Results of the memswap() smackdown from the thread "Sorting"assignment - Programming Languages

spinoza1111 <spinoza1111@yahoo.com> writes:
[color=blue]
> On Feb 16, 11:27 am, Richard Heathfield <r...@see.sig.invalid> wrote:[color=green]
>> Randy Howard said:
>>
>> <snip>
>>[color=darkred]
>> > I think this is hopeless, Clive.  Nilges doesn't see "broken" if his
>> > code runs to completion.  Regardless what answer is spat out, if it
>> > compiles and doesn't crash, he "declares victory".[/color][/color]
>
> In addition to confusing && and || last month, you posted code in
> response to Malcolm without disclosing (perhaps not knowing) (perhaps
> forgetting) that the failure of memcpy() to handle overlap in a
> defined fashion meant that your solution was undefined for overlapping
> blocks.[/color]

This is very tiresome. Whatever the history, your clutching at this
sort of straw looks foolish.

--
Ben.

On Sat, 16 Feb 2008 12:05:11 -0600, spinoza1111 wrote
(in article
<37a2efad-2285-42d4-b0cd-bbf521d556c4@s8g2000prg.googlegroups.com>):
[color=blue][color=green][color=darkred]
>>> You're lying when you pretend that you knew about this problem,[/color]
>>
>> Read the article where I specifically mentioned there being an issue
>> with the functions being discussed, and I asked if you would like to
>> know what it was.[/color]
>
> That "issue" was raised after the code was posted. You and your friend
> Heathfield are now claiming that you knew it all along.[/color]

Quick question: How could I have told you about an issue with a piece
of code until after I had seen it? How could I possibly have raised
the issue /before/ it was posted? Am I supposed to have a crystal ball
now?
[color=blue]
> I make errors in detail, to be sure.[/color]

By the bucket.
[color=blue]
> I am less practised at C than Ben who is by far the best programmer here
> (and whom you haven't had the decency or grace to recognize as such),[/color]

If Ben wants to know my opinion of his programming abilities, all he
has to do is ask me. He doesn't seem to clamor for attention (the
positive or negative varieties) as you do.
[color=blue]
> You haven't contributed a single line of code.[/color]

I saw no reason to do so. Someone else started this whole thing,
asking about the ability to do something in C. Richard posted a
function that would do what was asked about. It could have ended right
there, and in my opinion, it likely should have, but some others posted
alternative implementations which didn't fundamentally do any more to
address the original question. Then you decided to turn it into a
benchmark competition, and promptly screwed the first several attempts,
didn't even test your code, didn't even post benchmarked code that
actually tested the results, it up, and it all went downhill from
there.

Since the original question was answered, there doesn't seem to be any
need for Yet Another Solution" to it.
[color=blue]
> Instead, you've
> disrupted this and other threads constantly with unfounded statements
> about the competence of others based not on evidence but on what
> appears to be homophobia and poor anger management.[/color]

Homophobia? Is your AFDB in need of replacement, or did I miss a memo?

--
Randy Howard (2reply remove FOOBAR)
"The power of accurate observation is called cynicism by those
who have not got it." - George Bernard Shaw

On Sat, 16 Feb 2008 12:45:21 -0600, Richard Heathfield wrote
(in article <mf2dnWhZ-8fJsyranZ2dnUVZ8rednZ2d@bt.com>):
[color=blue]
> Clive D. W. Feather said [in reply to Ben, but presumably addressing Mr
> Nilges]:
>[color=green]
>> If you think that's comparable to writing:
>>
>> for (i = 0; i < strlen (s); i++)
>> // something that doesn't affect the value of or contents of s
>>
>> then you're even more deluded than we thought.[/color]
>
> No, he isn't - you just haven't known him long enough.[/color]

You beat me to it.


--
Randy Howard (2reply remove FOOBAR)
"The power of accurate observation is called cynicism by those
who have not got it." - George Bernard Shaw

On Feb 17, 2:45 am, Richard Heathfield <r...@see.sig.invalid> wrote:[color=blue]
> Clive D. W. Feather said [in reply to Ben, but presumably addressing Mr
> Nilges]:
>[color=green]
> > If you think that's comparable to writing:[/color]
>[color=green]
> >      for (i = 0; i < strlen (s); i++)
> >          // something that doesn't affect the value of or contents of s[/color]
>[color=green]
> > then you're even more deluded than we thought.[/color]
>
> No, he isn't - you just haven't known him long enough.[/color]

Richard, please see my online petition for your removal from this
newsgroup.

