Pull from develop - #172

…d antimony binaries in libroadrunner-deps repo

…d antimony binaries in libroadrunner-deps repo

…dedicated antimony binaries in libroadrunner-deps repo"

This reverts commit e11d443.

Update version number.

There are actually a ton of other things you similarly can't get from the API, but these are the ones he needs.  I think in general, people should get the SBML model and use the libsbml API, so we didn't bother adding in the other missing functions.

Also added missing tests for other API calls, and found a few other bugs, which are also fixed (mostly concerning calling functions outside of their scope).

Several places we would just return NULL instead of setting an error and returning NULL.
Also don't let elements with assignment rules or rate rules be constant.

* Actually enable ccache
* No more UNIVERSAL_BINARIES
* modernize using setup.py

Supplement api

* Events with no assignment rules
* Reactions with no reactants nor products.
* Negative stoichiometries.

Also test to make sure the conversion factor (for the model or a species) is constant.

This tests a bug in libsbml, not Antimony, so it should fail right now, until we update libroadrunner-deps.

…with roadrunner.

Cover test suite gaps.

Merging now so that the new test suite can become a CI.  But in general, lots of things now work that used to fail!

…ror both times.

…on factors.

Need to special-case this, since 'time' is a different valid thing to ask for.

The problem was actually that when writing the Antimony file, it wouldn't take into consideration a conversion factor if one was present.  So if the submodel had t=1, and the containing model also had t=1, but converted, Antimony would think it didn't need to write out t=1, even though it was a 'different' 1.

… present.

It actually was wrong!  The file should always have had a 't1=1', since the submodel's version of '1' was different, since there was a conversion factor.

Instead of trying to get 'sub1.E0' to auto-promote, just import it as a promotion from the get-go.  This means that someone using the 'sub1.E0' will still get wrong SBML, but the fix was ridiculous and didn't fully work.

When we promote things behind the scenes, they get different names, but the structure is the same.

Test roundtrip test suite v2

added build from source instructions to the README.md

Had to update maddy to use standard table formats instead of its own bespoke version.

We only need to make sure they're unique IDs when we translate to (and from) SBML.

Implement support for stoichiometries with the same name.