Re: [netlogo-devel] Procedure call graph

[Steve Upton]

Thanks again, Seth!!

steve

On Monday, June 29, 2020 at 1:30:54 PM UTC-4, Seth Tisue wrote:
>
> On Thu, Jun 25, 2020 at 7:46 AM Steve Upton  > wrote: 
> > BTW, what is __ignore? I don't see it in the Netlogo dictionary. 
>
> see https://github.com/NetLogo/NetLogo/wiki/Unofficial-features 
>
> Seth 
>

-- 
You received this message because you are subscribed to the Google Groups 
"netlogo-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to netlogo-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/netlogo-devel/aabd1633-dea3-460c-9380-590ca320886fo%40googlegroups.com.

Re: [netlogo-devel] Procedure call graph

[Seth Tisue]

On Thu, Jun 25, 2020 at 7:46 AM Steve Upton  wrote:
> BTW, what is __ignore? I don't see it in the Netlogo dictionary.

see https://github.com/NetLogo/NetLogo/wiki/Unofficial-features

Seth

-- 
You received this message because you are subscribed to the Google Groups 
"netlogo-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to netlogo-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/netlogo-devel/CAOoO2vgK8pWGteXpnyuESo4LB-2Opx7ar%2BjEma4PKMYm1eJocw%40mail.gmail.com.

Re: [netlogo-devel] Procedure call graph

[Steve Upton]

Seth,

Thank you for following up! Learning something new about Netlogo and Scala 
everyday!  

You are correct - I am only looking at the first level of the tree. I had 
assumed I only needed to go one level to capture all the procedures that 
called a set of procedures within its code block, and that all procedures 
would be included in the workspace.procedures map (and interested in 
user-defined procedures mainly). For your test case (I made a slight mod to 
make an edge from a procedure that didn't make any calls to a "Nothing" 
node), I get:
FOO -> Nothing
BAR -> Nothing
BAZ -> FOO

so clearly missed the call to Bar. 

BTW, what is __ignore? I don't see it in the Netlogo dictionary.

Also, thanks for the nitpick - always good to clean up code.

steve

On Wednesday, June 17, 2020 at 2:00:46 PM UTC-4, Seth Tisue wrote:
>
> This looks it will produce correct output in many cases, but I'm not 
> convinced it's correct in general.
>
> Does it work on a case like this, and on even more deeply nested cases?
>
> to-report foo [x] report x end
> to-report bar [x] report x end
> to baz __ignore foo bar 0 end
>
> I'm looking at your logic and it isn't clear to me that all possible 
> nestings (including recursive nestings) of _call and _callreport are 
> handled properly, hence the test case. I don't see anywhere in your code 
> that recursively processes entire argument trees, and not just the first 
> level of the tree.
>
> Also a nitpick that won't usually affect correctness: you shouldn't need 
> to resort to looking for "_call" and "_callreport' as strings — _call and 
> _callreport are classes, you should be able to detect instances of them via 
> pattern matching (or `isInstanceOf`).
>
> Seth
>

-- 
You received this message because you are subscribed to the Google Groups 
"netlogo-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to netlogo-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/netlogo-devel/7bb92c80-6dd6-4255-98ee-300a25b7b198o%40googlegroups.com.

Re: [netlogo-devel] Procedure call graph

[Seth Tisue]

This looks it will produce correct output in many cases, but I'm not 
convinced it's correct in general.

Does it work on a case like this, and on even more deeply nested cases?

to-report foo [x] report x end
to-report bar [x] report x end
to baz __ignore foo bar 0 end

I'm looking at your logic and it isn't clear to me that all possible 
nestings (including recursive nestings) of _call and _callreport are 
handled properly, hence the test case. I don't see anywhere in your code 
that recursively processes entire argument trees, and not just the first 
level of the tree.

Also a nitpick that won't usually affect correctness: you shouldn't need to 
resort to looking for "_call" and "_callreport' as strings — _call and 
_callreport are classes, you should be able to detect instances of them via 
pattern matching (or `isInstanceOf`).

Seth

-- 
You received this message because you are subscribed to the Google Groups 
"netlogo-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to netlogo-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/netlogo-devel/9d8d89fe-66e8-494f-9786-154217a5b94fn%40googlegroups.com.

