ioam / topographica Goto Github PK

Converted from SourceForge issue 1752461, submitted by ceball
Submit Date: 2007-07-12 08:04 GMT

The minimum needs to be calculated independently for the
Sheet and non-Sheet coordinate cases. Somewhat tricky due to how
_calculate_minimum_height_of_tallest_plot() also sets minimum_height;
it all needs to be cleaned up so that the actual minimum is enforced,
so that plots don't end up being so large.

Converted from SourceForge issue 1647715, submitted by ceball
Submit Date: 2007-01-30 08:24 GMT

On Mac OS X, plots appear to be displaced vertically; that is, there is a gap at the top of the canvas before the plot starts.

Additionally, the right-click-coordinates calculation does not return the correct results. It appears to be incorrect by an amount larger than the gap at the top.

We aren't doing anything differently on the Mac (it's just the standard unix version of Topographica), so maybe it is a bug in Tk?

Please see the attached screenshot for an example of the problem from the test pattern window.

Converted from SourceForge issue 1884341, submitted by jbednar
Submit Date: 2008-02-01 11:29 GMT

In a plot window that is much larger than the screen, the entire window scrolls when the horizontal scrollbar is used. For most of the content that is fine, but the dynamic information in the status bar at the bottom should not scroll, because there's no way to see the dynamic info for items that are off the screen to the right. Once the scrollbar has been moved over there, the status bar is no longer visible. Maybe the status bar can somehow be fixed to the size of the actual window only?

Converted from SourceForge issue 1771197, submitted by ceball
Submit Date: 2007-08-09 20:09 GMT

Clicking the "Strength only" on a PlotGroupPanel undoes any resizing that has been done with Enlarge or Reduce.

I suspect this is because Enlarge and Reduce are implemented in tkgui rather than in plotting (but there is a HACKALERT about the implementation of strength-only itself...).

Converted from SourceForge issue 1350353, submitted by jbednar
Submit Date: 2005-11-07 14:49 GMT

If I do ./topographica -g examples/hierarchical.ty, the
Projection plot for V3 shows an array of Afferent
ConnectionFields, but the Unit Weights plot is blank,
with no Projection selectable.

Converted from SourceForge issue 1370157, submitted by ceball
Submit Date: 2005-11-30 15:42 GMT

Sometimes a new browser window opens, sometimes a new
browser opens, sometimes the browser's current webpage
changes to the Topographica one (which is annoying).

I think it should probably:

• open a new window on the current browswer, if one exists
• open a new browser if one isn't already open

Converted from SourceForge issue 1876245, submitted by nobody
Submit Date: 2008-01-21 06:43 GMT

Should profile the plotting code to see why it's so slow.

e.g.
topo.plotting.bitmap.HSVBitmap contains a loop which should probably be woven in C.

Converted from SourceForge issue 1479591, submitted by tookaway
Submit Date: 2006-05-01 05:22 GMT

connectionfield.py line 384:
"cf = cfs[cols/2][rows/2]" should be "cf =
cfs[rows/2][cols/2]"

Converted from SourceForge issue 1020010, submitted by jprovost
Submit Date: 2004-08-31 21:08 GMT

This is a test bug. It's not really a bug. Really.
It's not.

Converted from SourceForge issue 1752465, submitted by ceball
Submit Date: 2007-07-12 08:12 GMT

A gradient of the rectangular Retina Activity bitmaps in a
hierarchical.ty Activity plot gives an error:

Error: 2
ValueError Exception in Tk callback
Function: <function at 0xb72ec764> (type: <type 'function'>)
Args: ()
Traceback (innermost last):
File "/disk/home/lodestar1/jbednar/topographica//lib/python2.4/site-packages/Pmw/Pmw_1_2/lib/PmwBase.py", line 1747, in call
return apply(self.func, args)
File "/disk/home/lodestar1/jbednar/topographica/topo/tkgui/templateplotgrouppanel.py", line 180, in
command=lambda: self.__gradient('Strength'))
File "/disk/home/lodestar1/jbednar/topographica/topo/tkgui/templateplotgrouppanel.py", line 300, in __gradient
topo.commands.pylabplots.gradientplot(m,title="Gradient: " + description)
File "/disk/home/lodestar1/jbednar/topographica/topo/commands/pylabplots.py", line 137, in gradientplot
matrixplot((cos(dx2pi)+cos(dy2pi)),title=title)
ValueError: shape mismatch: objects cannot be broadcast to a single shape

Is one of the arguments swapped?

