Missing Tests for Models with Variables with SId `INF` or `NaN` about sbml-test-suite HOT 7 CLOSED

@FFroehlich if a variable in SBML is called 'INF' or 'NaN' (which is a moronic idea and of no use at all), then that is just on identifier for the variable. And it is not to be confused with the MathML construct. In mathML it can be easily differentiated by the fact, that there will be a NaN , so an identifier called NaN rather than '' ...

so i think the test is still valid, and since it falls into the allowable subset of strings that represent a valid SId, i think it is ok to keep.

from sbml-test-suite.

@FFroehlich if a variable in SBML is called 'INF' or 'NaN' (which is a moronic idea and of no use at all),

I agree !

then that is just on identifier for the variable. And it is not to be confused with the MathML construct. In mathML it can be easily differentiated by the fact, that there will be a NaN , so an identifier called NaN rather than '' ...

But that information is lost after processing with libsbml.formulaToL3String, correct? http://sbml.org/Software/libSBML/5.18.0/docs/python-api/namespacelibsbml.html#a5aee8f98b5db063cadbae96c5e000cb3

so i think the test is still valid, and since it falls into the allowable subset of strings that represent a valid SId, i think it is ok to keep.

I am arguing that there actually should be a test! I just wasn't sure whether it is technically possible to pass it, but it sounds like it is.

from sbml-test-suite.

indeed, that information won't be there when converting the ASTNode to infix. So what would need to happen, is that before running the tests, all oddly named Species / Parameters / Compartments / Reactions would have to be renamed:

• true, false, nan, inf, time, avogadro, pi

prior to running the simulation. I'm quite sure there were tests for variables named time at least.

And i thought there would test for that ... maybe @luciansmith can provide some insight here. Since all the tests in the 'semantic' branch, are supposed to be simulatable and numerically comparable, maybe there was an issue with comparing inf / nan values?

from sbml-test-suite.

indeed, that information won't be there when converting the ASTNode to infix. So what would need to happen, is that before running the tests, all oddly named Species / Parameters / Compartments / Reactions would have to be renamed:

• true, false, nan, inf, time, avogadro, pi

prior to running the simulation. I'm quite sure there were tests for variables named time at least.

There are tests for time/avogadro but those are csymbols, so they can be correctly parsed by libsbml.formulaToL3StringWithSettings with avocsymbol set to True (this seems to generally affect all csymbols, including time). I don't think there are any tests for pi/true/false/inf/nan

And i thought there would test for that ... maybe @luciansmith can provide some insight here. Since all the tests in the 'semantic' branch, are supposed to be simulatable and numerically comparable, maybe there was an issue with comparing inf / nan values?

from sbml-test-suite.

Yup, these would be good tests to add. As you say, they would be clear in the XML, and not clear at all in the infix!

from sbml-test-suite.

OK, added these tests to the 'develop' branch: c6b515a

The last two crash the SBML Test Runner, but I'd love to hear back from you if you get these to work on your own system.

from sbml-test-suite.

These tests are now in the 3.4.0 release.

from sbml-test-suite.