sbml-arrays Mailing List for Systems Biology Markup Language (SBML) (Page 2)

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 422-6466

Hi,

I’m switching from SBML-team to SBML-arrays mailing list, since this is more a discussion about using arrays.  I think many of those on sbml-team are on sbml-arrays and likely those that might be following this discussion: Sarah/Lucian, maybe.

I’m bcc’ing sbml-team, so you are on team but not arrays speak up now about not switching lists, if you want to stay on this thread or better yet subscribe to arrays.  If no one complains though, replying to this message should move the conversation to a more appropriate list.

Chris

> On May 25, 2017, at 1:09 PM, Leandro Watanabe <lea...@gm...> wrote:
> 
> Hi Jacob,
> 
> Sounds good. If you replace 0 by something like random < 0.2 then it makes sense, but leaving the 0 there effectively deletes that event
> 
> In Output2, State0 to State1 Transition_0 event is switching state but the instruction is set to 3. However, none of the events seem to fire on instruction number 3.
> Although your code can generate initial assignments using vector quite easily, I would recommend trying to use other types of math because vector won't allow you to be flexible with changing array sizes in SBML. 
> 
> This message will be archived for the SBML team but you and I don't have access to that Google group. SBML discuss and SBML arrays mailing list are publicly archived. 
> 
> You can continue to attach them for SBML related questions unless they request otherwise.
> 
> iBioSim questions can be addressed to us separately. Flattening is already embedded within iBioSim.
> 
> Leandro 
> 
> On Thu, May 25, 2017 at 10:21 AM, Jacob Barhak <jac...@gm... <mailto:jac...@gm...>> wrote:
> Ok Leandro,
> 
> The instruction you saw should never be executed.  It is the false trigger part of the instruction with the true trigger that comes before that actually increases age. It seems redundant now, yet it has use in some future cases where 0 is replaced by a probability. 
> 
> And let us start another thread. By the way, is the SBML Team email necessary for the new thread? Is it publicly archived somewhere? 
> 
> I already install iBioSim, I just need to know how to pass it the SBML  I generate.  Do I need flattening ?
> 
> I'll try it tomorrow and hopefully have some results for you. 
> 
>          Jacob
> 
> On May 25, 2017 10:40 AM, "Leandro Watanabe" <lea...@gm... <mailto:lea...@gm...>> wrote:
> Hi Jacob,
> 
> I will look over the problems but the problem with events like Age Increase_1 is that it will never fire because you have
> 
> instructionNumber[d0] == 1.5 && 0
> 
> Booleans in SBML is evaluated like weakly typed languages: 0 is false and everything else is true. I will let you know if I find any problem. When you have a 1 instead of 0, it is not doing anything to the math because x && true == x.
> 
> If you want, you can install iBioSim following the instructions: https://github.com/MyersResearchGroup/iBioSim <https://github.com/MyersResearchGroup/iBioSim>
> 
> If you have questions about iBioSim you can e-mail me separately as well since this thread is getting quite long.
> 
> Leandro
> 
>   
> 
> On Thu, May 25, 2017 at 6:35 AM, Jacob Barhak <jac...@gm... <mailto:jac...@gm...>> wrote:
> Hi Leandro,
> 
> You are very helpful and your responses are very timely - I wish most folk will be as responsive as you are.
> 
> The funny triggers you see is because I must always take care of both true and false conditions of an event generated for a rule or for a state transition. For Rules such as Age increase that always executes there is no false condition, yet I create it anyway while making sure it will never trigger and put a zero there. It is possible that such rules in the future may have a probability of being triggered. For example, the rule can be: if a patient takes their medication Blood Pressure drops and they take their medication with a certain probability that sets the trigger - if they do not take their medication, I still have to increase the Instruction number - therefore I need the false trigger event for the same instruction line but not changing the Blood Pressure.
> 
> However, for some reason, it seems that the code does not work - since it worked in the past, I assume this is one of my latest changes of adding an event states or joiner/splitter states that messed the code.
> 
> I went over the code and did find one bug in the event reset instruction - it was 1 instead of 2 - I made the fix, yet it does not explain what you are experiencing on your side - so something else must be wrong.
> 
> Can you please then run example 1 in the code I am providing below - if it works for the 20 people in the simulation, then try example 2 and if 2 works fine, please try example 10. Yet I suspect things ill break at example 1 and since we have previous copies of it, we can go back in time and see on what date it broke to pin point the offending code.
> 
> At this point, things become too complex to just read through the code, so I am seeking explanation by looking at behavior of the code when running it - since I know what to expect at the end, its a good debug strategy.
> 
> If you can guide me through the process of running the code through iBioSim I may be able to do it on my machine to save you time, until then, I will really appreciate another eye - you have been more than helpful so far.
> 
> 
>                   Jacob
> 
> 
> 
> On Wed, May 24, 2017 at 1:52 PM, Leandro Watanabe <lea...@gm... <mailto:lea...@gm...>> wrote:
> Hi Jacob,
> 
> Sorry for the delay. This model is a lot more complex than the others. 
> 
> When I simulated the model, nothing seems to be happening. All of them aged to 31 years old and then nothing happens afterward. Instruction gets stuck at instruction number 2. Seems to be the same for Output2.
> 
> Why are you have setting some triggers to instructionNumber[d0] == 1 && 0, instructionNumber[d0] == 1 && 1 ?
> Did you mean to have a different check?
