[Nix-dev] New-style Marc Weber proposal

[Peter Simons]

Marc Weber 2
============

If you like it, we can deprecate Marc Weber soon. Right now, both can be used.

Many of you have gotten used to the version of Marc Weber that has been around
for so long, but lets face the fact that there are problems: (a) it actually
works and (b) I haven't written it. So I decided to do something about that.
First, I tried to patch the existing version, but that only made things worse,
so I re-wrote everything from scratch. This allowed me to format the entire
source code according to my personal taste without being bothered by style
guides or group consensus.

Previously, Marc Weber had the serious disadvantage that

  [       a + 3                 1                  1         ]
  [  ---------------   - ---------------  - ---------------  ]
  [  (a - 1) (a + 4)     (a - 1) (a + 4)    (a - 1) (a + 4)  ]
  [                                                          ]
  [          1               a + 3                 1         ]
  [ - ---------------   ---------------   - ---------------  ]
  [   (a - 1) (a + 4)   (a - 1) (a + 4)     (a - 1) (a + 4)  ]
  [                                                          ]
  [          1                  1               a + 3        ]
  [ - ---------------  - ---------------   ---------------   ]
  [   (a - 1) (a + 4)    (a - 1) (a + 4)   (a - 1) (a + 4)   ]

while also being

  (mmm-add-classes
   '((literate-haskell-bird
      :submode text-mode
      :front "^[^>]"
      :include-front true
      :back "^>\\|$"
      )))

and this lead to

  template<unsigned int i, typename x, typename xs>
  struct sieve_multiples_of< i, typelist<x,xs> >
  {
    typedef typename ifelse
      < x::value % i != 0u
      , typelist<x, typename sieve_multiples_of<i, xs>::result>
      , typename sieve_multiples_of<i, xs>::result
      >::result result;
  };

which is really bad! In my new version, however, this all becomes:

    a!#_{a]|_aa23 23 / 8ZZZZ{}````5@ a!#_{a]|_aa23 23 / 8ZZZZ{}``
    ``5@ a!#_{a]|_aa23 23 / 8ZZZZ{}````5@ a!#_{a]|_aa23 23 /
    8ZZZZ{}`` ``5@ a!#_{a]|_aa23 23 / 8ZZZZ{}````5@ a!#_{a]|_aa23
    23 / 8ZZZZ{}`` ``5@ a!#_{a]|_aa23 23 / 8ZZZZ{}````5@
    a!#_{a]|_aa23 23 / 8ZZZZ{}`` ``5@ a!#_{a]|_aa23 23 /
    8ZZZZ{}````5@ a!#_{a]|_aa23 23 / 8ZZZZ{}`` ``5@

I hope this explanation helps you understand why the re-written version is
superior.

At this point, my new code is undocumented, untested, unfinished, unreadable,
unnecessarily complex, unmaintained, and it doesn't work, so I checked it right
in for your convenience. If the change causes any problems for you, it's
probably easiest if you try to fix them yourself. I may be able to assist you
with your efforts. If it so happens that you can't fix the problems yourself,
please just let me know! I'll make an effort to deflect all responsibility with
flimsy excuses in the nick of time.

Generally speaking, I plan to develop this project into something really major,
like ... I dunno ... maybe a car or a moon rocket, but because of the other
five dozen unfinished projects I'm involved in, I'm not actually going to do
it. As always, I appreciate feedback (unless I don't like it) and I'll try to
fix my bugs ASAP (unless they don't bother me).

Have a nice day,
Peter Simons