h.n3d() returns 0

When Python is the interpreter, what is a good
design for the interface to the basic NEURON
concepts.

Moderator: hines

ioannisv

h.n3d() returns 0

Post by ioannisv » Tue Sep 03, 2013 2:25 am

Hi

I have followed your instructions about the temperature dependence.
(moderator's comment: see viewtopic.php?f=15&t=2107)
The problem is that every time I compile the mod file the h.n3d() function returns 0
I thought it was because I did something wrong but apparently every file that I compile returns zero. I even took some mod and hoc files from the modeldb database, like the original fh.mod and the are not compiling correctly.
Any idea what might be not working correctly?

Thanks!

ted
Site Admin
Posts: 5632
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: h.n3d() returns 0

Post by ted » Tue Sep 03, 2013 9:32 am

ioannisv wrote:every time I compile the mod file the h.n3d() function returns 0
n3d() has nothing to do with any mod file. Was there something in this thread that suggested otherwise? For n3d() to return a nonzero value, a section's geometry must either have been specified with pt3dadd statements, or define_shape() must have been called. For more information about these and related facts, please read the Programmer's Reference documentation regarding Geometry.
http://www.neuron.yale.edu/neuron/stati ... l#Geometry
I even took some mod and hoc files from the modeldb database, like the original fh.mod and the are not compiling correctly.
This sounds like a serious problem. Specifically what files did you evaluate, and what was the evidence that they "are not compiling correctly"?

ioannisv

Re: h.n3d() returns 0

Post by ioannisv » Tue Sep 03, 2013 9:44 am

I am more than shure that I am defining geometries in the hoc file.


I compiled both the original Frakenhauser Huxley (fh.mod) and the original Hodgkin Huxley (hh.mod) from the database.

After loading the dll and the hoc files every time when I called the h.n3d() command it kept returning zero.

ted
Site Admin
Posts: 5632
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: h.n3d() returns 0

Post by ted » Tue Sep 03, 2013 10:02 am

ioannisv wrote:I am more than shure that I am defining geometries in the hoc file.
That's not the issue. The issue is whether you defined section geometry by executing pt3dadd statements or simply by assigning values to L and diam (and then didn't call define_shape()). Suggest you read the material that I mentioned.
I compiled both the original Frakenhauser Huxley (fh.mod) and the original Hodgkin Huxley (hh.mod) from the database.

After loading the dll and the hoc files every time when I called the h.n3d() command it kept returning zero.
That's like saying "I put air in the tires, but the gas tank is still empty." Code in mod files is used to add new biophysical mechanisms to NEURON, or to define new classes of artificial spiking cells, or to add arbitrary C or NMODL code. It has no effect on the specification of geometry.

ioannisv

Re: h.n3d() returns 0

Post by ioannisv » Tue Sep 03, 2013 10:21 am

pt3dadd is called in my hoc file:

Code: Select all

proc basic_shape() { local x, y, z
	x = -0.5*(_axoninter*_Li - _w)
	y = _dEl
	z = 0
	for i=0, _axonnodes-1 {
		internode[i] { pt3dclear()
			pt3dadd(x, y, z, _d)
			x = x + _Li - _w
			pt3dadd(x, y, z, _d)
		}
		node[i] { pt3dclear()
			pt3dadd(x, y, z, _d)
			x = x + _w
			pt3dadd(x, y, z, _d)
		}
	}
	internode[_axonnodes] { pt3dclear()
		pt3dadd(x, y, z, _d)
		x = x + _Li - _w
		pt3dadd(x, y, z, _d)
	}
}
_axonnodes is defined as a predetermined variable of value 9, _axoninter = _axonnodes+1

ted
Site Admin
Posts: 5632
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: h.n3d() returns 0

Post by ted » Tue Sep 03, 2013 10:45 am

Looks like those sections should have nonzero n3d(). How are you calling n3d()?

ioannisv

Re: h.n3d() returns 0

Post by ioannisv » Tue Sep 03, 2013 11:06 am

Code: Select all

from neuron import h
Section = h.Section

h.load_file(0, "C:\nrn73w64\lib\hoc\stdlib.hoc")
h.nrn_load_dll("C:/Users/Ioannis/Desktop/fh(T)/nrnmech.dll")
h.load_file(1,"C:/Users/Ioannis/Desktop/fhT)/fht.hoc")
nn = h.n3d()

ted
Site Admin
Posts: 5632
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: h.n3d() returns 0

Post by ted » Tue Sep 03, 2013 11:05 pm

I really don't know what to say, except that it seems likely that the geometry of the default section that your hoc code creates was specified with the stylized method and therefore has no 3d points. Here is a sequence of commands I just now executed, along with NEURON's replies:

Code: Select all

>>> from neuron import h
>>> h.load_file("cell.hoc") # has many sections, all geometries specified with 3d points
>>> h.psection()
soma { nseg=1  L=37.101  Ra=35.4
	/*location 0 attached to cell 0*/
	/* First segment only */
	insert morphology { diam=22.0682}
	insert capacitance { cm=1}
}
1.0
>>> h.n3d() # I already know that soma has 28 pt3d points
28.0
>>> for sec in h.allsec():
...   print h.secname(), h.n3d()
... 
soma 28.0
dendrite_8 38.0
dendrite_7[0] 6.0
dendrite_7[1] 7.0
 . . . etc. usw ad not quite infinitum . . .
It would be quite a surprise to discover that someone has managed to create a mod file, written entirely in NMODL code and including no vicous C statements in VERBATIM blocks, that destroys pt3d data. Have you examined the hoc file that contains the statements that define your model cell's geometry? Is the problem perhaps that there is no default section (in which case I would expect h.psection() to return an error message).

