM (bound)
MOST-POSITIVE-LONG-FLOAT (bound)
MOST-POSITIVE-SHORT-FLOAT (bound)
MOST-POSITIVE-SINGLE-FLOAT (bound)
SB-C::MOST-POSITIVE-BOUND
MOST-POSITIVE-WORD (bound)
SB-KERNEL:MOST-POSITIVE-EXACTLY-DOUBLE-FLOAT-FIXNUM (bound)
SB-KERNEL:MOST-POSITIVE-EXACTLY-SINGLE-FLOAT-FIXNUM (bound)
SB-VM:MOST-POSITIVE-COST

* (apropos '#:most-positive '#:sb-c)
SB-C::MOST-POSITIVE-BOUND


Sample CLisp output:

[1]> (apropos '#:most-positive)
MOST-POSITIVE-DOUBLE-FLOAT                 constant
MOST-POSITIVE-FIXNUM                       constant
MOST-POSITIVE-LONG-FLOAT                   variable
MOST-POSITIVE-SHORT-FLOAT                  constant
MOST-POSITIVE-SINGLE-FLOAT                 constant

[2]> (apropos '#:push '#:system)
SYSTEM::APPLY&PUSH                      
SYSTEM::CALL&PUSH                       
SYSTEM::CALL1&PUSH                      
SYSTEM::CALL2&PUSH                      
SYSTEM::CALLC&PUSH                      
SYSTEM::CALLCKEY&PUSH                   
SYSTEM::CALLS1&PUSH                     
SYSTEM::CALLS2&PUSH                     
SYSTEM::CALLSR&PUSH                     
SYSTEM::CAR&PUSH                        
SYSTEM::CDR&PUSH                        
SYSTEM::CONS&PUSH                       
SYSTEM::CONST&PUSH                      
SYSTEM::CONST&SYMBOL-FUNCTION&PUSH      
SYSTEM::COPY-CLOSURE&PUSH               
SYSTEM::FUNCALL&PUSH                    
SYSTEM::GETVALUE&PUSH                   
SYSTEM::HANDLER-BEGIN&PUSH              
SYSTEM::INSERT-COMBINED-LAPS-NILPUSHER-P 
SYSTEM::JSR&PUSH                        
SYSTEM::LIST&PUSH                       
SYSTEM::LIST*&PUSH                      
SYSTEM::LOAD&CAR&PUSH                   
SYSTEM::LOAD&CDR&PUSH                   
SYSTEM::LOAD&DEC&PUSH                   
SYSTEM::LOAD&INC&PUSH                   
SYSTEM::LOAD&PUSH                       
SYSTEM::LOADC&PUSH                      
SYSTEM::LOADI&PUSH                      
SYSTEM::LOADV&PUSH                      
SYSTEM::MAKE-STRING-PUSH-STREAM            function
SYSTEM::MAKE-VECTOR1&PUSH               
SYSTEM::NIL&PUSH                        
PUSH                                       macro
SYSTEM::PUSH-*DENV*                        function
SYSTEM::PUSH-*VENV*                        function
SYSTEM::PUSH-NIL                        
SYSTEM::PUSH-SPECIALS                      function
SYSTEM::PUSH-UNBOUND                    
PUSHNEW                                    macro
SYSTEM::T&PUSH                          
VECTOR-PUSH                                function
VECTOR-PUSH-EXTEND                         function

In this case above some packages outside of SYSTEM are still shown,
though...


Here is a replacement which doesn't care that APROPOS-LIST might return
duplicates, but which strips duplicates and sorts results for output.
Perhaps somewhat of a hack; I initially tried using PUSHNEW in vein,
then thought about using a hash table to eliminate duplicates, but
since I wanted sorted results anyway, it seemed best to simply sort
backwards and remove duplicates in a second pass.  Using the existing
PRINT-SYMBOL-APROPOS and later dealing with strings was simpler than
dealing with qualified and unqualified, exported and unexported symbols
for matching and sorting...


(defun apropos (string &optional package)
  "Args: (string &optional (package nil))
Prints those symbols whose print-names contain STRING as substring.  If
PACKAGE is non-NIL, then only the specified PACKAGE is searched."
  (setq string (string string))
  (mapc #'princ
        (loop
           with dupes = (sort
                         (mapcar #'(lambda (symbol)
                                     (with-output-to-string (stream)
                                       (let ((*standard-output* stream))
                                         (print-symbol-apropos symbol))))
                                 (apropos-list string package))
                         #'string-greaterp)
           with set = '()
           with last = nil
           for item in dupes
           do
             (unless (and last (string-equal last item))
               (push item set))
             (setq last item)
           finally (return set)))
  (values))


New results with it:

> (apropos '#:most-positive)
MOST-POSITIVE-DOUBLE-FLOAT  Constant: 1.7976931348623157d308
MOST-POSITIVE-FIXNUM  Constant: 2305843009213693951
MOST-POSITIVE-LONG-FLOAT  Constant: 1.7976931348623157d308
MOST-POSITIVE-SHORT-FLOAT  Constant: 3.4028235e38
MOST-POSITIVE-SINGLE-FLOAT  Constant: 3.4028235e38

> (apropos '#:push '#:c)
C::C1STACK-PUSH  Function
C::C1STACK-PUSH-VALUES  Function
C::C2STACK-PUSH-VALUES  Function
C::COMPILER-PUSH-EVENTS  Function
C::P1STACK-PUSH-VALUES  Function
C::PUSH-VARS  Function
C::STACK-PUSH
C::STACK-PUSH-VALUES
PUSH  Macro
PUSHNEW  Macro
VECTOR-PUSH  Function
VECTOR-PUSH-EXTEND  Function


Thanks,
-- 
Matt