Changed non-destructive list-insert for 0th position to a desctructiv…

[Hunter11zolomon]

Changed non-destructive list-insert for 0th position to a destructive one

[Affonso-Gui]

Thanks for your contribution.

However, there are some problems that should be fixed before merging:

1. You modify a total of 5 files, but we only have interest in common.l. Please remove the other 4 files from this PR (eusmap, include, Makefile and common.l~)
2. It would also be nice to have destructive effect for null lists (not only for zero pos).
3. nreverse is allowed to have arbitrary side effects. From CLHS:

 (setq l (list 1 2 3)) =>  (1 2 3)
 (nreverse l) =>  (3 2 1)
 l =>  implementation-dependent

In eus it appears to be granted to have the list match the result, but code should not depend on this.

EusLisp/lisp/c/sequence.c

Lines 249 to 301 in 9cfd655

	pointer NREVERSE(ctx,n,argv)
	register context *ctx;
	int n;
	pointer argv[];
	{ register pointer a=argv[0],r=NIL, *vp;
	register eusinteger_t i,k,s,kk,x,y;
	register byte *cp;
	ckarg(1);
	if (a==NIL) return(NIL);
	else if (islist(a)) {
	i=0; r=a;
	while (islist(r)) { ckpush(ccar(r)); r=ccdr(r); i++;}
	r=a;
	while (i>0) { pointer_update(ccar(r),vpop()); r=ccdr(r); i--;}
	return(a);}
	else if (isarray(a)) {
	if (a->c.ary.rank>makeint(1)) error(E_NOSEQ);
	s=intval(a->c.ary.offset);
	k=intval(a->c.ary.dim[0]);
	a=a->c.ary.entity;}
	else if (isvector(a)) { k=vecsize(a); s=0;}
	else error(E_NOSEQ);
	kk=k/2; vp=a->c.vec.v;
	cp=a->c.str.chars;
	switch(elmtypeof(a)) {
	case ELM_BIT:
	for(i=0; i<kk; i++, s++) {
	x=bitref(a,s);
	y=bitref(a,k-i-1);
	#if (WORD_SIZE == 64)
	if (y) a->c.ivec.iv[s/64] |= (1L<<(s%64));
	else a->c.ivec.iv[s/64] &= ~(1L<<(s%64));
	if (x) a->c.ivec.iv[(k-i-1)/64] |= (1L<<((k-i-1)%64));
	else a->c.ivec.iv[(k-i-1)/64] &= ~(1L<<((k-i-1)%64));
	#else
	if (y) a->c.ivec.iv[s/32] |= (1<<(s%32));
	else a->c.ivec.iv[s/32] &= ~(1<<(s%32));
	if (x) a->c.ivec.iv[(k-i-1)/32] |= (1<<((k-i-1)%32));
	else a->c.ivec.iv[(k-i-1)/32] &= ~(1<<((k-i-1)%32));
	#endif
	}
	break;
	case ELM_FOREIGN: cp=a->c.foreign.chars;
	case ELM_CHAR: case ELM_BYTE:
	for(i=0; i<kk; i++, s++) {
	x=cp[s]; cp[s]=cp[k-i-1]; cp[k-i-1]=x;}
	break;
	default:
	vp=a->c.vec.v;
	for(i=0; i<kk; i++, s++) {
	r=vp[s]; pointer_update(vp[s],vp[k-i-1]); vp[k-i-1]=r;}
	break; }
	return(a);}

[Hunter11zolomon]

Ok, Thanks

I will look into the changes. And can you please elaborate on the side effects of nreverse?

[Affonso-Gui]

For nreverse, sequence might be destroyed and re-used to produce the result. 
The result might or might not be identical to sequence. 

In other words, the input may be destroyed as a way of optimizing the algorithm, having no guarantee that the result will correctly be stored in the input back again. In the case of sbcl, for example:

CL-USER> (setq lst (list 1 2 3))
(1 2 3)
CL-USER> (nreverse lst)
(3 2 1)
CL-USER> lst
(1)

[Hunter11zolomon]

I will try to find a work around.
Thanks for the input