Fix issue #270

[lohedges]

This PR closes #270 by using a list rather than a set to store the names of protein and nucleic acid residues. This makes the search string for backbone atoms reproducible, hence the tests now repeatedly pass. I'm still not sure why this issue has only cropped up recently, since _prot_res has been a set since the code was first added.

In [1]: x = {"1", "2", "3"}

In [2]: ",".join(x)
Out[2]: '2,3,1'

In [3]: x = ["1", "2", "3"]

In [4]: ",".join(x)
Out[4]: '1,2,3'

• I confirm that I have merged the latest version of devel into this branch before issuing this pull request (e.g. by running git pull origin devel): [y]
• I confirm that I have permission to release this code under the GPL3 license: [y]

Suggested reviewers:

@chryswoods

[lohedges]

Thinking more about this, I had a similar problem when searching for ions during parameterisation. It wasn't possible to do a search for all ions in one go due to the limitation of the search syntax. Instead I just looped over the ions and searched one-by-one. Perhaps I need to update the code to do something similar here, or maybe it could be caught and handled in the sire search grammar.

[lohedges]

Actually, it appears that the issue was that the protein and nucleic acid residue strings weren't being joined correctly. No comma was added between then so the last protein residue name was being combined with the first nucleic acid residue. Depending on the ordering from the set, this could result in removing some of the matching residues from the test molecule, e.g. ACE combined with T would create a nonsense residue called ACET, so we would no longer match ACE.

[chryswoods]

All good - the change from { } to [ ] is very subtle - good spot that this was one of the causes of the inconsistent behaviour. Same with the missing "," in the join.