ioannisv

Re: h.n3d() returns 0

Post by ioannisv » Wed Sep 04, 2013 2:16 am

The thing is that I have some files from previous people and they are working, they are not returning zero.
In the beginning I thought I had something wrong with my hoc/mod files but then when I recompiled the files that were working before they are now returning zero as well
That's why I said at my first post about compiling "error".
Thanks anyway.

ted
Site Admin
Posts: 5632
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: h.n3d() returns 0

Post by ted » Wed Sep 04, 2013 11:06 am

Something is definitely wrong, but it looks like I'll have to be able to reproduce the problem myself in order to advise you properly. Could you please zip up just enough source code (hoc, mod, ses, py files) to reproduce the problem and email it to ted dot carnevale at yale dot edu?

ioannisv

Re: h.n3d() returns 0

Post by ioannisv » Thu Sep 05, 2013 4:49 am

They are in your email with subject h.n3d()
Thanks

ted
Site Admin
Posts: 5632
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: h.n3d() returns 0

Post by ted » Thu Sep 05, 2013 12:44 pm

First I just compiled the mod files and executed the hoc file that defines the model cell.

Code: Select all

[ted@loki ioannis]$ nrngui SENNT.hoc 
NEURON -- Release 7.3 (927:f3dcf53fbe74) 2013-07-27
Duke, Yale, and the BlueBrain Project -- Copyright 1984-2013
See http://www.neuron.yale.edu/neuron/credits

loading membrane mechanisms from /home/ted/myhoc/ioannis/i686/.libs/libnrnmech.so
Additional mechanisms from files
 SENNT.mod
Note that there was no need for load_dll because I was in the same directory as the hoc, py, and compiled mod file, so NEURON automatically knew where to find the compiled mechanisms.

Next I verified that a default section exists and that all sections have pt3d data.

Code: Select all

oc>secname()
node[0]
oc>forall print secname(), " ", n3d()
node[0] 2 
node[1] 2 
node[2] 2 
node[3] 2 
node[4] 2 
node[5] 2 
node[6] 2 
node[7] 2 
node[8] 2 
internode[0] 2 
internode[1] 2 
internode[2] 2 
internode[3] 2 
internode[4] 2 
internode[5] 2 
internode[6] 2 
internode[7] 2 
internode[8] 2 
internode[9] 2 
Next I used test.py. Since I was running NEURON from the directory that contained the hoc, py, and compiled mod file, I changed test.py to

Code: Select all

from neuron import h
Section = h.Section
# h.load_file("C:/nrn73w64/lib/hoc/stdlib.hoc")
# h.nrn_load_dll("C:/Users/Ioannis/Desktop/SENN(T)/nrnmech.dll")
# h.load_file(1,"C:/Users/Ioannis/Desktop/SENN(T)/SENNT.hoc")
h.load_file("SENNT.hoc")
print h.n3d()
After making those changes, I executed test.py by

Code: Select all

[ted@loki ioannis]$ nrngui -python test.py
NEURON -- Release 7.3 (927:f3dcf53fbe74) 2013-07-27
Duke, Yale, and the BlueBrain Project -- Copyright 1984-2013
See http://www.neuron.yale.edu/neuron/credits

loading membrane mechanisms from /home/ted/myhoc/ioannis/i686/.libs/libnrnmech.so
Additional mechanisms from files
 SENNT.mod
2.0
So h.n3d() returned a nonzero value.

Next I wanted to double check that the default section existed and was known to Python

Code: Select all

>>> h.secname()
'node[0]'
so the answers are yes and yes.

Finally, I wanted to verify that all sections had pt3d data

Code: Select all

>>> for sec in h.allsec():
...   print h.secname(), h.n3d()
... 
node[0] 2.0
node[1] 2.0
node[2] 2.0
node[3] 2.0
node[4] 2.0
node[5] 2.0
node[6] 2.0
node[7] 2.0
node[8] 2.0
internode[0] 2.0
internode[1] 2.0
internode[2] 2.0
internode[3] 2.0
internode[4] 2.0
internode[5] 2.0
internode[6] 2.0
internode[7] 2.0
internode[8] 2.0
internode[9] 2.0
and indeed they do.

So now the question is why your experience was different. Are you sure that NEURON loaded your dll (should have printed a msg like "additional mechanisms from ...") and your hoc code? Were there any error messages, like "section access not specified"?

ioannisv

Re: h.n3d() returns 0

Post by ioannisv » Mon Sep 09, 2013 5:18 am

running the functions through hoc returned 2
running the functions through python when trying to load the compiled dll file the console now returns
Ladlibrary failed with error 193 dlopen failed

ted
Site Admin
Posts: 5632
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: h.n3d() returns 0

Post by ted » Mon Sep 09, 2013 10:09 am

Exactly what is the first line that NEURON prints to its terminal when it starts? Should look something like this:
NEURON -- Release 7.3 (927:f3dcf53fbe74) 2013-07-27

Why do you have to explicity load the dll? Aren't your hoc and py files in the same directory as the dll, so that NEURON loads the dll automatically? Does NEURON print a statement that looks like this?

Code: Select all

loading membrane mechanisms from . . . [name of some file here]
Additional mechanisms from files
SENNT.mod
(should be a few lines after the one that tells which release of NEURON you're using)

ioannisv

Re: h.n3d() returns 0

Post by ioannisv » Wed Sep 11, 2013 3:11 am

I finally figured it out.
It was a conflict with my pythonpath, the neuron module imported and the python platform I was using.
Thanks for all the help.

Post Reply