> 
> Leandro
> 
> 
> 
> On Wed, May 24, 2017 at 2:37 AM, Jacob Barhak <jac...@gm... <mailto:jac...@gm...>> wrote:
> Hi Leandro, Hi Chris,
> 
> Again, the attached file is untested except from passing flattening, yet it seems I was able to implement example 10 which involves splitter and joiner states. Once a splitter state is reached it creates parallel disease processes and moves the individual to those states - like a fork in Linux. One a joiner state was reached it then terminate those processes. In disease modeling this is used many times to indicate what kind of death the person died from, for example a stroke process can compete with a MI process and with an other death process. There is one joiner state that all death lead to, so when an individual dies, all processes collapse. Splitter state are typically used a the beginning of simulation to start multiple processes is a person reaches a certain state such as diabetes that can start other disease processes a person can later die from.
> 
> Implementing this required little changes in the code - it is mostly a change in approach - the event mechanism SBML provides seems to have good support to implement what I am trying to do - once you introduced arrays, it seems to add the extra functionality towards disease modeling at the individual level.
> 
> Yet again I have to finish the entire test suite and see that it does work before I jump to any conclusions. I'll keep on adding examples in hope I can finish this.
> 
> Leandro, if you find a few minutes, can you run example 10 in iBioSim and tell me if it produces similar results to example 2 where state 1,2 in example 10 = state 2 in example 2 - they should be equivalent by design.
> 
> Hopefully this will work.
> 
>                Jacob
> 
> On Tue, May 23, 2017 at 7:18 AM, Jacob Barhak <jac...@gm... <mailto:jac...@gm...>> wrote:
> Thanks Leandro, Thanks Lucian,
> 
> Your fix seems to work.
> 
> I think I also figured out how to represent MIST event states in SBML
> 
> All state transitions events need to be persistent. Instructions numbers will guide execution of rules, so there only the instruction n number matters. For state transitions, all occur simultaneously and the trigger will define which will be executed.
> 
> There are 3 parts in each transition there is a delay of 1 and :
> X: generate Random + advance instruction to X+0.5
> X+0.5: if in proper state and random< threshold of transition, then Change state + advance instruction to Next Instruction
> X+0.5: if in proper state and random>= threshold of transition, then Change state to another state or stay in same state is desired + advance instruction to Next Instructios
> 
> 
> For event states there is a slight difference, the delay is 0 and
> X: generate Random + advance instruction to X+0.5
> X+0.5: if in proper state and random< threshold of transition, then Change state + Reset instruction to X
> X+0.5: if in proper state and random>= threshold of transition, then Change state to another state + Reset instruction to X
> 
> Basically this means that event states just return backwards to the start of the instruction line to allow other state transitions to be processed - so you can potentially process multiple event states - and once a non event state is processed, the system will advance to the next instruction.
> 
> I am attaching a new version with the SBML files having only 2 repetitions per individual - which seems to flatten nicely for all examples.
> 
> Leandro, if you have a minute, I will appreciate if it you look at Examples 1 and 8 - they should work similarly - Where State 2= State1_Entered in Example 9, while State 1 is never set since it is reset upon exit.
> 
> There is a chance it will not work - I may have broken something without knowing it - perhaps my solution is not that elegant - this is why it needs testing.
> 
> If it works, can you see if example 9 also works?
> 
> Hopefully it all works now.
> 
>                  Jacob
> 
> 
> 
> On Mon, May 22, 2017 at 12:30 PM, Leandro Watanabe <lea...@gm... <mailto:lea...@gm...>> wrote:
> Thanks Lucian. 
> 
> Currently in Jacob's code he is using libsbml.parseL3Formula(formula)
> for formula = 'plus(1, selector(Time,d0))'.
> 
> In this case, Time is still csymbol. 
> 
> It looks like he needs to use libsbml.parseL3FormulaWithModel(formula, Model) instead.
> Seems to fix this problem.
> 
> Leandro
> 
> On Mon, May 22, 2017 at 10:14 AM, Lucian Smith <luc...@gm... <mailto:luc...@gm...>> wrote:
> The libsbml ASTNode parser can easily distinguish csymbol time from 'cn' time.  The *infix* parser cannot do this, because you lose information when you go to text-only.  Currently, the infix parser translates all capitalizations of 'time' to the csymbol time, so if you're planning to use infix, you should not have any parameter/variable with the same name as a csymbol or MathML constant (pi, avogadro, true, false, etc.)
> 
> -Lucian
> 
> On Mon, May 22, 2017 at 9:01 AM, Leandro Watanabe <lea...@gm... <mailto:lea...@gm...>> wrote:
> SBML ids should be case sensitive but it looks like the ASTNode parser in libsbml is case insensitive. Someone from the libsbml team can clarify on this.
> 
> Your instructions were previously running serially because of the chain of instructionNumber. Every step was moving from 1 to 1.5 to 2...
> However, one cycle was not a unit delay. Each event was a unit delay. If your cycle has a chain of n events, then one year = n SBML time units. 
> If you say time is in year units, then every time an event fires one year has elapsed because of delay=1.
> 
> I guess if you want to track the year when an individual has reached a terminal state, then it makes sense to add a variable. It just might be a little confusing because time != Time so I would pick another name (also due to the bug). 
> 
> As for event states, I would say that if you want to indicate event states (active and inactive) then you can annotate an event to indicate it is a terminal state. Not sure a better way but maybe someone knows.
> 
> Leandro
> 
> 
> On Mon, May 22, 2017 at 8:57 AM, Jacob Barhak <jac...@gm... <mailto:jac...@gm...>> wrote:
> So Leandro,
> 
> You lost me there.
> 
> My python code has "Time" with a capital T defined. - do not use and small caps "time" in the code.
