Issue by luciansmith
Wednesday Jul 02, 2014 at 15:47 GMT
Originally opened as https://github.com/AndySomogyi/roadrunner-backup/pull/6

luciansmith included the following code: https://github.com/AndySomogyi/roadrunner-backup/pull/6/commits

Issue by hsauro
Tuesday Jul 08, 2014 at 23:31 GMT
Originally opened as https://github.com/AndySomogyi/roadrunner-backup/issues/9

Currently getStoichiometryMatrix returns the reduced stoich matrix, it should return the full matrix. getNrMatrix() is used to return the reduced matrix.

To match the C++/Python and SBW behavior, simulate should not call reset before carrying out a simulation. Current it calls

rri->getSimulateOptions().flags |= SimulateOptions::RESET_MODEL;

This should be removed so that a reset does not occur.

Tried the following:

• Added TEMP_DIR_PATH=/tmp to my config file

Result: No effect:

$ rr ... 2>&1 | grep -i temp
Debug: Temp data folder: /home/jkm/devel/install/roadrunner/bin
Debug: rr::RoadRunner::RoadRunner(const string &, const string &, const string &)compiler: GPUSIM, tempDir:/home/jkm/devel/install/roadrunner/bin, supportCodeDir: 
Debug: getTempDir(): /home/jkm/devel/install/roadrunner/bin/
Debug: getTempDir(): /home/jkm/devel/install/roadrunner/bin/

• Passed -u to the rr exe (this should be the default behavior IMO)

Result: Sort of recognized but not used:

$ rr -u ... 2>&1 | grep -i temp
Debug: Temp data folder: /tmp
Debug: rr::RoadRunner::RoadRunner(const string &, const string &, const string &)compiler: GPUSIM, tempDir:/tmp, supportCodeDir: 
Debug: getTempDir(): /home/jkm/devel/install/roadrunner/bin/
Debug: getTempDir(): /home/jkm/devel/install/roadrunner/bin/

• Passed -t /tmp to rr exe

Result: same as above

The output for model 1148 is slightly off, according to the SBML test suite.

Compute rates of change for all species when moiety conserved cycles active

Line 1119 in rrCompiledExecutableModel.cpp:
std::copy (additionalEvents.begin(), additionalEvents.end(), firedEvents.end());

will crash in some situations on Windows, such as when running model 952. It needs to change to:

            firedEvents.insert(firedEvents.end(), additionalEvents.begin(), additionalEvents.end());

Warning messages written to C stdout likely causing errors with QTConsole on OS 10.8

Test suite models 950 and 951 both successfully run through the simulator, but the comparison code doesn't know how to compare NaN and INF, thereby incorrectly flagging the test as failing.

The python API changes the name of the function to more accurately reflect what's happening under the hood. Ideally, the same change could be made in the C API, but care would have to be taken to ensure that this didn't break SBW compatibility.

rr.model.getConservedMoietyIds returns empty even when there are conserved moieties in the model.

So I was trying to build roadrunner on OSX Mavericks. Apparently the tr1 headers no longer have the tr1 prefix, even in non-c++11 mode (e.g. tr1/memory -> memory). Rather than add even more #ifdefs to each site where the headers are included, I suggest making proxy headers and including those instead. Example: replace

#include <tr1/unordered_map>

with

#include <rr_unordered_map>

which contains:

// usual headers & guards...
#if __APPLE__
#if MAC_OS_X_VERSION_MIN_REQUIRED >= MAC_OS_X_VERSION_10_9 // is Mavericks
#include <unordered_map>
#else
#include <tr1/unordered_map>
#endif // OSX ver
#endif // Apple

// additional #ifdef zoo as necessary

I could implement this fix if desired.