Clive, in 2003, I used precisely the same code, and Richard initiated,
enabled, and continually refreshed a campaign of personal destruction
based on this which included written libel in the form of letters to
Apress, my publisher. My book "Build Your Own .Net Language and
Compiler" was nonetheless published, Apress having the good sense to
ignore such malign letters, and this book has been often in the top
fifty books on "compilers" at USA Amazon.

You are getting sucked in to a Lynch mob. You hounded Schildt, and so
the bullies want you on their team.

What's really sad is that programming is still so much fun. I
demonstrated a visual quicksort in C Sharp to my class today (far more
than C could accomplish in reasonable time) which even for small
arrays demonstrated dramatically the speed of quicksort which gave the
lie to a crude claim made here that "I only see a difference, in C,
for 1000 entries". The visual sort also demonstrated how toplevel
entries in the interchange sort are quickly available.

Here I have to see other people, good people, LEAVE after being
assaulted, sometimes for having an obviously female, Hispanic, or
African-American name, and I have to endure constant harassment in
comp.programming (not, mind you, comp.lang.c) for not "drinking the
Kool-Ade" on C.

You wouldn't pull this shit in person. Like most people in cities
today, you're probably too soft and too scared. Even the people who
don't participate in the destruction are too soft and too scared to do
other than make apologies for the clear bullies lest we all rock the
boat.

Please add your name, Clive, to my petition for the voluntary removal
of Heathfield from this network.
[color=blue]
>
> --
> Richard Heathfield <http://www.cpax.org.uk>
> Email: -http://www. +rjh@
> Google users: <http://www.cpax.org.uk/prg/writings/googly.php>
> "Usenet is a strange place" - dmr 29 July 1999[/color]

Ben Bacarisse said:
[color=blue]
> spinoza1111 <spinoza1111@yahoo.com> writes:
>[/color]
<snip>[color=blue][color=green]
>> printf("Swapping %d bytes at %d and %d\n",
>> intLength, ptrToA, ptrToB);[/color]
>
> %o is the format for printing pointers.[/color]

No, it's %p - and %p is only defined for void *. Whether it will also work
for char * (because of guarantees about representation and alignment) has
been a matter of debate, but the safest course is to supply a conversion.

--
Richard Heathfield <http://www.cpax.org.uk>
Email: -http://www. +rjh@
Google users: <http://www.cpax.org.uk/prg/writings/googly.php>
"Usenet is a strange place" - dmr 29 July 1999

spinoza1111 <spinoza1111@yahoo.com> writes:
[color=blue]
> On Feb 16, 10:00 pm, Ben Bacarisse <ben.use...@bsb.me.uk> wrote:[color=green]
>> Richard Heathfield <r...@see.sig.invalid> writes:[color=darkred]
>> > Ben Bacarisse said:[/color]
>>[color=darkred]
>> >>spinoza1111<spinoza1...@yahoo.com> writes:[/color]
>>[color=darkred]
>> > <snip>[/color]
>>[color=darkred]
>> >>> <snip> he has made a significant error in
>> >>> recommending memcpy without remembering, as the C expert here, that it
>> >>> is undefined for overlap.[/color]
>>[color=darkred]
>> >> Did he?[/color]
>>[color=darkred]
>> > No. <snip>[/color]
>>
>> Of course not.  I obviously failed to indicate the sarcastic tone of
>> the question.  Next time I will try "Right, and I am Marie of
>> Romania"[1].[/color]
>
> He did, Ben. He replied to Malcolm's claim that C cannot do a
> multibyte memory swap with code that used memcpy() and he did not
> disclose this limitation at the time. When he remembered or perhaps
> learned the behavior a week later, he told Randy Howard that this was
> so, and then he proceeded to collude with Howard in a cover up.[/color]

This is getting silly. I know exactly what has been posted (I have
read all of these threads). My incredulity related to two things:
that he made "a serious error" and that he did so "without
remembering" what memcpy does. If you think the subsequent exchanges
about overlapping swaps suggest that he'd forgotten anything about
memcpy, I doubt I can persuade you otherwise, but I did not read it
that way. If you think the initial dismissal of overlap was a serious
error, then so be it. I don't see it that way -- I think it is a
non-issue and leaving it undefined is entirely satisfactory to me.

--
Ben.

"Randy Howard" <randyhoward@FOOverizonBAR.net> wrote in message[color=blue]
> You will kindly point out how the original questioner specified your
> convention for an implementation of memswap(), and how Richard and
> others were in error by failing to meet this requirement.
>[/color]
I made a combsort(), using the same prototype as qsort(), and comented that
C doesn't have facilities for swapping except byte-by-byte.