> 
> Is SBML not case sensitive? Or am I missing something there?
> 
> Also, the delay causes thins to happen one time step ahead and in my case we want instructions to happen serially, so time is actually the instruction timer if I understand correctly - in my simulations typically a set of instructions and transitions all happen in one year which is the regular time step which should be several SBML_time steps if I understand correctly.
> 
> I also think that the scheme we came up with may change due to the introduction of event states - hopefully there is a simple elegant solution there.
> 
>                Jacob
> 
> 
> 
> 
> 
> 
> On Mon, May 22, 2017 at 9:45 AM, Leandro Watanabe <lea...@gm... <mailto:lea...@gm...>> wrote:
> Hi Jacob,
> 
> I think the copyright should be fine as it is. 
> 
> The global simulation time is implied in the model through csymbol time. You can use the value but you won't be able to set a value to it. Time is monotonically increasing and progresses according to your simulator. 
> 
> In your example, you introduced a parameter Time (different from simulation time). However, in the event assignment you are using csymbol time (referring to simulation time). Parameter Time is an array but symbol time is not. Since you are using a selector on the symbol, which is scalar, flattening is failing.
> 
> In your examples, time is moving at the same time for every individual so I don't think you need to introduce a new variable. 
> 
> Leandro
> 
> On Mon, May 22, 2017 at 7:30 AM, Chris Myers <my...@ec... <mailto:my...@ec...>> wrote:
> Hi Jacob,
> 
> I will just add my comments on persistent.  In many cases, the flag setting will not matter.  However, let’s consider the following example:
> 
> Event 1
> Trigger a==1
> TriggerInitiallyTrue false
> Persistent <true/false>
> Delay 1
> Assign a:=0
> Assign b:=1
> 
> Event 2
> Trigger a==1
> TriggerInitiallyTrue false
> Persistent <true/false>
> Delay 2
> Assign a:=0
> Assign c:=1
> 
> Let’s assume a=1 and b=c=0 at time 0.  If persistent is true, then at time 1 b=1 and at time 2 c=1.  However, if persistent is false, then at time 1 b=1, but Event 2 is disabled by Event 1, so c will not change value after all.
> 
> Basically, this flag, when false, allows events to disable each other.
> 
> Does this make sense?
> 
> Chris
> 
>> On May 22, 2017, at 9:10 AM, Jacob Barhak <jac...@gm... <mailto:jac...@gm...>> wrote:
>> 
>> Hi Leandro,
>> 
>> First about the copyright. You and Chris deserve recognition since you guided this transformation of examples to SBML - and I had to add the copyright since MIST is GPL and the code was too close to MIST examples at this point so it was necessary to add it so it won't be forgotten in the future.
>> 
>> However, if all is well, when I start transforming MIST to generate example like the ones you see, I will reuse some of the helper functions I am creating, I will ask that MIST copyright remains under my name - and the U of M for historical reasons.  This may not matter , yet under certain circumstances it may have future implications - although everyone can use the GPL code once I release it I and plan to acknowledge your contribution - let me know if you or Chris have any objections to this or if you want to go into a more elaborate discussion on this copyright issue.
>> 
>> After your explanations, I understood that I need to add a local Time parameter for each individual and keep track of time separately - since the events will be resolved during SBML_Time+1 - so my simulation time is not SBML_Time. 
>> 
>> I have not dealt with the persistent flag for events - it is really too much information for me to handle right now - I really cannot see how all these events will trigger with the elaborate structure of instructions order we added - I think that persistent should be set as true fro all my events, yet it is currently set to false and it seemed to work so far, so I am unsure.
>> 
>> However, I do need help in a more critical point - I added some code to allow creating smaller examples by allowing to override number of repetitions. You can now run my examples with a command line parameter to set the number of repetitions.
>> 
>> Below you will find a transcript of what happens when I try to flatten SBML that has only 2 repetitions = 20 records for example 1. I get an error, yet I cannot figure out at what SBML file this is happening - the error message does not point to the parameter name or XML line.
>> 
>> The only parameter I added is Time and it seems to be similar to others, yet I cannot pinpoint the issue.
>> 
>> Since it will probably more visible to you, I will appreciate it if you can look at the XML and help figure it out.
>> 
>>               Jacob
>> 
>> ############# Transcript of execution and error #############
>> 
>> C:\Users\Work\Desktop\SBML>winpdb SBMLExamples.py 2
>> C:\Users\Work\Desktop\SBML>java -jar runner.jar -Xmx4096 OutSBML1.xml
>> Exception in thread "main" java.lang.IndexOutOfBoundsException: Index: 19, Size:
>>  0
>>         at java.util.ArrayList.rangeCheck(Unknown Source)
>>         at java.util.ArrayList.get(Unknown Source)
>>         at org.sbml.jsbml.ASTNode.getChild(ASTNode.java:2489)
>>         at org.sbml.jsbml.ext.arrays.compiler.VectorCompiler.selector(VectorComp
>> iler.java:4132)
>>         at org.sbml.jsbml.ASTNode.compile(ASTNode.java:2159)
>>         at org.sbml.jsbml.ext.arrays.compiler.VectorCompiler.getScalarsAndVector
>> s(VectorCompiler.java:2515)
>>         at org.sbml.jsbml.ext.arrays.compiler.VectorCompiler.plus(VectorCompiler
>> .java:3642)
>>         at org.sbml.jsbml.ASTNode.compile(ASTNode.java:1952)
>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.convertArraysMa
>> th(ArraysFlattening.java:204)
>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.convertMath(Arr
>> aysFlattening.java:270)
>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.convertMath(Arr
>> aysFlattening.java:271)
>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.convertMath(Arr
>> aysFlattening.java:271)
>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.convertMath(Arr
>> aysFlattening.java:271)
>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.convertMath(Arr
>> aysFlattening.java:285)
>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.convert(ArraysF
>> lattening.java:90)
>>         at io.github.leandrohw.Runner.main(Runner.java:68)
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> On Sun, May 21, 2017 at 2:16 PM, Leandro Watanabe <lea...@gm... <mailto:lea...@gm...>> wrote:
>> Hi Jacob,
>> 
>> Sounds good. This is interesting work thanks for including us.