Ref: https://code.google.com/p/alembic/issues/detail?id=322

Issue by hsauro
Tuesday Jul 08, 2014 at 23:32 GMT
Originally opened as https://github.com/AndySomogyi/roadrunner-backup/issues/11

Issue by 0u812
Wednesday Jul 09, 2014 at 23:37 GMT
Originally opened as https://github.com/AndySomogyi/roadrunner-backup/pull/12

Minor tweaks for compiling with gcc 4.8.2 on TK1 (ARM) with LLVM 3.5

0u812 included the following code: https://github.com/AndySomogyi/roadrunner-backup/pull/12/commits

Even if you uncheck 'BUILD_PYTHON' in CMake, it still tries to build the python bindings.

rr.model.getConservedMoietyValues() returns empty for models with conserved moieties. Try this model.

model = '''
J1: A -> B; k1_A - k2_B;
J2: B -> C; k3_B - k4_C;

k1 = 0.1; k2 = 0.02;
k3 = 0.3; k4 = 0.04;
A = 10; B = 0; C = 0;
'''

rr = te.loadAntimonyModel (model)

The models in the test suite that use hierarchical model composition will run in roadrunner if they are flattened, which can happen when importing the model. The model will need to know its filename at that point, so that external model definitions can be read in.

NOTE: libsbml provides functions on a document that will say whether the comp namespace is enabled, and can validate it, etc. These should be used in lieu of regexp tests for the namespace.

Integrator.cpp does not compile for me on windows as-is because of the use of std::transform, which is defined in .

Run this script:

import tellurium as te

model = '''
S1 -> S2; k1_S1;
S2 -> S3; k2_S2;
S1 = 10; k1 = 0.1; k2 = 0.2;
'''
rr = te.loadAntimonyModel (model)

Save the SBML

str = rr.getCurrentSBML()

and you'll get:

return _roadrunner.RoadRunner_getCurrentSBML(self)

RuntimeError: Unknown exception

Issue by AndySomogyi
Sunday May 11, 2014 at 19:30 GMT
Originally opened as https://github.com/AndySomogyi/roadrunner-backup/pull/1

AndySomogyi included the following code: https://github.com/AndySomogyi/roadrunner-backup/pull/1/commits

Bug rr.model.getStoichiometryMatrix(), returns the unordered matrix for models with moiety cycles and moiety conservation is on.

model = '''
J1: A -> B; k1_A - k2_B;
J2: B -> C; k3_B - k4_C;

k1 = 0.1; k2 = 0.02;
k3 = 0.3; k4 = 0.04;
A = 10; B = 0; C = 0;
'''

rr = te.loadAntimonyModel (model)
print rr.getFullStoichiometryMatrix()

The docs for setConfigurationXML is wrong. The tag name for the main section is returned as capabilities but the docs says is caps

A variety of tests were commented out of the python testing module (roadrunner\wrappers\Python\roadrunner\testing\tester.py, line 916 and onward) because roadrunner changed its interface. If commented back in, those tests fail due to syntax errors. They need to be re-instated.

Issue by hsauro
Tuesday Jul 08, 2014 at 23:29 GMT
Originally opened as https://github.com/AndySomogyi/roadrunner-backup/issues/8

Compartmental systems with moiety conservation on given incorrect results. this is most likely due to the fact that the conservation laws are computed using concentrations rather than amounts. eg when computing

S2 = T - S1

The units for T and S1 must be amounts not concentrations. Only amounts are conserved.

Request for feature in the next release. getSBML currently returns Level 3 SBML. However users may require Level 2 in order to run models on older software. Please add an optional argument to getSBML() that specifies the Level and Version for the SBML to be returned. By default getSBML() can return the latest level supported.

Note this change will also need to be added to the C API.

There are a variety of places in the code that warn about unused variables, where lines like:

catch(const std::exception& e)

can be changed to

catch(const std::exception&)

when 'e' is never used. Here are the warnings:

2>....\source\rrRoadRunner.cpp(2593): warning C4101: 'e' : unreferenced local variable

4>......\wrappers\C\rrc_api.cpp(1822): warning C4101: 'e' : unreferenced local variable
4>......\wrappers\C\rrc_api.cpp(1835): warning C4101: 'e' : unreferenced local variable
4>......\wrappers\C\rrc_api.cpp(1849): warning C4101: 'e' : unreferenced local variable
4>......\wrappers\C\rrc_api.cpp(1860): warning C4101: 'e' : unreferenced local variable
4>......\wrappers\C\rrc_api.cpp(1874): warning C4101: 'e' : unreferenced local variable
4>......\wrappers\C\rrc_api.cpp(1886): warning C4101: 'e' : unreferenced local variable
4>......\wrappers\C\rrc_api.cpp(1916): warning C4101: 'e' : unreferenced local variable
4>......\wrappers\C\rrc_api.cpp(1948): warning C4101: 'e' : unreferenced local variable