There's also the potential of a divide-by-zero error (noted in the code). In fact, a gradient of the rectangular Retina Activity bitmaps in a
hierarchical.ty Activity plot gives:

Error: 2
ValueError Exception in Tk callback
Function: <function at 0xb72ec764> (type: <type 'function'>)
Args: ()
Traceback (innermost last):
File "/disk/home/lodestar1/jbednar/topographica//lib/python2.4/site-packages/Pmw/Pmw_1_2/lib/PmwBase.py", line 1747, in call
return apply(self.func, args)
File "/disk/home/lodestar1/jbednar/topographica/topo/tkgui/templateplotgrouppanel.py", line 180, in
command=lambda: self.__gradient('Strength'))
File "/disk/home/lodestar1/jbednar/topographica/topo/tkgui/templateplotgrouppanel.py", line 300, in __gradient
topo.commands.pylabplots.gradientplot(m,title="Gradient: " + description)
File "/disk/home/lodestar1/jbednar/topographica/topo/commands/pylabplots.py", line 137, in gradientplot
matrixplot((cos(dx2pi)+cos(dy2pi)),title=title)
ValueError: shape mismatch: objects cannot be broadcast to a single shape

Converted from SourceForge issue 1663345, submitted by ceball
Submit Date: 2007-02-19 10:39 GMT

Pressing 'Reset' with a Composite pattern causes the list of generators to disappear.

Converted from SourceForge issue 1881582, submitted by ceball
Submit Date: 2008-01-29 04:31 GMT

Check situate works normally:

./topographica -c default_density=8 -a examples/cfsom_or.ty -c "save_plotgroup('Projection',projection=topo.sim['V1'].projections('Afferent'),situate=False)"

Save a snapshot:

./topographica -c default_density=8 -a examples/cfsom_or.ty -c "save_snapshot('test.typ')"

Repeat first test, but with snapshot:

./topographica -a -c "load_snapshot('test.typ')" -c "save_plotgroup('Projection',projection=topo.sim['V1'].projections('Afferent'),situate=False)"

The second time, the plots are situated (but shouldn't be).

Same happens in the GUI.

Converted from SourceForge issue 1856728, submitted by nobody
Submit Date: 2007-12-23 08:02 GMT

Executing the following two commands:

$ ./topographica examples/hierarchical.ty -a -c "save_script_repr('hierarchical_script_repr.ty')"

$ ./topographica hierarchical_script_repr.ty

leads to lots of warnings about being unable to set lbound and ubound parameters on various UniformRandom instances.

The problem is caused by a name clash:

$ grep "UniformRandom" hierarchical_script_repr.ty
from topo.misc.numbergenerators import UniformRandom
from topo.patterns.random import UniformRandom
...

Converted from SourceForge issue 1760374, submitted by ceball
Submit Date: 2007-07-25 15:16 GMT

With psyco enabled, 'make tests' gives a segmentation fault.

(More information in 'Tests failing?' e-mails between JAB and CB.)

Converted from SourceForge issue 1508786, submitted by nobody
Submit Date: 2006-06-19 17:24 GMT

I tried to install Topographica0.9.0 under debian-linux,
and got this error-message from make:

...
checking for connect... yes
checking for gethostbyname... yes
checking how to build libraries... shared
checking for ranlib... ranlib
checking if 64bit support is requested... no
checking if 64bit Sparc VIS support is requested... no
checking system version (for dynamic loading)...
./configure: line 7338: syntax error near unexpected
token )' ./configure: line 7338: OSF*)'
make[1]: *** [tcl] Error 2
make[1]: Leaving directory
`/home/agdp/download/Med/topographica-0.9.0/external'
make: *** [ext-packages] Error 2
<<<

"uname -a" yields:

Linux Kanotix1 2.6.16.5-kanotix-2 #1 SMP PREEMPT Sat
Apr 15 20:20:23 CEST 2006 i686 GNU/Linux
<<<

-HaJo Gurt

Converted from SourceForge issue 1821527, submitted by ceball
Submit Date: 2007-10-28 09:13 GMT

The slider is set, but the binding to "Go" has been lost.