>> 
>> I do not see any problem with Output7. The problem is the model is too large so maybe the flattening algorithm needs some redesigning to scale out. I would say do not worry about flattening for such large models. If it flattens for small sizes, then you can assume it works for 2000 or so. This is one of the advantages of arrays. You can express large models more concisely. Especially because SBML is XML-based, the flattened models can get quite large.
>> 
>> About some of your questions:
>> - You cannot remove the event from the model. The simulator will continually check for the trigger. However, you can setup the trigger such that the event will never fire. You might be able to design a simulator to make assumptions about specifics properties of your models though.
>> 
>> - You are correct about setting delay to 0 make the event instantaneous. 
>> 
>> - Event delays are the used to indicate when an event fire. If an event is triggered at time n, and has a delay of d, then the event only happens at time n+d.
>> 
>> - Events can be in two states: triggered and not triggered. An event is triggered if the trigger was previously false and then evaluated to true.
>> 
>> - The persistent flag is used to indicate whether you want to rechecked the trigger value again. In your model, you have a delay of 1 in many events. If you want to ensure that the event's trigger stays true for n and n+1, then you use the persistent flag (true if you fire regardless and false if you want to recheck). 
>> 
>> Thanks,
>> Leandro
>> 
>> On Sun, May 21, 2017 at 3:04 AM, Jacob Barhak <jac...@gm... <mailto:jac...@gm...>> wrote:
>> Hi Leandro, Hi Chris,
>> 
>> Attached you will find a new version of the SBML MIST examples. I added the event example #8 and I also added a copyright statement for this code to remain GPL - after all I am copying from the MIST GPL code examples, so I must keep this GPL. I added your names since you contributed to this - yet I can remove them in the next version if you do not want any responsibility for this code.
>> 
>> I made some changes to accommodate event states. Yet this also makes flattening impossible - I can check example 8, yet I cannot flatten it anymore.
>> 
>> Leandro, it will really help if you can run the XML through your system - Example 8 is relatively trivial - the same as example 1 only with an event state in the middle that just triggers and passes to the next state immediately.
>> 
>> If this seems reasonable to you, I will continue and complicate examples.
>> 
>> Hopefully the changes I made will work ok.
>> 
>>         Jacob
>> 
>> 
>> 
>> 
>> 
>> On Fri, May 19, 2017 at 9:37 AM, Leandro Watanabe <lea...@gm... <mailto:lea...@gm...>> wrote:
>> Hi Jacob,
>> 
>> Sounds good. Looks like you got pretty good with SBML. Your models look fine. I haven't been able to see if they behave as you described but since you can test in MIST then it is all good. 
>> 
>> This is looking quite promising.
>> 
>> Thanks,
>> Leandro
>> 
>> On Fri, May 19, 2017 at 4:35 AM, Jacob Barhak <jac...@gm... <mailto:jac...@gm...>> wrote:
>> Hi Leandro,
>> 
>> You will find attached even a new version of the test code. I am recreating MIST examples serially and it seems that the current structure allows moving fast - it takes only  a few minutes to create a new example. Examples 4,5,6 add some little complexity of functions and covariates and size - simulations are now a bit larger.
>> 
>> No need to spend any time of those. I will just continue moving along - you see MIST has test code to test the simulation results of these examples - this is not something a human should do - yet I am sending those to you so that you will have a copy once I run into trouble and I expect this will happen pretty soon.
>> 
>> Yet I am making progress fast, so its a good sign.
>> 
>>       Jacob
>> 
>> 
>> On Thu, May 18, 2017 at 8:47 AM, Jacob Barhak <jac...@gm... <mailto:jac...@gm...>> wrote:
>> Many thanks Leandro,
>> 
>> You were very helpful - I was overwhelmed with the number of changes and missed some texts you and Chris wrote- I think I am all caught up now.
>> 
>> Your new runner is working nicely and the system actually does flatten the XML files, yet I am sending only the code with the original XMLs because the flattened files are just too large.
>> 
>> If this looks ok to you, I will proceed to the next example.
>> 
>> I foresee that some examples will take much more time. Yet if I manage to finish them all, I will be able to write a converter from MIST eventually.
>> 
>> Hopefully I am not taking too much of your time with these questions.
>> 
>>          Jacob
>> 
>> 
>> On Wed, May 17, 2017 at 10:18 AM, Leandro Watanabe <lea...@gm... <mailto:lea...@gm...>> wrote:
>> That's right. You can run the validator with a -c flag for checking validation rules.
>> java -jar runner.jar -c OutSBML2.xml
>> 
>> NOTE: Do not worry about the errors about selectors. It is a bug.
>> 
>> Jacob: Output2 and Output3 still have references to Dead in the terminate event so flattening fails. Output1 can be flattened.
>> 
>> BTW, try running with -Xmx2048m (or 512m, 1024m etc). You are running out of heap space. 
>> 
>> The running time of the flattening is not so bad, but maybe we need to take a look on how to improve memory space in the future. Maybe there are dangling references to unused objects. 
