Future of chemfiles reader in MDA

[Luthaf]

In #1862, I added coordinate & topology readers and writers based on chemfiles, which are disabled by default. I am opening this issue to discuss the future of this code, and how to make it available to as many people as possible.

A possible starting point would be enabling chemfiles as a reader for formats not yet supported by MDA, which currently are:

• CML: http://www.xml-cml.org/
• CSSR: http://www.chem.cmu.edu/courses/09-560/docs/msi/modenv/D_Files.html#944777
• mmCIF from the PDB archive
• Molden: http://cheminf.cmbi.ru.nl/molden/molden_format.html
• SMI/OpenSMILES (topology only)
• GROMACS TNG is supported as a read-only format

Would you be interested in activating chemfiles by default for some/all of these formats?

---

Another area where chemfiles can help with MDA is in formats which both implement, where chemfiles might be a bit faster (thanks to @richardjgowers tricks of faster line by line reading). This includes formats like PDB, XYZ, Tinker XYZ and potentially others.

Here I would leave it to the users to pick which parser they want, but potentially add some documentations so that the existing alternative is more visible to end users.

---

Related to the previous point, chemfiles is able to parse some files (such as non standard PDB files) not supported by MDA (#988, #1133, #1966 are examples of issues with the PDB parser which should work with chemfiles). Since chemfiles also has bugs (just different ones), I would take the same action as the previous point, telling users there is an alternative parser that they can use if the default one do not works.

This alternative parser suggestion could also work with other bridges to MDA such as the RDKit one from #2707

---

What do you think?

[orbeckst]

Sorry for the lack of discussion at the moment. I think at the moment many @MDAnalysis/coredevs are involved in GSoC and finalizing release 1.0 #2443 / fixing issues or preparing for 2.0 #2729 . Thank you for your patience.

I added coordinate & topology readers and writers based on chemfiles, which are disabled by default.

Can you explain what you mean by "disabled by default"? Do you mean there's not automatic detection of the file format and rerouting to ChemfilesReader?

---

As a sidenote, I just wanted to mention that interoperability (at the API level) has become a new theme in MDAnalysis, with the idea to use converters. I think in the longer run we should also consider making sure that the chemfiles python bindings can function within the converter framework as well.

[Luthaf]

Yes, by disabled by default I meant no automating routine to ChemfilesReader. As far as I understand, users need to request the use of ChemfilesReader explicitly.

Thanks for the pointer on converters, I'll read the documentation on this subject!