That's not entirely true, as Heathfield pointed out, because a swap can be
built on top of memcpy(). However it is not entirely false either, because
we are supposing that memcpy() itself isn't written in portable C.

To be fair to Nilges, he did point out that memory accesses could be
minimised by comparing bytes for equality before writing. My function did
show a very tiny speedup when swapped variables were pointer-sized, or 4
bytes of typcial current architectures, in some tests, but in my own the
memcpy() version was actually faster, though it made only a tiny difference
to the runtime of the whole sort function.

--
Free games and programming goodies.
[url]http://www.personal.leeds.ac.uk/~bgy1mm[/url]

spinoza1111 <spinoza1111@yahoo.com> writes:
<snip>[color=blue]
> I make errors in detail, to be sure.[/color]

Why are all your errors ones of detail but everyone else's are
serious? At least serious enough for you to keep going on about them?
[color=blue]
> I am less practised at C than Ben
> who is by far the best programmer here (and whom you haven't had the
> decency or grace to recognize as such),[/color]

I would be shocked if anyone did so. I can't see why you say that
either. There is no evidence that I am "the best programmer here" and
there is much evidence (on the net) that I am not. I won't embarrass
anyone by naming names.

But I am embarrassed that you make such claims. It does not help
toknow who id good and who is bad. Good programmers can have bad
ideas and bad programmers can have excellent ones. Let's try to
discuss posted ideas and not compare people.

--
Ben.

On Feb 17, 2:45 am, Ben Bacarisse <ben.use...@bsb.me.uk> wrote:[color=blue]
> spinoza1111 <spinoza1...@yahoo.com> writes:[color=green]
> > On Feb 16, 3:08 am, "Bartc" <b...@freeuk.com> wrote:[/color]
>
> [I don't know what has happened to your attribution lines, but you are
> replying to BartC without, it would seem, quoting anything he wrote.
> Normally I'd snip the above line because I'm not quoting any of it,
> but since I'm commenting on its inclusion, I think I have to leave it.]
>
> <snip>
>[color=green]
> > "Overlap swapping" may not be sensible but it's deterministic and
> > reversible in the sense of undoable because deterministic: it might be
> > useful in encoding.[/color]
>[color=green]
> > The problem: "overlap swapping" is NOT reversible like this[/color]
>[color=green]
> > swap(swap(blockA, blockB)) == blockA,blockB[/color]
>[color=green]
> > It would seem to me that a memory swap should share this property with
> > swap of bytes where swap(swap(byte1, byte2)) = byte1,bytes[/color]
>[color=green]
> > Fortunately, if you do not swap overlapping bytes then the swap is
> > reversible.[/color]
>
> If you think that useful, go for it.  I can't see the purpose.
>[color=green]
> > I believe the following code handles possible overlap.[/color]
>
> Nope.  Why don't you post in C#?  What purpose does trying to write C
> serve?  Most people here can read a wide range of languages and your
> examples would be clear enough to most.  Some of the errors seem to be
> logical, but using a language you know will surely reduce the number.
> Can C# do overlapping memory swaps?[/color]

No it cannot.

When in Rome...actually, I enjoy getting back up to speed in C. It's a
challenge. Thanks for your criticism.

I program in C with amateur status, and make and fix errors
repeatedly, in order to "deconstruct" a completely false image of
programming masculinity which is used here by Heathfield and Howard to
manipulate and control. You see, I don't "buy" their implicit claim,
that there are "real" programmers who program close to the metal,
versus inferior script kiddies and girliemen. I don't buy it, Ben,
because I established chops far earlier than they by debugging a
compiler in machine language, and matured rapidly out of what I have
come to believe is not a truly scientific endeavor, but a form of
social control.

Since I've been there and done that in a range of languages that
included C in the early 1990s, I find it very ugly when Heathfield/
Howard throw their weight around, since they seem in fact to know very
little about computer science, object-oriented coding, or other topics
outside of C, and I think this is why Heathfield constantly snipes at
better men than he: a profound personal insecurity.

Therefore I thought it would be amusing to ramp quickly back up to C
by making online mistakes and using you as a resource, and demonstrate
the shallowness of an "expertise" which is (in many cases I've seen)
quietly and without fanfare acquired by "nonprofessional" programmers,
especially, in my experience, doctors and lawyers with far more
culture, self-discipline and collegial amiability than are displayed
by the "professional" programmer, who so often deludes himself that he
is other than a corporate slave raised for food.

When I see the insecurity and nastiness here including not only that
directed at me but also at Schildt, I thank my lucky stars that I am
no longer in the computer business. At the same time, as I get older,
I persist in learning and here relearning, time permitting, for the
same reason I run or work out the equivalent of 20 miles a week.