>> 
>> Leandro
>> 
>> 
>> 
>> On Wed, May 17, 2017 at 7:03 AM, Chris Myers <my...@ec... <mailto:my...@ec...>> wrote:
>> The arrays validator like the arrays flattening is only available in Java.  I think Leandro in his last message said he add this to the arrays flatten code he gave you.
>> 
>> Chris
>> 
>>> On May 17, 2017, at 7:29 AM, Jacob Barhak <jac...@gm... <mailto:jac...@gm...>> wrote:
>>> 
>>> Thanks Leandro, Thanks Chris,
>>> 
>>> First Chris, your code is in java and I cannot find the validator function in python. The closest I can see at the top level is: SBMLExternalValidator I do not see the function ArraysValidator - under what object should I be able to find it?
>>> 
>>> Leandro, thanks for noticing the death - it was a bug - I actually needed to dive deeper into the code and define a terminating event that starts the stimulation for Example3. The new event just checks at the beginning of each time loop if  a terminal state is set, if it is, then the system skips assigning the next instruction line. This is how an individual simulation is terminated ahead of time in case of death.
>>> 
>>> This also meant I needed to change the structure of all transition events and rule events - now rule events also have true and false elements with Instruction number advancing in both true and false parts.
>>> 
>>> Attached is a new version of the file and you will also find Example 3 there with 2 terminal states to test this is working.
>>> 
>>> However, your java code does not help flattening - below is the error I get with your new runner version. 
>>> 
>>> Since I made many changes, you can assume this version has some bugs in it. I will look at this more closely tomorrow - yet if you have a few minutes and can lend me another eye on this code, it would help me advance faster.
>>> 
>>>             Jacob
>>> 
>>> ########### The Error I am getting when flattening ###########
>>> 
>>> Exception in thread "main" java.lang.OutOfMemoryError: Java heap space
>>>         at java.util.Arrays.copyOf(Unknown Source)
>>>         at java.util.ArrayList.grow(Unknown Source)
>>>         at java.util.ArrayList.ensureExplicitCapacity(Unknown Source)
>>>         at java.util.ArrayList.ensureCapacityInternal(Unknown Source)
>>>         at java.util.ArrayList.addAll(Unknown Source)
>>>         at org.sbml.jsbml.AbstractTreeNode.addAllChangeListeners(AbstractTreeNod
>>> e.java:200)
>>>         at org.sbml.jsbml.AbstractTreeNode.addAllChangeListeners(AbstractTreeNod
>>> e.java:207)
>>>         at org.sbml.jsbml.AbstractTreeNode.addAllChangeListeners(AbstractTreeNod
>>> e.java:207)
>>>         at org.sbml.jsbml.AbstractTreeNode.addAllChangeListeners(AbstractTreeNod
>>> e.java:190)
>>>         at org.sbml.jsbml.AbstractTreeNode.setParent(AbstractTreeNode.java:796)
>>>         at org.sbml.jsbml.ASTNode.setParentSBMLObject(ASTNode.java:1342)
>>>         at org.sbml.jsbml.AbstractMathContainer.setMath(AbstractMathContainer.ja
>>> va:322)
>>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.updateMath(Arra
>>> ysFlattening.java:580)
>>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.expandDim(Array
>>> sFlattening.java:376)
>>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.expandDim(Array
>>> sFlattening.java:416)
>>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.convert(ArraysF
>>> lattening.java:168)
>>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.convert(ArraysF
>>> lattening.java:170)
>>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.convert(ArraysF
>>> lattening.java:187)
>>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.convert(ArraysF
>>> lattening.java:89)
>>>         at io.github.leandrohw.Runner.main(Runner.java:68)
>>> 
>>> 
>>> 
>>> 
>>> On Tue, May 16, 2017 at 1:00 PM, Leandro Watanabe <lea...@gm... <mailto:lea...@gm...>> wrote:
>>> The only problem I found was that event1 had a reference to Dead in its trigger but you don't have a variable with that name.
>>> 
>>> I updated the jar on the Dropbox link I sent you earlier. If you want to run validation, you can do:
>>> 
>>> java -jar runner.jar -c Output1.xml
>>> 
>>> the -c flag is for checking validation rules. If you don't specify it, then it will just do flattening.
>>> 
>>> Note: there are some bugs in the validator that I have to fix. I am planning to work on them during Harmony so you need to ignore some of them. 
>>> 
>>> If you run into this: Selector arguments other than first should not go out of bounds 
>>> then don't worry about it. This is because the event assignment is using the dimensions from the parent event and the validator is not picking it up.
>>> 
>>> Leandro
>>> 
>>> 
>>> 
>>> On Tue, May 16, 2017 at 10:34 AM, Chris Myers <my...@ec... <mailto:my...@ec...>> wrote:
>>> Hi Jacob,
>>> 
>>> It is usually best to validate before you flatten.  When there are errors, flatten can crash as shown.  I think something like this:
>>> 
>>>   			List<SBMLError> arraysErrors = ArraysValidator.validate(doc);
>>>   			for (int i = 0; i < arraysErrors.size(); i++)
>>>   			{
>>>   				String error = arraysErrors.get(i).getMessage();
>>>   				message += numErrors + ":" + error + "\n";
>>>   				numErrors++;
>>>   			}
>>> 
>>> Chris
>>> 
>>>> On May 16, 2017, at 12:24 PM, Jacob Barhak <jac...@gm... <mailto:jac...@gm...>> wrote:
>>>> 
>>>> Hi Leandro,
>>>> 
>>>> Attaches is the python code for the first attempt to model two examples using SBML array.
