Re: Re: More on Delete Problems

• To: mathgroup at smc.vnet.net
• Subject: [mg51715] Re: [mg51691] Re: More on Delete Problems
• From: Andrzej Kozlowski <andrzej at akikoz.net>
• Date: Sun, 31 Oct 2004 01:15:19 -0500 (EST)
• References: <cl9t75\$7nl\$1@smc.vnet.net> <200410270554.BAA24546@smc.vnet.net> <clpq73\$9qe\$1@smc.vnet.net> <200410290739.DAA03434@smc.vnet.net>
• Sender: owner-wri-mathgroup at wolfram.com

```No, I have good reason to believe that the problem is relate to Packed
Arrays. I also have reaosn to beleive that your suggestion is just a
red herring.

Theorem 1.
Packed arrays are relevant here.

a = {{1., 2., 3.}, {4., 5., 6.}};
b = {{1, 0, 1}, {0, -2, 3}, {0, 1, 0}};
c = a.b

Developer`PackedArrayQ[c]

True

Now if we unpack the packed array and use delete we do not get a crash,
if we do not we get a crash every time.

Now, if you understand what packed arrays do you will see how this may
well be related to the crash. Of course this is not a strict
mathemaitcal  proof but it far more convincing than your suggestion, as
I will prove next.

Theorem 2
Your suggestion is a "red herring", and has nothing to do with this
problem.

Proof.

a={{1.,2.,3.},{4.,5.,6.}};

b={{1,0,1},{0,-2,3},{0,1,0}};

c=a.b;

<<Algebra`AlgebraicInequalities`

Delete[c,{1,1}]

At this point I get a kernel crash every time!

\$Version

In[1]:=
\$Version

Out[1]=
5.0 for Mac OS X (June 10, 2003)

If my conclusion was "premature' what would you call yours?

Andrzej Kozlowski
Chiba, Japan
http://www.akikoz.net/~andrzej/
http://www.mimuw.edu.pl/~akoz/

On 29 Oct 2004, at 16:39, Peter Valko wrote:

> *This message was transferred with a trial version of CommuniGate(tm)
> Pro*
> Dear Andrej,
>
> I do not really see why you used packed arrays here.
> If your message is intended to suggest that the crash problem is
> somehow related to the PackedArrays, your conclusion is premature. To
> show this I do:
>
> a = {{1., 2., 3.}, {4., 5., 6.}};
> b = {{1, 0, 1}, {0, -2, 3}, {0, 1, 0}};
> c = a . b
> {{1., -1., 7.}, {4., -4., 19.}}
> << Algebra`AlgebraicInequalities`
> Delete[c, {1, 1}]
> {{-1., 7.}, {4., -4., 19.}}
>
> Without including the "<< Algebra`AlgebraicInequalities`" line,
> Mathematica crashes.
>
> Including the "<<Algebra`AlgebraicInequalities`" line eliminated the
> crash problem, but it does not mean the
> "Algebra`AlgebraicInequalities" has anything to do with the root of
> the problem. (I selected it because it is clearly a neutral,
> ambivalent package for the given case.) My conclusion is that with
> including a package (any package) more initialization is done and that
> helps to avoid the root of the crash problem.
>
> This example is good, because it is simple and reproducible and it
> shows the existence of a problem (I suspect one related to dynamic
> allocation of space) that might be more difficult to discover in a
> real situation. It is possible, that a simple fix can be built in
> temporarily into the init.m file, for instance. A light
> acknowledgement and advice from somebody associated with WRI would
> help us a lot, even if the disclamer ("this is not the official
> opinion of my employer" might show up with it.)
>
> Regards
> Peter
>
>
>
> Andrzej Kozlowski <andrzej at akikoz.net> wrote in message
> news:<clpq73\$9qe\$1 at smc.vnet.net>...
>> Packed Arrays strike again!
>>
>> Try this:
>>
>>
>> In[1]:=
>> a = {{1., 2., 3.}, {4., 5., 6.}};
>>
>> In[2]:=
>> b = {{1, 0, 1}, {0, -2, 3}, {0, 1, 0}};
>>
>> In[3]:=
>> c = Developer`FromPackedArray[a . b];
>>
>> In[4]:=
>> Delete[c, {1, 1}]
>>
>> Out[4]=
>> {{-1., 7.}, {4., -4., 19.}}
>>
>> No crash.
>>
>> Andrzej Kozlowski
>>
>> Andrzej Kozlowski
>> Chiba, Japan
>> http://www.akikoz.net/~andrzej/
>> http://www.mimuw.edu.pl/~akoz/
>>
>>
>>
>>
>> On 27 Oct 2004, at 14:54, Peter Valko wrote:
>>
>>> For the record:
>>>
>>> \$Version
>>> 5.0 for Microsoft Windows (June 11, 2003)
>>>
>>> a={{1.,2.,3.},{4.,5.,6.}}
>>> b = {{1,0,1},{0,-2,3}, {0,1,0}}
>>> c = a . b
>>> Delete[c,{1,1}]
>>>
>>> crashes Mathematica on my machine, exactly as Give Effinger said.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> Gove Effinger <effinger at skidmore.edu> wrote in message
>>> news:<cl9t75\$7nl\$1 at smc.vnet.net>...
>>>> Here's simple code which bombs for me:
>>>>
>>>> a={{1.,2.,3.},{4.,5.,6.}}
>>>>
>>>> b = {{1,0,1},{0,-2,3}, {0,1,0}}
>>>>
>>>> c = a . b
>>>>
>>>> Delete[c,{1,1}]
>>>>
>>>> Why?
>>>>
>>>> G. Effinger
>>>
>>>
>
>

```

• Prev by Date: Re: Re: More on Delete Problems
• Next by Date: Re: Calabi-Yau Manifold visualization
• Previous by thread: Re: Re: More on Delete Problems
• Next by thread: Re: More on Delete Problems