Re: [netlogo-devel] Procedure call graph

[Steve Upton]

In case anyone is interested, I came up with this simple code to generate 
the call graph:
import org.nlogo.headless.HeadlessWorkspace
import org.nlogo.nvm._

case class GraphEdge(from : String, to: String, isolate : Boolean) {
override def toString = s"$from -> $to"
def toCSV = s"$from,$to,$isolate"
}

object GraphEdge {
def apply(from: String, to: String) : GraphEdge = {
val isolate = to match {
case "" => true
case _ => false
}
apply(from, to, isolate)
}

}

 object GenerateCallGraph {
  def extractCallGraph(procedures : Procedure.ProceduresMap) : 
Set[GraphEdge] = {
val procNames = procedures.keySet
println("printing the procNames")
procNames foreach println
println("***printing the calls")
val edges = procedures.map{case(parent,child) => {
println(s"working procedure: $parent")
val (gcall,gargs) = child.code.partition(_.toString.contains("_call"))
val topEdges = gcall.map(c => GraphEdge(from=parent,to=c.displayName))
// greports : Array[Array[org.nlogo.nvm.Reporter]] 
val greports = gargs.map(_.args).filter(a => 
a.exists(_.toString.contains("_callreport")))
val argEdges = greports.map(r => {
val edge = GraphEdge(from=parent, to=r(0).displayName)
println(s"making edge in proc: $parent => $edge")
edge
})
argEdges match {
case e if e.isEmpty => topEdges.toList
case _ => topEdges.toList ::: argEdges.toList
}
}}  
val edgeSet = edges.toList.flatten.toSet
edgeSet foreach println
edgeSet

  }

def run(modelFileName : String, callGraphOutFileName : String) : Unit = {
val workspace = HeadlessWorkspace.newInstance
workspace.open(modelFileName)
val procs = workspace.procedures 
val callGraph = extractCallGraph(procs)
val output = callGraph.map(_.toCSV).toList
Utils.writeFile(callGraphOutFileName, output)
workspace.dispose()
}


} 

On Thursday, June 11, 2020 at 11:41:03 AM UTC-4, Steve Upton wrote:
>
> Shoot, has it been that long??
>
> Your answers, and references, are enough to get me going. I'll parse the 
> Commands since I don't see where to get the FrontEndResults - maybe they 
> have been discarded. Or maybe someone else will pipe in.
>
> thanx again
> steve
>
> On Thursday, June 11, 2020 at 11:20:30 AM UTC-4, Seth Tisue wrote:
>>
>> It's now been more than five years since I worked on NetLogo; there's a 
>> limit to how much I still remember.
>>
>> In particular, I don't remember whether it's possible to get back to the 
>> ASTs after the entire compiler has run, or whether they have already been 
>> discarded and it's too late.
>>
>> I found some prior discussions in this area in the NetLogo room on Gitter:
>>
>> * https://gitter.im/NetLogo/NetLogo?at=5be1a45dda57ff676c9d4305  — "you 
>> want `FrontEndResults` and not `CompilerResults`. "FrontEndResults contains 
>> ProcedureDefinition objects, these have the ASTs inside them. see 
>> AstNode.scala"
>> * https://gitter.im/NetLogo/NetLogo?at=5bba30ccc7bf7c3662cda279
>>
>> But anyway, although my first thought was to walk the ASTs, on second 
>> thought it might be easier and just as effective for your use case to walk 
>> the fully compiled model (the `Array[Command]` inside each procedure) and 
>> look for the `_call` and `_callreport` nodes there. Any `_call`s will be in 
>> the `Array[Command]` directly, but to find the `_callreport`s you'll need 
>> recursively walk the `args` tree under each `Command`.
>>
>> Seth
>>
>

-- 
You received this message because you are subscribed to the Google Groups 
"netlogo-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to netlogo-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/netlogo-devel/af675a50-601e-4a20-8d54-4deb5338c211o%40googlegroups.com.

Re: [netlogo-devel] Procedure call graph

[Steve Upton]

Shoot, has it been that long??

Your answers, and references, are enough to get me going. I'll parse the 
Commands since I don't see where to get the FrontEndResults - maybe they 
have been discarded. Or maybe someone else will pipe in.

thanx again
steve

On Thursday, June 11, 2020 at 11:20:30 AM UTC-4, Seth Tisue wrote:
>
> It's now been more than five years since I worked on NetLogo; there's a 
> limit to how much I still remember.
>
> In particular, I don't remember whether it's possible to get back to the 
> ASTs after the entire compiler has run, or whether they have already been 
> discarded and it's too late.
>
> I found some prior discussions in this area in the NetLogo room on Gitter:
>
> * https://gitter.im/NetLogo/NetLogo?at=5be1a45dda57ff676c9d4305  — "you 
> want `FrontEndResults` and not `CompilerResults`. "FrontEndResults contains 
> ProcedureDefinition objects, these have the ASTs inside them. see 
> AstNode.scala"
> * https://gitter.im/NetLogo/NetLogo?at=5bba30ccc7bf7c3662cda279
>
> But anyway, although my first thought was to walk the ASTs, on second 
> thought it might be easier and just as effective for your use case to walk 
> the fully compiled model (the `Array[Command]` inside each procedure) and 
> look for the `_call` and `_callreport` nodes there. Any `_call`s will be in 
> the `Array[Command]` directly, but to find the `_callreport`s you'll need 
> recursively walk the `args` tree under each `Command`.
>
> Seth
>

-- 
You received this message because you are subscribed to the Google Groups 
"netlogo-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to netlogo-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/netlogo-devel/e5c504e1-ca20-43b0-a0a4-488ec5f6f633o%40googlegroups.com.

