static VALUE
rb_ary_sample(int argc, VALUE *argv, VALUE ary)
{
    VALUE nv, result, *ptr;
    long n, len, i, j, k, idx[10];
    len = RARRAY_LEN(ary);
    if (argc == 0) {
        if (len == 0) return Qnil;
        i = len == 1 ? 0 : (long)(rb_genrand_real()*len);
        return RARRAY_PTR(ary)[i];
    }
    rb_scan_args(argc, argv, "1", &nv);
    n = NUM2LONG(nv);
    if (n < 0) rb_raise(rb_eArgError, "negative sample number");
    ptr = RARRAY_PTR(ary);
    len = RARRAY_LEN(ary);
    if (n > len) n = len;
    switch (n) {
      case 0: return rb_ary_new2(0);
      case 1:
        return rb_ary_new4(1, &ptr[(long)(rb_genrand_real()*len)]);
      case 2:
        i = (long)(rb_genrand_real()*len);
        j = (long)(rb_genrand_real()*(len-1));
        if (j >= i) j++;
        return rb_ary_new3(2, ptr[i], ptr[j]);
      case 3:
        i = (long)(rb_genrand_real()*len);
        j = (long)(rb_genrand_real()*(len-1));
        k = (long)(rb_genrand_real()*(len-2));
        {
            long l = j, g = i;
            if (j >= i) l = i, g = ++j;
            if (k >= l && (++k >= g)) ++k;
        }
        return rb_ary_new3(3, ptr[i], ptr[j], ptr[k]);
    }
    if ((size_t)n < sizeof(idx)/sizeof(idx[0])) {
        VALUE *ptr_result;
        long sorted[sizeof(idx)/sizeof(idx[0])];
        sorted[0] = idx[0] = (long)(rb_genrand_real()*len);
        for (i=1; i<n; i++) {
            k = (long)(rb_genrand_real()*--len);
            for (j = 0; j < i; ++j) {
                if (k < sorted[j]) break;
                ++k;
            }
            memmove(&sorted[j+1], &sorted[j], sizeof(sorted[0])*(i-j));
            sorted[j] = idx[i] = k;
        }
        result = rb_ary_new2(n);
        ptr_result = RARRAY_PTR(result);
        for (i=0; i<n; i++) {
            ptr_result[i] = ptr[idx[i]];
        }
    }
    else {
        VALUE *ptr_result;
        result = rb_ary_new4(len, ptr);
        ptr_result = RARRAY_PTR(result);
        RB_GC_GUARD(ary);
        for (i=0; i<n; i++) {
            j = (long)(rb_genrand_real()*(len-i)) + i;
            nv = ptr_result[j];
            ptr_result[j] = ptr_result[i];
            ptr_result[i] = nv;
        }
    }
    ARY_SET_LEN(result, n);
    return result;
}