Issue by luciansmith
Monday Jun 02, 2014 at 17:28 GMT
Originally opened as https://github.com/AndySomogyi/roadrunner-backup/pull/4

luciansmith included the following code: https://github.com/AndySomogyi/roadrunner-backup/pull/4/commits

Remove rr.model.getStoichiometryMatrix()

Issue by AndySomogyi
Friday May 23, 2014 at 21:52 GMT
Originally opened as https://github.com/AndySomogyi/roadrunner-backup/pull/2

AndySomogyi included the following code: https://github.com/AndySomogyi/roadrunner-backup/pull/2/commits

Must use MCJIT. A sampling of compiler errors:

[  1%] Building CXX object source/CMakeFiles/roadrunner.dir/llvm/ASTNodeCodeGen.cpp.o
In file included from /home/poltergeist/home/devel/src/roadrunner/source/llvm/ASTNodeCodeGen.cpp:13:
In file included from /home/poltergeist/home/devel/src/roadrunner/source/llvm/ASTNodeCodeGen.h:13:
/home/poltergeist/home/devel/src/roadrunner/source/llvm/LLVMIncludes.h:73:10: fatal error: 'llvm/ExecutionEngine/JIT.h' file not found
#include <llvm/ExecutionEngine/JIT.h>

Issue by 0u812
Thursday Jul 10, 2014 at 19:46 GMT
Originally opened as https://github.com/AndySomogyi/roadrunner-backup/pull/13

Comparison needed a char cast on ARM since char is unsigned

0u812 included the following code: https://github.com/AndySomogyi/roadrunner-backup/pull/13/commits

So I found the cause of the problems Stan and I were running into whenever we tried to use the Python wrapper on arm. It turns out that Python 2 breaks, you know, like basically everything. Anyway I managed to get it working using Python 3 and ctypes. Verified using my perennial decay model.

My purposes for reporting this issue are twofold: to inform the team that IMHO we really should be planning on dropping Python 2 at some point in the future, and to document it to save someone else the trouble of discovering it the hard way. Something like "beware the prehistoric snake" would be a good admonition.

Btw can someone add me as a contrib so I can label these things? This is more of a long-term feature request / enhancement.

Issue by luciansmith
Monday Jun 30, 2014 at 16:34 GMT
Originally opened as https://github.com/AndySomogyi/roadrunner-backup/pull/5

luciansmith included the following code: https://github.com/AndySomogyi/roadrunner-backup/pull/5/commits

In a model with a reducable Jacobian matrix such as a simple one with two reactions; S1->S2 and S2->S1, the full Jacobian is 2x2 and the reduced Jacobian is 1x1. This means that there are two potential Eigenvalue matrices; one for the full Jacobian (which will have a row that is '0 0'), and one for the reduced Jacobian (which will have only one row).

However, there does not seem to be a way to get the 'full' Eigenvalue matrix out of roadrunner, merely the reduced one. Furthermore, if you request the Eigenvalue for S2 ("eigen(S2)"), you get 'NaN', instead of '0 0'.

The Copasi API has a way to view either eigenvalue matrix, though I don't know what the underlying API calls are. I also don't know what Copasi would return for 'the eigenvalue of S2' (or even if this is possible), but it does seem reasonable that if a user requests it, they should get '0' instead of 'NaN'.

Issue by hsauro
Tuesday Jul 08, 2014 at 23:31 GMT
Originally opened as https://github.com/AndySomogyi/roadrunner-backup/issues/10

The internal comparison code claims that models 1120, 1170, and 1175 fail, when they actually succeed, but the order of the columns is not what the comparison code expected.

in 'jacobian_test', if the number of arguments is wrong (line 173), the routine should return -1 or a similar error code.

Also, there are a few typos: 'jabobian' -> 'jacobian' in the strings.

Issue by luciansmith
Wednesday Jul 02, 2014 at 16:51 GMT
Originally opened as https://github.com/AndySomogyi/roadrunner-backup/pull/7

luciansmith included the following code: https://github.com/AndySomogyi/roadrunner-backup/pull/7/commits

Visual Studio complains if you take a reference to the first element of an empty vector, if you're compiling in debug mode.

