API typos, outdated info, broken links, etc (ongoing)

[lilyminium]

• Bond.__eq__ does not check bond order in core.topologyobjects:
  

  mdanalysis/package/MDAnalysis/core/topologyobjects.py

  Lines 163 to 167 in e110108

  	Two :class:`Bond` instances can be compared with the ``==`` and
  	``!=`` operators. A bond is equal to another if the same atom
  	numbers are connected and they have the same bond order. The
  	ordering of the two atom numbers is ignored as is the fact that a
  	bond was guessed.

  
  

  mdanalysis/package/MDAnalysis/core/topologyobjects.py

  Lines 135 to 140 in e110108

  	def __eq__(self, other):
  	"""Check whether two bonds have identical contents"""
  	if not self.universe == other.universe:
  	return False
  	return (np.array_equal(self.indices, other.indices) or
  	np.array_equal(self.indices[::-1], other.indices))

• All the "Reads the following Attributes" and "Guesses the following Attributes" should be double-checked in topology.__init__

  • __init__ does not include ITPParser
  • PSFParser: .. _PSF: http://www.charmm.org/documentation/c35b1/struct.html link broken.
  • MinimalParser: Ends in the middle of a sentence:
    

    mdanalysis/package/MDAnalysis/topology/MinimalParser.py

    Line 58 in e110108

    This requires that the coordinate format has

  • TRJReader outdated: #2398

• dist is missing documentation of the box parameter (analysis.distances.dist). Note that AtomGroups can be (usually are?) from different Universes. Note that even if the Universe has dimensions associated, dist does not use them unless explicitly passed into box

• Add lines to initialise k and dist in distances.self_distance_array returns example

• Update Molnums documentation, which is identical to Moltypes

• All of the helanal documentation. Best to leave this after #2452 is addressed

• #2395 fix up guess_bonds documentation

• All of the PSA documentation. Perhaps it will get AnalysisBased?

• encore.hes returns unreadable details which are not documented

• encore.ClusteringMethod.KMeans documents precompute_distances without actually having it as a kwarg

• encore.DimensionalityReductionMethod.StochasticProximityEmbeddingNative documents the default number of dimensions as 3, but it's actually 2

• for encore.hes, ces, dres: document that the method returns averages and standard deviations if estimate_error=True

• for encore.ces_convergence document the output. The shape does not include preference_values.

• for rdf.InterRDF_s document density kwarg

• fix base.AnalysisFromFunction raises ValueError

• document kwargs in Ramachandran.plot

• PCA.mean_atoms is not the mean positions, PCA.mean is

[richardjgowers]

Yep. So something pandas does is to mush around doc strings with decorators, so we could even do that with topology.__init__ so that when Topology Parsers are defined, they append to that table, so they're automagically done. Or just write some more rst if you don't want to do it programmatically.

[lilyminium]

The recently-merged #2390 was created to make a script that auto-docs parsers and the attributes associated. It's currently in the UserGuide but I could move it over to the main repo?

[orbeckst]

@lilyminium does the UG have a list of all pre-defined topology attributes from core.topologyattrs? Would be useful for Universe.empty() trickery (as I found out when I wasted a few minutes answering a 4-year old Stackoverflow question How to instantiate Atoms using MDAnalysis?).

EDIT: Yes – at https://www.mdanalysis.org/UserGuide/topology_system.html#topology-attributes

[lilyminium]

This example in rms.RMSF doesn't work, as align.AlignTraj requires a reference. Also, could use a trajectory that doesn't need to be made whole. And should either explain what order='afc' means or leave it out or use order='fac'; a 1-frame trajectory doesn't need reshaping.

Edit: as pointed out by @jbarnoud, it is easy to forget to clean or preprocess your trajectories before analysis. In that case, putting in the #TODO is now on TODO list 😁

import MDAnalysis as mda
from MDAnalysis.analysis import align

from MDAnalysis.tests.datafiles import TPR, XTC

u = mda.Universe(TPR, XTC, in_memory=True)
protein = u.select_atoms("protein")

# 1) need a step to center and make whole: this trajectory
#    contains the protein being split across periodic boundaries
#
# TODO

# 2) fit to the initial frame to get a better average structure
#    (the trajectory is changed in memory)
prealigner = align.AlignTraj(u, select="protein and name CA", in_memory=True).run()

# 3) reference = average structure
reference_coordinates = u.trajectory.timeseries(asel=protein).mean(axis=1)
# make a reference structure (need to reshape into a 1-frame "trajectory")
reference = mda.Merge(protein).load_new(
            reference_coordinates[:, None, :], order="afc")

[jbarnoud]

Also, could use a trajectory that doesn't need to be made whole.

Not dealing with PBC is a frequent misuse of the analysis tools. Showing it has to be done seems a good idea to me. Our examples will be used more or less verbatim in papers.