Converted from SourceForge issue 1451421, submitted by ceball
Submit Date: 2006-03-16 16:25 GMT

Once you have started to enter a command at the
command prompt (not in the Topographica console, but
at the command prompt), the GUI is frozen.

The GUI remains frozen until the command is entered
(or deleted, and return is pressed).

E.g. type a space at the command prompt. Now you
can't interact with the Topographica console.
Continue to wait. After a while the Topographica
console window will show '[Not Responding]' in the
title bar, making it look like Topographica has
crashed. Press enter at the command prompt and the
window will become active again.

Converted from SourceForge issue 1535580, submitted by tookaway
Submit Date: 2006-08-06 22:30 GMT

I find that the external\mswin\util directory of 0.9.1
is empty, so the setup under MSWin would always fail.

Converted from SourceForge issue 1875377, submitted by nobody
Submit Date: 2008-01-19 17:59 GMT

If I save a snapshot s, close Topographica, define a new plotgroup "Direction Preference" in topo.commands.analysis, and then load snapshot s, I get:

...
File "/home/chris/topographica/topo/tkgui/topoconsole.py", line 130, in call
new_plotgroup.plot_templates = topo.plotting.plotgroup.plotgroups[self.plotgroup.name].plot_templates
File "/home/chris/topographica/topo/misc/keyedlist.py", line 69, in getitem
raise KeyError(key)
KeyError: 'Direction Preference'

Presumably related to task in Future Work regarding cleanup of plotgroups dictionary.

Converted from SourceForge issue 1877167, submitted by ceball
Submit Date: 2008-01-22 10:57 GMT

./topographica -g -c "topo.guimain['Simulation']'Command prompt'"

Type the following, followed by to see one of the possible errors:
1

(I can't paste the error here because I can't copy it from the Command Prompt box.)

Please can we just remove the Command Prompt? Does anyone use it? My guess is nobody uses it.

Maybe there should be a feature request "Embed IPython in tkgui", although I am not sure it would ever be done. It looks like IPython is being split up into 'backend' and 'frontend' (http://projects.scipy.org/ipython/ipython/wiki/NewDesign/FrontEnds),
which would help.

Converted from SourceForge issue 1806249, submitted by ceball
Submit Date: 2007-10-02 10:48 GMT

The images on buttons are not centered.

If you load the .png images into a viewer like xv and zoom in, you'll see that they are very nearly centered, but off by some fraction of a pixel. At most they should be off by a pixel, but it seems like Tkinter has shifted them by more than that.

Converted from SourceForge issue 1829843, submitted by ceball
Submit Date: 2007-11-11 08:28 GMT

Strange Projection plots.

Converted from SourceForge issue 1433073, submitted by chrispalmer
Submit Date: 2006-02-16 18:55 GMT

For two sheets with same bounds but different
densities, the sheet with the higher density is larger
on the activity plot, when they should be the same size.

And on the orientation plot the problem is the same but
reversed.

Converted from SourceForge issue 1504496, submitted by ceball
Submit Date: 2006-06-11 21:46 GMT

When I load a saved snapshot, the default prefix for
the lookup path (i.e. the topographica root
directory) is not updated.

e.g.
I load a network saved from one machine onto another,
where topographica is in a different directory. Then
I open a test pattern window and choose image. I get
an error that the file can't be found.

Converted from SourceForge issue 1910078, submitted by ceball
Submit Date: 2008-03-08 13:53 GMT

Editing a parameter in the GUI has no effect on the parameterizedobject.

e.g.
$ ./topographica -c default_density=2 examples/lissom_oo_or.ty -g

topo> from topo.tkgui.parametersframe import edit_parameters
topo> edit_parameters(topo.sim['V1'])

Changing a parameter in the GUI has no effect on topo.sim['V1']!

The tests are passing, indicating that tkparameterizedobject itself is still working.