The following lines need to be fixed to accommodate this:

line 1905-6 of LLVMExecutableModel.cpp
unsigned char *p1 = &prevEventState[0];
unsigned char *p2 = &currEventStatus[0];

to
unsigned char *p1 = (prevEventState.size() ==0 ? NULL : &prevEventState[0]);
unsigned char *p2 = (currEventStatus.size()==0 ? NULL : &currEventStatus[0]);

line 1934-6:
getEventTriggers(prevEventState.size(), 0, &prevEventState[0]);

unsigned char *p1 = &prevEventState[0];
unsigned char *p2 = &currEventStatus[0];

to

getEventTriggers(prevEventState.size(), 0, (prevEventState.size()==0 ? NULL : &prevEventState[0]));

unsigned char *p1 = (prevEventState.size() ==0 ? NULL : &prevEventState[0]);
unsigned char *p2 = (currEventStatus.size()==0 ? NULL : &currEventStatus[0]);

Line 117-9 of rrSparse.cpp:
memcpy(mat->rowptr, &mrowptr[0], (mat->m+1)_sizeof(unsigned));
memcpy(mat->colidx, &mcolidx[0], nnz_sizeof(unsigned));
memcpy(mat->values, &mvalues[0], nnz*sizeof(double));

to

memcpy(mat->rowptr, &mrowptr[0], (mat->m+1)*sizeof(unsigned));
if (nnz) {
  memcpy(mat->colidx, &mcolidx[0], nnz*sizeof(unsigned));
  memcpy(mat->values, &mvalues[0], nnz*sizeof(double));
}

line 272 of rrCompiledExecutableModel.cpp
getRateRuleValues(&dTemp[0]);

to
getRateRuleValues(mData.numRateRules==0 ? NULL : &dTemp[0]);

line 644-6
vector currentState(getStateVector(0), 0.0);
getStateVector(&currentState[0]);
cevalEvents(&mData, timeIn, &currentState[0]);

to

    int sv = getStateVector(0);
    vector<double> currentState(sv, 0.0);
    getStateVector(sv==0 ? NULL : &currentState[0]);
    cevalEvents(&mData, timeIn, sv==0 ? NULL : &currentState[0]);

line 614-6:
vector currentState(getStateVector(0), 0.0);
getStateVector(&currentState[0]);
cevalModel(&mData, timein, &currentState[0]);

to
int sv = getStateVector(0);
vector currentState(sv, 0.0);
getStateVector(sv==0 ? NULL : &currentState[0]);
cevalModel(&mData, timein, sv==0 ? NULL : &currentState[0]);

line 419 of CVODEIntegrator.cpp
mModel->getEventTriggers(initialEventStatus.size(), 0, &initialEventStatus[0]);

to

mModel->getEventTriggers(initialEventStatus.size(), 0, (initialEventStatus.size()==0 ? NULL : &initialEventStatus[0]));

line 427:
mModel->applyEvents(timeEnd, &previousEventStatus[0], stateVector, stateVector);

to

mModel->applyEvents(timeEnd, (previousEventStatus.size()==0 ? NULL : &previousEventStatus[0]), stateVector, stateVector);

line 227:
mModel->getEventTriggers(eventStatus.size(), 0, &eventStatus[0]);

to

    mModel->getEventTriggers(eventStatus.size(), 0, (eventStatus.size()==0 ? NULL : &eventStatus[0]));

Issue by AndySomogyi
Friday May 23, 2014 at 22:42 GMT
Originally opened as https://github.com/AndySomogyi/roadrunner-backup/pull/3

AndySomogyi included the following code: https://github.com/AndySomogyi/roadrunner-backup/pull/3/commits

I noticed that we have conflicting statements about how the code is licensed. In most of the files we have Apache 2.0, but the doxygen main page (generated from rrRoadRunner.h) has BSD. We do want Apache 2, correct? I can fix this, I just need confirmation from Herbert.

Found this when trying to execute run_test_suite. Causes the entire test suite to fail in CORE_TESTS: LOGGING. Problem is that enableLoggingToFile (rrc_logging_api.cpp) tries to get temp folder. Result is "not used/RoadRunner.log" after path join. Throws Poco FileNotFoundException.