Robust SBML parser

[daemontus]

Complete rework of the SBML parser. In terms of the public API, this should change almost nothing - although the warnings are no longer printed to stdout, but there is a new method (from_sbml_strict) that allows you to parse and emit warnings to a given argument. However, there are some critical internal changes that should be noted:

• The implementation switched from xml-rs (which seems to be mostly abandoned at this point) to roxmltree. This may be slightly slower, but is much more modular and future proof. As a consequence, we do not have to deal with the /> bug (e47f246778a84d6a95bdad20c5beee6ef0d68007) and we don't depend on random git repos.
• The architecture of the parser is much more modular and much safer, so it should be easy to adopt to multi-valued models in the future as well as find and fix bugs when they arise.
• We now have a differential test that compares the SBML parser output to the output of the AEON parser on a large collection of models. However, the AEON models are generated using the previous version of the parser, so this can only uncover newly introduced bugs in future revisions.
• Breaking change: If there is a <defaultTerm> with targetLevel=1, the parser correctly outputs a constant true function (previously, no function was emitted).
• Breaking change: If there is a function with that has an input species, but (syntactically) does not use this input anywhere, the input is marked as non-observable, as this is most likely what the author has intended. (If the function uses the input, but the input has no effect, this is not addressed in parsing)
• If a function uses input id instead of the specie id, this is properly recognized.
• Name collision resolution should be much more robust.

Of course, this is not the end (for example, it would be cool to have a standalone SBML library and then convert between in-memory representations) and the SBML export is also not exactly spectacular (although it does not have to be as robust as the import since it has much better defined inputs). But for now this is reasonable and I believe it can be released into the wild :)

Closes #13 and #10.

[codecov[bot]]

Codecov Report

Merging #16 (6847d73) into master (ef7aaa7) will increase coverage by 1.00%. The diff coverage is n/a.

@@            Coverage Diff             @@
##           master      #16      +/-   ##
==========================================
+ Coverage   83.00%   84.00%   +1.00%     
==========================================
  Files          30       36       +6     
  Lines        1871     1951      +80     
==========================================
+ Hits         1553     1639      +86     
+ Misses        318      312       -6     

Impacted Files	Coverage Δ
src/bdd_params/impl_witness_generator.rs	79.27% <0.00%> (-1.81%)	:arrow_down:
src/bdd_params/impl_static_constraints.rs	97.56% <0.00%> (-1.22%)	:arrow_down:
src/parser/impl_regulatory_graph.rs	98.61% <0.00%> (-0.08%)	:arrow_down:
src/sbml/export.rs	94.73% <0.00%> (-0.05%)	:arrow_down:
src/sbml/mod.rs
src/sbml/import/_read_species.rs	92.85% <0.00%> (ø)
src/sbml/import/_convert_mathml_to_fn_update.rs	65.07% <0.00%> (ø)
src/sbml/import/_read_mathml.rs	80.39% <0.00%> (ø)
src/bin/gen_diff_tests.rs	0.00% <0.00%> (ø)
src/sbml/import/mod.rs	85.83% <0.00%> (ø)
... and 4 more

---

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 10d6bda...6847d73. Read the comment docs.