[Molvis-list] rasmol connect bug

[Herbert J. Bernstein]

Dear Michael,

   Thank you for the quick analysis.  Nikolay Darakev has also been looking
as this.  I'll leave it to him to look into the best reasonable upper
bound allowing for metal coordination, but for the moment, I agree that
4 Angstroms (#define AbsMaxBondDist 1000) would make sense on modern
machines.  Long term, we probably should use two limits, one short one
for purely organic bonds, and one longer one for everything else.

   Regards,
     Herbert




At 12:16 PM -0500 4/20/08, Michael Sternberg wrote:
>Hi,
>
>Thanks Herbert.  I located the "not-really" bug - it is due to 
>optimization and rather simple to fix:
>
>In molecule.c, change
>	#define AbsMaxBondDist 600
>to
>	#define AbsMaxBondDist 800
>
>Roger implemented fast progressive short-circuit evaluation of  dx^2 
>+ dy^2 + dz^2 > AbsMaxBondDist^2, and 600 (=2.4 A) was just a tad 
>too short for metal bonds that are aligned to the axes; 800 (3.2 
>Angstrom) works fine now, but could possibly be even 1000.
>
>The performance impact of a larger value is negligible on today's machines ;-)
>
>
>Regards, Michael
>
>On Apr 16, 2008, at 21:27 , Herbert J. Bernstein wrote:
>>Thank you for the problem report.  Until we locate the bug, you can force
>>bonding one atom pair at a time with the bond command, e.g.
>>
>>  bond 0 1
>>  bond 0 2
>>  bond 1 2
>>
>>I have entered the report into the tracker on sourceforge for you.
>>
>>Regards,
>>  H. J. Bernstein
>>
>>
>>
>>At 7:04 PM -0500 4/16/08, Michael Sternberg wrote:
>>>Dear Rasmol users,
>>>
>>>I'm encountering a bug in the "connect" command, wherein certain 
>>>bonds are not detected, come what may.
>>>
>>>Consider the following xyz-file, which is an excerpt from a larger 
>>>model of a surface I'm trying to render:
>>>
>>>==== test.xyz ================================
>>>   3
>>>
>>>Ni  2.74059   4.93685   0.02301
>>>Ni  1.31849   7.41686   0.03767
>>>Ni  4.16141   7.41082   0.01675
>>>==============================================
>>>
>>>These 3 atoms are arranged in a roughly equilateral triangle with 
>>>distances 2.843, 2.853 and 2.859 Angstroms.
>>>
>>>Run any rasmol version 2.6.4 to the very latest RasMol_2_7_4_2_10Apr08:
>>>
>>>     rasmol -xyz test.xyz
>>>	RasMol>	connect on
>>>
>>>This will draw just *one* bond, not three.  Curiously, it's the 
>>>*middle* length bond that's detected, not the shorter or longer 
>>>ones.
>>>
>>>This shortcoming of rasmol's bond detection has bugged me on and 
>>>off for quite some time.  Usually, I got around it by re-declaring 
>>>the offending atoms to be types with larger covalent radii.  That 
>>>won't work in the full structure that I'm handling here, as it 
>>>introduces unphysical bonds in other parts of the structure.  I 
>>>may have some recourse in supplying bond info trough, say, the 
>>>alchemy format, but I wonder if the root of this isn't a shortcut 
>>>taken in the bond detection routine?
>>>
>>>
>>>Regards, Michael
>>>
>>>_______________________________________________
>>>Molvis-list mailing list
>>>Molvis-list at bioinformatics.org
>>>http://www.bioinformatics.org/mailman/listinfo/molvis-list
>>
>>
>>--
>>=====================================================
>>Herbert J. Bernstein, Professor of Computer Science
>>   Dowling College, Kramer Science Center, KSC 121
>>        Idle Hour Blvd, Oakdale, NY, 11769
>>
>>                 +1-631-244-3035
>>                 yaya at dowling.edu
>>=====================================================


-- 
=====================================================
  Herbert J. Bernstein, Professor of Computer Science
    Dowling College, Kramer Science Center, KSC 121
         Idle Hour Blvd, Oakdale, NY, 11769

                  +1-631-244-3035
                  yaya at dowling.edu
=====================================================