>>>> 
>>>> If you look at the code you will find out that I was able to identify and consolidate repetitive code into functions. You will also find example 2 from:
>>>> 
>>>> https://github.com/Jacob-Barhak/MIST/blob/master/SimulationExamples.pdf <https://github.com/Jacob-Barhak/MIST/blob/master/SimulationExamples.pdf>
>>>> 
>>>> I also increase the number of individuals of Example 1 to match the example in the document.
>>>> 
>>>> I am attaching the code and output, yet I can no longer flatten the code - Below is the error message I get:
>>>> 
>>>> I know it is not realistic to flatten large arrays, so I am not worried. However, I lost the mechanism that allowed me to check for errors.
>>>> 
>>>> Is there any other idea on how to debug other than decreasing array size?
>>>> 
>>>> Hopefully the code looks ok to you.
>>>> 
>>>>           Jacob
>>>> 
>>>> 
>>>> ###### The Error I get #######
>>>> 
>>>> Exception in thread "main" java.lang.OutOfMemoryError: GC overhead limit exceede
>>>> d
>>>>         at org.sbml.jsbml.ASTNode.initDefaults(ASTNode.java:3023)
>>>>         at org.sbml.jsbml.ASTNode.<init>(ASTNode.java:1581)
>>>>         at org.sbml.jsbml.ASTNode.clone(ASTNode.java:1908)
>>>>         at org.sbml.jsbml.ASTNode.<init>(ASTNode.java:1603)
>>>>         at org.sbml.jsbml.ASTNode.clone(ASTNode.java:1908)
>>>>         at org.sbml.jsbml.ASTNode.<init>(ASTNode.java:1603)
>>>>         at org.sbml.jsbml.ASTNode.clone(ASTNode.java:1908)
>>>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.replaceDimensio
>>>> nId(ArraysFlattening.java:563)
>>>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.updateMath(Arra
>>>> ysFlattening.java:580)
>>>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.expandDim(Array
>>>> sFlattening.java:376)
>>>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.expandDim(Array
>>>> sFlattening.java:416)
>>>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.convert(ArraysF
>>>> lattening.java:168)
>>>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.convert(ArraysF
>>>> lattening.java:170)
>>>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.convert(ArraysF
>>>> lattening.java:187)
>>>>         at org.sbml.jsbml.ext.arrays.flattening.ArraysFlattening.convert(ArraysF
>>>> lattening.java:89)
>>>>         at io.github.leandrohw.Runner.main(Runner.java:29)
>>>> 
>>>> 
>>>> 
>>>> 
>>>> On Sun, May 14, 2017 at 1:26 PM, Leandro Watanabe <lea...@gm... <mailto:lea...@gm...>> wrote:
>>>> Hi Jacob,
>>>> 
>>>> This looks great. The model can be simulated without a problem. Your code is in good state to move on to other examples.
>>>> 
>>>> Leandro
>>>> 
>>>> On Sun, May 14, 2017 at 4:50 AM, Jacob Barhak <jac...@gm... <mailto:jac...@gm...>> wrote:
>>>> Thanks Leandro,
>>>> 
>>>> You are very observant. I overlooked the difference between state transition events and rule events that change parameter values that do not have an else clause.
>>>> 
>>>> I made the changes in the python file. Attached is the file and its products.
>>>> 
>>>> If you think it is ready, I can move on to a more complicated disease modeling example.
>>>> 
>>>> Hopefully it seems ready to you.
>>>> 
>>>>               Jacob
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> On Thu, May 11, 2017 at 12:44 PM, Leandro Watanabe <lea...@gm... <mailto:lea...@gm...>> wrote:
>>>> I'm looking at the models using a GUI so it makes it easier to spot these things.
>>>> 
>>>> It looks like your event1 is not switching states. You need InstructionNumber[d0] = 2
>>>> 
>>>> Other than that, it looks pretty good.
>>>> 
>>>> Here is some simulation result. 
>>>> <Screen Shot 2017-05-11 at 11.32.03 AM.png>
>>>> 
>>>> Sorry about the axis labels. Time is SBML simulation time but the cycle number is not really the same as simulation time because of the delay 1. Each year is about 4 time steps.
>>>> 
>>>> <Screen Shot 2017-05-11 at 11.12.57 AM.png>
>>>> 
>>>> This is not quite the results you get because of the iterations and repetitions in your simulation rules. In general, simulation rules are specified in a SED-ML file. However, you can have a counter in your model that will only repeat the loop three times and check the counter in the events trigger or use SBML constraints, 
>>>> 
>>>> Thanks,
>>>> Leandro 
>>>> 
>>>> On Thu, May 11, 2017 at 8:13 AM, Jacob Barhak <jac...@gm... <mailto:jac...@gm...>> wrote:
>>>> Thanks Leandro,
>>>> 
>>>> You are faster than me with testing this code. Your are also very observant - this is truly helpful and brings another issue - terminal states. If you look at the Example 1 diagram in https://github.com/Jacob-Barhak/MIST/blob/master/SimulationExamples.pdf <https://github.com/Jacob-Barhak/MIST/blob/master/SimulationExamples.pdf> you will see that the dead state has a red box around it - this indicates a terminal state, meaning that once this state is reached, there is no next time step - the individual record does not exist in the next time step.
>>>> 
>>>> Also in the future, I may need event states where there are split assignments of what may happen. So I chose the option of true event and false event.
>>>> 
>>>> I am attaching the new version with output and flattened XML.
>>>> 
>>>> Hopefully it will behave better - however, it did not pass rigorous testing, so another eye on this will be truly helpful.