Converted from SourceForge issue 1600059, submitted by fdemertzis
Submit Date: 2006-11-20 22:53 GMT

On the model editor when trying to add a new projection OR a new sheet, if the user closed the new object parameters window it will throw a confusing tk callback exception. Normally should be handled a bit better.

Converted from SourceForge issue 1836828, submitted by ceball
Submit Date: 2007-11-23 04:48 GMT

make slow-tests includes speed tests to check that topographica's performance does not get worse, but currently these results are only returned as information (i.e. do not generate test failures if the performance worsens). The reason for that is because the python timing functions used to not seem to compensate well for processor activity (i.e. if the computer is also doing something else, the timings become much worse).

We need to check that the correct functions are being used*, and then uncomment the error-raising code: see the ALERT in topo/tests/test_script.py, line 194 ("whatever compensations the python timing functions are supposed to make [...]").

* and decide what to do about garbage collection: see note in topo/tests/test_script.py, line 142 ("not clear whether to turn off GC?")

Converted from SourceForge issue 1451415, submitted by ceball
Submit Date: 2006-03-16 16:21 GMT

On Windows, the learning iteration counter sometimes
doesn't appear (stays showing 'Status: OK'), and
sometimes freezes part-way through a group of
learning iterations. In both cases Topographica
continues to run, but appears to have frozen.

This problem appears to depend on what else the user
is doing at the same time (e.g. clicking around
opening other windows, dragging windows across the
Topographica console, etc.).

Converted from SourceForge issue 1855882, submitted by jprovost
Submit Date: 2007-12-21 17:54 GMT

How to hang the topo GUI:

1. topographica examples/lissom_or.ty -g
2. "Run for" 20000 from console.
3. While running, go to shell prompt and press ENTER.
4. Execution stops, GUI is fscked.

Works on Mac and Linux.

Converted from SourceForge issue 1479594, submitted by tookaway
Submit Date: 2006-05-01 05:30 GMT

When the simulator contains events added by
schedule_action(), it can't be saved as a snapshot.

Converted from SourceForge issue 1831123, submitted by rczhao
Submit Date: 2007-11-13 15:07 GMT

line 48, topo/patterns/image.py,

if pattern_array and image

if pattern_array or image is an numpy array, this comparison may return actual value of them, rather than the truth value, causing error:

ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

Converted from SourceForge issue 1848915, submitted by jprovost
Submit Date: 2007-12-11 21:19 GMT

CFProjectionPanel.populate_projection_param raises an error when a script contains a CFSheet that doesn't have any incoming CFProjections, preventing the viewing of an Projections for any Sheets.

In my case, I have an instance of topo.sheets.saccade.SaccadeController (a subclass of CFSheet) that doesn't have any incoming CFProjections but does have a OneToOneProjection, as well as another incoming connection that's not a projection. It causes the crash because of these lines:

    prjns = \[x for x in self\.plotgroup\.sheet\.projections\(\)\.values\(\)
             if isinstance\(x,CFProjection\)\]
    self\.plotgroup\.params\(\)\['projection'\]\.objects = prjns
    self\.plotgroup\.projection = prjns\[0\] \# CB: necessary?

The variable prjns refers to an empty list.

I suppose that CFSheets should only be included in the ProjectionPanel if they have at least one incoming CFProjection. Or maybe we need the GUI to support OneToOneProjections.

Converted from SourceForge issue 1864419, submitted by ceball
Submit Date: 2008-01-05 10:37 GMT

In a ParametersFrame, clearing out the value of a FixedPoint number causes an error.

This happens because the GUI automatically checks to see if a Parameter has changed, but passing an empty string to FixedPoint (for the comparison) causes an error (FixedPoint can't represent the value of an empty string).

(E.g. Edit properties of tiny.ty's projection and clear out the 'delay' field.)

Converted from SourceForge issue 1862101, submitted by ceball
Submit Date: 2008-01-02 10:53 GMT

Run this code:

./topographica -g examples/hierarchical.ty -c "p=topo.guimain['Plots']'Projection' ; p.sheet=topo.sim['V2'] ; p.Refresh()"

to see that the number of bitmaps seems confused.

Mismatch between the number of CFs returned to the projection arranging code and the number it expects?

