Feature/implement groups

[gregmedlock]

Implements the group class and the Model.groups attribute for better compliance with the SBML FBC package, fixes #543.

Some key implementation choices:

• reactions, metabolites, or genes can belong to groups
• groups can only take "kind" values as specified in the FBC specification
• each Group has a members attribute, which is a set of reactions, metabolites, and/or genes.
• whenever reactions, metabolites, or genes are removed from a model, group membership is updated.
• a Group that is not associated with a model can be generated in isolation and associated with cobrapy objects; upon addition of the Group to the model, all cobrapy objects associated with the Group are also added.
• when a Group is removed from a model, no modifications are made to cobrapy objects that were previously members of the group.
• model.get_associated_groups() is used to dynamically look up group membership. This PR does not have a model-based helper function for retrieving group information en masse (e.g. group membership info for all reactions, metabolites and genes with a single function)

Models needed to be repickled for tests to pass; one interesting artifact is a "version" tag in mini.json that is now being regenerated with a string in the corresponding field as "1" instead of the int (1) that is compliant with the jsonschema. I'm not sure what the source of this is, since it seems like it's been a while since the models were repickled in a PR. So, here I just manually corrected the issue in mini.json (changing "1" to 1).

I think I will add documentation for groups in a separate PR; would it make the most sense to add a section in "building a model" and "reading and writing a model"?

Code review and comments for improvement are much appreciated; I tried to keep things consistent with the rest of the codebase where possible. @matthiaskoenig please let me know if I missed some major functionality that would make the libsbml integration smoother!

[codecov-io]

Codecov Report

Merging #754 into devel will decrease coverage by 0.33%.
The diff coverage is 68.68%.

@@            Coverage Diff             @@
##            devel     #754      +/-   ##
==========================================
- Coverage   82.87%   82.54%   -0.34%     
==========================================
  Files          40       41       +1     
  Lines        3883     3981      +98     
  Branches      885      911      +26     
==========================================
+ Hits         3218     3286      +68     
- Misses        463      489      +26     
- Partials      202      206       +4

Impacted Files	Coverage Δ
cobra/core/gene.py	73.91% <100%> (+0.69%)	⬆️
cobra/manipulation/delete.py	82.4% <33.33%> (-1.41%)	⬇️
cobra/core/model.py	86.12% <58.33%> (-4.08%)	⬇️
cobra/core/group.py	87.87% <87.87%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 8ee8eaa...220c1ed. Read the comment docs.

[Midnighter]

I haven't had the time yet to review your code but your description of the PR sounds really promising!

The problem with the JSON version that you see is related to #720 and will be fixed in #725 where we switch to only strings. So no manual editing to stick with the current schema is necessary.

[gregmedlock]

Ok, I reran update_pickles.py so the files represent their true state. Also did it to retrigger the travis build; looks like the python 2.7 build timed out without any error.

edit: leaving mini.json as is makes the travis build fail, but tests were fine before.

[Midnighter]

I made a few changes to the class interface, for example, no assignment to .members only via .add_members. The only problem right now is that cobra.Model has no add_genes methods. We need to fix that first. Also, the other tests will pass once the schema is updated via the other open PR.

[Midnighter]

Hi @gregmedlock, thanks again for your work here.Your code has been merged as part of #685. I have some code locally that will extend the JSON schema with groups. So that still needs to be done but this code is finally in devel. I'm closing the PR because your code was merged locally.