Re: [netlogo-devel] Procedure call graph

[Seth Tisue]

I think you will need to walk the ASTs representing the procedure bodies
and look for _call and _callreport nodes.

Seth

-- 
You received this message because you are subscribed to the Google Groups 
"netlogo-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to netlogo-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/netlogo-devel/CAOoO2vgQfVZa9Fjm4M_t_-VP5vcEzOMP1zqjsH28fgw2Xak6sg%40mail.gmail.com.

Re: [netlogo-devel] Procedure call graph

[Steve Upton]

Thank you for the pointer. Hadn't looked at that yet. The profiler 
extension is a start, but it only gives me the calls to the procedures, not 
what other procedures called that procedure, etc., in a graph. From your 
pointer, I found the Reflection-Extension, which looked like it produce 
what I want, but if I'm reading it correctly, I'd have to put calls to 
"reflection:procedure" and "reflection:callers" in each procedure. I was 
hoping for something more like a static analysis of the netlogo code.

Seth Tissue had posted code back in 2012 that got me started, and I had 
been looking at the procedures map that is returned. I can get a dump but 
don't see how I can just get the children's names (i.e., the callers). With 
the dump,  I can at least parse that to get what I'm looking for.

thanx
steve

On Tuesday, June 9, 2020 at 8:45:40 PM UTC-4, Forrest wrote:
>
> You might take a look at the profiler extension (or source code thereof), 
> if you haven't already...
>
> https://github.com/NetLogo/Profiler-Extension
>
> Cheers,
>
> Dr. Forrest Stonedahl
> Computer Science
> Augustana College
>
> On Tue, Jun 9, 2020, 11:39 AM Steve Upton > 
> wrote:
>
>> Hi All,
>>
>> Fairly new to Netlogo. We have a good sized model and would like to 
>> create a procedure call graph, i.e., for each procedure, identify what 
>> procedures it calls, and so forth. I've found a post on reflection and can 
>> get the procedures (as well as globals, turtles, breeds), so about to 
>> explore further, but wondered if this wasn't already available.
>>
>> thank you
>> steve upton
>>
>> -- 
>> You received this message because you are subscribed to the Google Groups 
>> "netlogo-devel" group.
>> To unsubscribe from this group and stop receiving emails from it, send an 
>> email to netlog...@googlegroups.com .
>> To view this discussion on the web visit 
>> https://groups.google.com/d/msgid/netlogo-devel/4cbab5a6-b52a-48d5-87fa-e272397c4e7bo%40googlegroups.com
>>  
>> 
>> .
>>
>

-- 
You received this message because you are subscribed to the Google Groups 
"netlogo-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to netlogo-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/netlogo-devel/e35605f1-2fa9-488c-9675-fa540073d462o%40googlegroups.com.

Re: [netlogo-devel] Procedure call graph

[Forrest Stonedahl]

You might take a look at the profiler extension (or source code thereof),
if you haven't already...

https://github.com/NetLogo/Profiler-Extension

Cheers,

Dr. Forrest Stonedahl
Computer Science
Augustana College

On Tue, Jun 9, 2020, 11:39 AM Steve Upton  wrote:

> Hi All,
>
> Fairly new to Netlogo. We have a good sized model and would like to create
> a procedure call graph, i.e., for each procedure, identify what procedures
> it calls, and so forth. I've found a post on reflection and can get the
> procedures (as well as globals, turtles, breeds), so about to explore
> further, but wondered if this wasn't already available.
>
> thank you
> steve upton
>
> --
> You received this message because you are subscribed to the Google Groups
> "netlogo-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to netlogo-devel+unsubscr...@googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/netlogo-devel/4cbab5a6-b52a-48d5-87fa-e272397c4e7bo%40googlegroups.com
> 
> .
>

-- 
You received this message because you are subscribed to the Google Groups 
"netlogo-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to netlogo-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/netlogo-devel/CAG_mutPoXYOAdiZUr3-WAi9Bp%3DQy6MXp5r0fxju5jTtUUC_QLg%40mail.gmail.com.