>>>> 
>>>>                Jacob
>>>> 
>>>> On Wed, May 10, 2017 at 10:29 AM, Leandro Watanabe <lea...@gm... <mailto:lea...@gm...>> wrote:
>>>> Hi Jacob, 
>>>> 
>>>> Looks good. This is great progress. Make sure that InstructionNumber constant=false though.
>>>> 
>>>> Now that the syntax looks good, we can talk about semantics.
>>>> 
>>>> Good news: I was able to simulate your model. However, I encounter one problem. 
>>>> 
>>>> One thing I am not sure about is event3. If the probability < 0.0717, you say the person dies. Otherwise, you continue onto the next year. The problem with your model now is that if event3 is false, then the model halts because you are not setting instruction number back to 1 for probability >= 0.0717. In this case, all events become disabled.
>>>> 
>>>> I recommend changing event3 to either:
>>>> 
>>>> Solution 1:
>>>> 
>>>> event3:
>>>> - Trigger: InstructionNumber[d0] == 2.5
>>>> - Delay: 1
>>>> - Priority: 1
>>>> - Event Assignments:
>>>>      * Alive[d0] = piecewise(0, Alive[d0] == 1 && Random[d0] < 0.0717, 1)
>>>>      * Dead[d0] = piecewise(1, Alive[d0] == 1 && Random[d0] < 0.0717, 0)
>>>>      * InstructionNumber[d0] = 1
>>>> 
>>>> Piecewise is like if statements: 0 if Alive[d0] == 1 && Random[d0] < 0.0717 else 1.
>>>> In L3V2, you would be able to use booleans in arithmetic operations instead of piecewise. However, it would be nice to get L3V1 done first.
>>>> 
>>>> This way you can cycle back and do all the instructions again. This way you wouldn't need the Random variable anymore. You can just call uniform within the event assignment. 
>>>> 
>>>> Solution 2:
>>>> Another way would be to create another event4 that would trigger when:
>>>> - InstructionNumber[d0] == 2.5 &&  Alive[d0] == 1 && Random[d0] >= 0.0717 
>>>> where the event assignment  would only reset instruction number back to 1.
>>>> 
>>>> Thanks,
>>>> Leandro
>>>> 
>>>> On Tue, May 9, 2017 at 7:18 AM, Chris Myers <my...@ec... <mailto:my...@ec...>> wrote:
>>>> Hi Jacob,
>>>> 
>>>> Sounds like great progress.  I would suggest not worrying too much about getting the UncertML portions of distrib, but to instead focus on simply the annotation scheme.  Other discussions may be taking distrib in a different direction, and in any case, the annotations are more widely supported at this time.
>>>> 
>>>> I do hope you can make it to Seattle for Harmony in June.  It would be a great opportunity for you to meet those working in this area and hack through any lingering issues.
>>>> 
>>>> Cheers,
>>>> 
>>>> Chris
>>>> 
>>>>> On May 9, 2017, at 7:49 AM, Jacob Barhak <jac...@gm... <mailto:jac...@gm...>> wrote:
>>>>> 
>>>>> Thanks Leandro,
>>>>> 
>>>>> Your guidance here is very much appreciated.
>>>>> 
>>>>> I rearranged the code a bit and it passed flattening - see attached.
>>>>> 
>>>>> Now it should be accessible by other systems if they support the features used .
>>>>> 
>>>>> I am going to try and run the attached code on some other systems to see if this example works fine.
>>>>> 
>>>>> If it does, I can go on and try more elaborate examples to see how far I can get with modeling disease models in SBML. I am organizing this example code in a way that I will be able in the future to export disease models from MIST to iBioSim.
>>>>> 
>>>>> Hopefully things will be easier from this point on.
>>>>> 
>>>>>              Jacob
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> On Mon, May 8, 2017 at 11:16 PM, Leandro Watanabe <lea...@gm... <mailto:lea...@gm...>> wrote:
>>>>> Hi Jacob,
>>>>> 
>>>>> Since nobody responded, I looked through the documentation and figured out how to do the distribution for your model.
>>>>> I don't know if it is the version I have or what, but the name of the python function to create uncert nodes was not consistent with the given examples.
>>>>> Maybe this is what caused you trouble.
>>>>> 
>>>>> This is consistent to what Chris sent earlier. It has the commonly used annotation and the distrib for your function definition.
>>>>> 
>>>>> Hope this helps.
>>>>> 
>>>>> Leandro
>>>>> 
>>>>> On Mon, May 8, 2017 at 1:59 AM, Jacob Barhak <jac...@gm... <mailto:jac...@gm...>> wrote:
>>>>> Sorry Brett,
>>>>> 
>>>>> The same message is returned after running the command you mentioned that cleared some disk space on my machine. For some reason, my system thinks the new file is corrupt.
>>>>> 
>>>>> The downloaded file size shown on my machine is:
>>>>> Size:             5.51 MB (5,781,328 bytes)
>>>>> Size on disk: 5.51 MB (5,783,552 bytes)
>>>>> 
>>>>> If there is a checksum I can test, I'll be happy to make sure it was downloaded correctly.
>>>>> 
>>>>> I noticed that the path the system reports uses unix style separators rather than windows: 'Lib/site-packages/libsbml/__init__.pyc' is this the reason that the path is not found?
>>>>> 
>>>>> The other option I was thinking is reinstalling anaconda, yet I rather first try to make sure that the downloaded file is ok.
>>>>> 
>>>>> However, this issue does not block my progress, I am more in need of help to define distributions in SBML in python. Yet it would be nice to fix this issue on the way.