As to C sharp. There's no such thing as a von Neumann addressible
contiguous memory in C sharp, therefore no such thing as "moving
blocks" at all. However, you can simulate the problem.
[color=blue]
>
>
>
>
>[color=green]
> > Once it hits a
> > point where block A byte x is inside block B, it hops to the end of
> > block B. Let's see, you don't need the opposite test (block B byte y
> > is inside block A), I think, because for any given byte, it "overlaps"
> > the corresponding byte iff the corresponding byte overlaps it.[/color]
>[color=green]
> > You don't need a byte by byte precheck, or even my assert, this code
> > handles the checking as part of its main work. The listing is followed
> > by a preliminary test result. The test is instrumented so you can
> > watch the swap step by step. It swaps and then it unswaps. The results
> > and the code are best viewed in a monospace font.[/color]
>[color=green]
> > void spinoza1111_memswap(char *ptrToA,
> >                          char *ptrToB,
> >                                             int intLength)
> > {
> > printf("Swapping %d bytes at %d and %d\n",
> >        intLength, ptrToA, ptrToB);[/color]
>
> %o is the format for printing pointers.[/color]

So? Why am I permitted to use %d? Either the language enforces it, or
it's not a rule.[color=blue]
>[color=green]
> > if (ptrToA + intLength > ptrToB
> >     ||
> >     ptrToB + intLength > ptrToA)
> >    printf("Overlap swap\n");[/color]
>
> This logic is plain wrong.  If we accept the restriction you are
> implicitly putting on this function (that it can swap only parts of
> the same object) the code above becomes legal C but is not a check for
> overlap (the condition is true unless the two pointers are the same --
> i.e. it is false only for total overlap).[/color]

Sorry, this code was info only and was left in. But it has a typo
which I didn't catch, because the only purpose of the code was to
print "overlap swap": the second greater than sign should have been a
less than sign.

I think you're a great coder, but rather inarticulate. Had I spotted
the error, I would have said "didn't you want that to be a greater
than sign?", not "this logic is plain wrong", because I would have
reflected upon and internalized INTENT, something you need to do when
you read code.
[color=blue]
>
>
>
>
>[color=green]
> > char chrExchange;
> > char *ptrToAend = ptrToA + intLength;
> > char *ptrToAwork = ptrToA;
> > char *ptrToBwork = ptrToB;
> > char *ptrToAstart;
> > char *ptrToBstart;
> > char *ptrToBblockLimit = ptrToB + intLength;
> > int intBlockLength;
> > printBlocks(ptrToA, ptrToB, intLength);
> > while (1)
> > {
> >    do
> >            {
> >                    if (ptrToAwork >= ptrToB && ptrToAwork < ptrToBblockLimit)
> >                            ptrToAwork += ptrToBblockLimit - ptrToB;
> >                    if (ptrToAwork >= ptrToAend) return;
> >                    if (*ptrToAwork != *ptrToBwork)break;
> >                    ptrToAwork++; ptrToBwork++;
> >            }
> >    while (1);
> >    chrExchange = *ptrToAwork;
> >    *ptrToAwork = *ptrToBwork;
> >    *ptrToBwork = chrExchange;
> >    printBlocks(ptrToA, ptrToB, intLength);
> >    ptrToAwork++; ptrToBwork++;
> > }
> > }[/color]
>
> And this code does not do what you claim.  Your solution to overlap
> was not to swap overlapping areas, but unless you also have an odd
> definition of these, this code does swap the overlapping area (though
> not all of it and not always).
>
> For example:
>
>   char t[] = "abcdefgh";
>   spinoza1111_memswap(t, t + 2, 6);
>
> will leave t containing "cdabefgh".  The overlap, is "cdef" in my
> book, so I'd expect "ghcdefgab" from your definition above.
>
> There are two other problems.  The code does not do the same thing
> when the pointers are reversed.  Surely swap(a, b, n) should have the
> same effect as swap(b, a, n)?  In the example above
>
>   spinoza1111_memswap(t + 2, t, 6);
>
> is a no-op.
>[/color]
Thanks as always. I will investigate and if possible fix these errors
if that is what they are. Meanwhile, you might try to express why the
code doesn't work not merely THAT it fails, unless you don't want to.
[color=blue]
> The other is the C issue.  Pointer comparison is only defined for
> pointer into (and just one past the end) the same object.  In trying
> to make it more general (do something with overlap) you have made it
> more restricted.  You can't, for example, swap the contents of two
> separate arrays.
>
> --
> Ben.- Hide quoted text -
>
> - Show quoted text -- Hide quoted text -
>
> - Show quoted text -[/color]