Default behavior of MDAnalysis.analysis.hbonds

[GoogleCodeExporter]

Long story short: the current default value for hbonds(update_selection1=False) 
leads to unexpected behavior when used to analyse hydrogen bonding with solvent.

Detailed story:
I think that update_selection1/2=False may be a troublesome default for
many users because it makes hidden assumptions about the analyzed system.
As a VMD user, I know I have to enable "update selection" whenever I use
distance-dependent selections with solvent because intermolecular distances
change all the time. However in this case I did not even suspect that the
analysis code depends on the initial geometry of my system, I just told the
module to perform hydrogen bonding analysis, and my selections did not
depend on distances. In a sense, the code does not respect my selection
string if I tell it to consider all water molecules but it only selects
those that are closest to selection1 in the first frame and then does not
update that selection each frame.

I understand that this is a very efficient optimisation, but I would value
safe default behaviors over efficient ones. Why not change the default to
True? Or reframe the flag by naming it something that communicates this
behavior more clearly, e.g. "assume_rigid_molecule" (you can probably come
up with something more fitting than me).

Original issue reported on code.google.com by cmiel...@zedat.fu-berlin.de on 2 May 2013 at 6:46

[GoogleCodeExporter]

This issue was closed by revision feb07226f516.

Original comment by orbeckst on 23 Aug 2013 at 2:12

• Changed state: Fixed