Current implementation of builtin time() function in BNGL gives inaccurate results

[jrfaeder]

The example file shows that the builtin time() function that can be invoked in the functions block (although gives an error if invoked in a rate law) give inaccurate results. The builtin time() function and the user-defined t() function give identical values at all of the points at which they are output, but the results of integrating a rate law that invokes them differ by an amount that greatly exceeds the expected integration error. The most likely cause of the problem is that the time() function is not correctly being used at intermediate points of the integration.
test_time2.bngl.txt

[lh64]

Yeah, the time() function wraps around the time variable t in run_network.cpp. For ODE simulations, CVODE has its own time variable, which can't be accessed (or at least I couldn't). So the time() function doesn't get updated while CVODE is running in between output steps. If you increase n_steps, I bet the results of the built-in time() and user-defined t() will converge.

Note that this isn't a problem for SSA simulations, so it should work fine for those.

[lh64]

BTW, this is why it says NOT FULLY TESTED for the time function in the "Built-in operators and functions" table we had on the BNG Wiki: https://web.archive.org/web/20131113212908/http://www.bionetgen.org/index.php/BNGManual:Table_2.

I could never figure out how to get around this issue for ODE sims.