>>>>> 
>>>>>              Jacob
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> On Tue, Apr 25, 2017 at 4:03 AM, Brett G. Olivier <bre...@gm... <mailto:bre...@gm...>> wrote:
>>>>> Dear Jacob
>>>>> 
>>>>> Just to follow up on this, indeed this was a bug in the Arrays implementation, thanks for reporting it. Fortunately this issue has been fixed and an update version is available for testing. 
>>>>> 
>>>>> I see that you are using Anaconda Python, here are the instructions for downloading and installing they updated libSBML Anaconda package.
>>>>> 
>>>>> Installing the Anaconda nightly build
>>>>> ----------------------------------------------------
>>>>> 
>>>>> In order to install the nightly build of the Python Anaconda packages
>>>>> 
>>>>> Visit http://raterule.caltech.edu/ <http://raterule.caltech.edu/> and browse to Section 6 at the bottom of the page::
>>>>> 
>>>>>  - select 32/64 bit directory
>>>>>  - select the latest experimental package, for Python 2.7 this is python-libsbml-experimental-5.15.1-py27_0.tar.bz2
>>>>>  - save it to a local directory 
>>>>>  
>>>>>  Install the package locally with (in the same directory as the downloaded file)::
>>>>>  
>>>>>   conda install python-libsbml-experimental-5.15.1-py27_0.tar.bz2
>>>>>   
>>>>> If it installs skip this if you receive an error, something like::
>>>>> 
>>>>>  CondaError: Cannot link a source that does not exist. C:\Anaconda\pkgs\python-libsbml-experimental-5.15.1-py27_0\Lib\site-packages\libsbml\__init__.pyc
>>>>>  
>>>>> Don't worry, this is an Anaconda issue that, fortunately, has a simple workaround. Change to the directory mentioned in the error::
>>>>> 
>>>>>  c:\Anaconda\pkgs\python-libsbml-experimental-5.15.1-py27_0\Lib\site-packages\libsbml\
>>>>>  
>>>>> copy __init__.py to __init__.pyc and rerun the installer::
>>>>> 
>>>>>  conda install python-libsbml-experimental-5.15.1-py27_0.tar.bz2
>>>>>  
>>>>> and you should be good to go. An alternative to the above installation method is to simply overwrite your installed libSBML with the one in the downloaded archive.
>>>>> 
>>>>> Either way, if your installation is successful you get the following::
>>>>> 
>>>>>  In [1]: import libsbml as sb
>>>>>  In [2]: print(sb.LIBSBML_VERSION_STRING)
>>>>>  51501
>>>>>  In [3]: D = sb.readSBMLFromFile('arrays1.xml')
>>>>>  In [4]: M = D.getModel()
>>>>>  In [5]: M.getSpecies(0).getPlugin('arrays').getListOfDimensions()[0].getSize()
>>>>>  Out[5]: 'n'
>>>>> 
>>>>> This is the output I received after I used the first method installing into a custom Anaconda environment.
>>>>> 
>>>>> HTH
>>>>> Brett (on behalf of the libSBML team)
>>>>> 
>>>>> 
>>>>> 
>>>>> On Wed, Apr 19, 2017 at 1:36 PM, Jacob Barhak <jac...@gm... <mailto:jac...@gm...>> wrote:
>>>>> 
>>>>> Here is what I tried in python:
>>>>> 
>>>>>  from libsbml import *
>>>>>  reader = SBMLReader()
>>>>>  reader.readSBMLFromFile('OutSBML_2017_04_17.xml')
>>>>> 
>>>>> The system just hung and the python process stopped rather than produce an exception - this should be fixed. I am using  Python 2.7.11 |Anaconda 2.4.1 (64-bit) with libsbml version '5.15.0' on windows 7.
>>>>> 
>>>>> At this point I need to know if the problem is something I need to fix or something on the SBML team side.
>>>>> 
>>>>> Hopefully its a silly issue with a quick fix.
>>>>> 
>>>>>           Jacob
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> <Flattened_2017_05_09.xml><OutSBML_2017_05_09.xml><SBMLExample1_2017_05_09.py>
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> <SBMLExamples_2017_05_16.zip>
>>> 
>>> 
>>> 
>>> <SBMLExamples_2017_05_17.zip>
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> <SBMLExamples_2017_05_22.zip>
> 
> 
> 
> 
> 
> -- 
> You received this message because you are subscribed to the Google Groups "sbml-team" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to sbm...@go... <mailto:sbm...@go...>.
> For more options, visit https://groups.google.com/d/optout <https://groups.google.com/d/optout>.
> 
> 
> 
> 
> 
> 
> 
> 
> 

2012	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug	Sep	Oct	Nov	Dec (1)
2013	Jan	Feb	Mar	Apr (2)	May	Jun	Jul	Aug	Sep (10)	Oct (27)	Nov	Dec (6)
2014	Jan (25)	Feb	Mar (44)	Apr (21)	May (1)	Jun (7)	Jul (3)	Aug (18)	Sep (7)	Oct (1)	Nov	Dec
2015	Jan	Feb	Mar	Apr (9)	May	Jun	Jul	Aug	Sep (15)	Oct	Nov	Dec
2017	Jan	Feb	Mar	Apr	May (31)	Jun (4)	Jul (9)	Aug	Sep	Oct	Nov	Dec
2018	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug (1)	Sep (2)	Oct	Nov	Dec

sbml-arrays Mailing List for Systems Biology Markup Language (SBML) (Page 2)

A file format for exchanging computational models in systems biology

sbml-arrays — The SBML L3 Arrays and Sets package discussion list