- IMP-dev - salilab.org

a common pointer for imp
by Daniel Russel 21 Feb '08

21 Feb '08

First, some minor (and nasty) patches: The Grid class has some slightly better hacks to handle roundoff error with computing the upper bounds. The issue is that the (num cells)*(cell width)+(min offset) is not necessarily as large as the point which was used to define the maximum coordinate. I now added a 10% slop factor. I don't really like it, but without it various other things break. The particle grid now caps the number of voxels used in the case where there are few particles scattered over a wide area. The formula used is undoubtedly not optimal. Any improvements are welcome. BondDecorators now print their length if they have one And the main event. I added a class internal::ObjectPointer which can be used to (guess what?) store pointers to Objects. Depending on a bool template argument it either deletes the object when the pointer is destroyed or does not. It checks various invariants like the object being OK and that only one pointer is going to delete any particular object. It is also null initialized to try to avoid problems like the model_ pointer bug from yesterday. I switched a few classes over to use it, but not an exhaustive set. I also moved Object into the internal namespace since it is just an implementation detail and you don't need to know about it do do much of anything.

2 1

Add accessor methods to Harmonic
by Ben Webb 20 Feb '08

20 Feb '08

The attached patch adds accessor methods to Harmonic (and subclasses) to get/set the mean and standard deviation. (Frido asked for this to help in tweaking his optimizations.) Ben -- ben(a)salilab.org http://salilab.org/~ben/ "It is a capital mistake to theorize before one has data." - Sir Arthur Conan Doyle

1 0

A few more patches
by Daniel Russel 19 Feb '08

19 Feb '08

I modified the log states to have a function to force writing of a file whether or not you are skipping files. Seems useful to have (and used in the next patch) and I wrote an example. Perhaps should have more documentation, but it is a start. finally, the BondDecoratorList now checks to make sure that both particles are contained in the list of particles. And really finally the SConscript files for the headers weren't all updated after the last move (the restraints one, for example).

1 0

sundry patches
by Daniel Russel 19 Feb '08

19 Feb '08

smallest first: bonddecoratorlistscorestate returns and prints the number of bonds define types for pairs in base_types (a later patch exports them to python) I forgot to have the spheres list update the bonded lists Finally, a pairlist restraint and minor tweaks to list restraint since I was mimicking it

1 0

DistanceRestraint
by Daniel Russel 19 Feb '08

19 Feb '08

It seems to me it should be PairRestraint and take a PairScore. Right? Currently I have issues if I want to apply it to two spheres.

2 3

NULL model pointer in opt
by Daniel Russel 19 Feb '08

19 Feb '08

Optimizer was not initializing its member model_ pointer. The patch also has a few checks I added while trying to track the problem down.

1 0

Re: [IMP-dev] cleanup of cruft
by Ben Webb 19 Feb '08

19 Feb '08

Jeremy Phillips wrote: > Frank's restrainer, yes, but nothing Bret did, which is what I assume > Ben meant. I think the consensus was to scrap that.... Yes, exactly - I meant Bret's web interface thingy: https://svn.salilab.org/impmod/trunk/restrainer/ Frank's Restrainer is, of course, its own little world, so nothing we do with IMP will break it. Ben -- ben(a)salilab.org http://salilab.org/~ben/ "It is a capital mistake to theorize before one has data." - Sir Arthur Conan Doyle

1 0

cleanup of cruft
by Daniel Russel 19 Feb '08

19 Feb '08

Can we just delete everything in kernel/doc/examples? The code I looked at doesn't work any more. Andrej pointed Marc at IMP and he was, quite reasonably, trying to figure things out based on code there. In addition, I Ben sets it up so that scons test runs the code in examples, I'll submit a few more reasonable instances. If the code doesn't get run, it will just rot again. Thanks. Finally, can we delete all the restraint IO stuff from the XML loader? It uses restraints we don't really want people to use, is quite inflexible etc. And I would not want someone to be tempted to use it.

2 1

Nonbonded list for spheres
by Daniel Russel 18 Feb '08

18 Feb '08

Here is a non-bonded list for spheres. It works by binning the spheres based on radius and creating a grid for each bin. To build the nbl it searches between the grids as needed. It can be made a lot more efficient, but at least it seems to work and is straight forward. Somewhere this patch making printing of particles more useful got lost.

1 0

Various minor patches
by Daniel Russel 18 Feb '08

18 Feb '08

In brief, NonbondedListScoreState is now an abstract class and the corresponding class is AllNonbondedListScoreScoreState. I don't like the name, so if you have a better one, please suggest it. I needed to have a base class for the lists so that I can implemented one which handles particles with varying radii nicely. In more depth: 1) Better string handling in the exception base class 2) added bounds check on vector access 3) Refactored the non-bonded list code. There is a new internal class which implements a voxel grid for particles. There is also a base class for non-bonded lists which AllNonbondedList and BipartiteNonbondedList both inherit from. I also got rid of the subdir for the bonded lists and fixed various documentation. SConscipt files and IMP.h changes are not submitted.

2 3