Converted from SourceForge issue 1370075, submitted by ceball
Submit Date: 2005-11-30 14:15 GMT

On the unit weights plot for a BoundingBox of -0.5 to 0.5,
typing -0.5 works, typing 0.5 gives an error. For 0.5
it's an
IndexError. I think that's ok. Typing something like
0.6 gives a
really cryptic error: "NameError: global name 'Pmw' is
not defined".

- We should have an informative error message for
coordinates out of bounds.

- We should be enforcing the bounds in the GUI -- there
are strict
bounds, and we can just make sure that the user can't
set any value outside of them.

Converted from SourceForge issue 1359195, submitted by arobert3434
Submit Date: 2005-11-17 18:49 GMT

CVS checkout 2005/11/17 w/LATEST_STABLE tag on Mac OS X 10.4.3
with Python 2.3.5. Type "make" in top-level directory, all goes
smoothly until around 230 MB have been accumulated under
topographica/external, and the package
Scipy_core-0.3.0_108.1820.tar.gz is reached, at which point:

make -C external
tar xzf Scipy_core-0.3.0_108.1820.tar.gz
cd Scipy_core-0.3.0_108.1820; env PREFIX=/Users/arobert/src/
topographica/ LD_LIB
RARY_PATH=/Users/arobert/src/topographica//lib /Users/arobert/
src/topographica//
bin/python setup.py install
sem_init: Function not implemented
sem_wait: Bad file descriptor
sem_post: Bad file descriptor
sem_wait: Bad file descriptor
sem_post: Bad file descriptor
sem_init: Function not implemented

After a few hundred of these, it ends in:

Traceback (most recent call last):
File "setup.py", line 17, in ?
from scipy_distutils.core import setup
File "/Users/arobert/src/topographica/external/
Scipy_core-0.3.0_108.1820/scipy_distutils/init.py", line 15, in ?
import ccompiler
File "/Users/arobert/src/topographica/external/
Scipy_core-0.3.0_108.1820/scipy_distutils/ccompiler.py", line 13, in ?
from exec_command import exec_command
File "/Users/arobert/src/topographica/external/
Scipy_core-0.3.0_108.1820/scipy_distutils/exec_command.py", line
53, in ?
import tempfile
File "/Users/arobert/src/topographica//lib/python2.4/tempfile.py",
line 83, in ?
_once_lock = _allocate_lock()
thread.error: can't allocate lock
sem_wait: Bad file descriptor
sem_post: Bad file descriptor
sem_wait: Bad file descriptor
sem_post: Bad file descriptor
sem_wait: Bad file descriptor
sem_post: Bad file descriptor
sem_wait: Bad file descriptor
sem_post: Bad file descriptor
make[1]: *** [weave] Error 1
make: *** [ext-packages] Error 2

Converted from SourceForge issue 1872761, submitted by ceball
Submit Date: 2008-01-16 13:58 GMT

E.g. on Windows, sometimes (by chance) Topographica's console opens behind the command prompt. This makes it look as though the GUI as not opened.

Should add something to raise the console to the front on opening.

Converted from SourceForge issue 1758936, submitted by antolikjan
Submit Date: 2007-07-23 15:45 GMT

I noticed that the measure_rfs() routine works strangely. After I run it on my model (which is big but still should be reasonably in memory) the machine starts to swap like crazy without seeing any progress with the measurment, and this goes forever until the machine completely hangs up.

Converted from SourceForge issue 1877068, submitted by ceball
Submit Date: 2008-01-22 08:28 GMT

I just tried on Sake, with and without osx-patch.

Without osx-patch:

sake:topographica ceball$ svn update
At revision 7684.
sake:topographica ceball$ make
make -C external
tar -xozf tcl8.4.13-src.tar.gz
cd tcl8.4.13/unix; ./configure --prefix=/Users/ceball/topographica/
creating cache ./config.cache
[...]
touch tk
echo /Users/ceball/topographica/
/Users/ceball/topographica/
tar -xozf Python-2.5.1.tgz
cd Python-2.5.1; ./configure --prefix=/Users/ceball/topographica/ LDFLAGS=-Wl,-rpath,/Users/ceball/topographica//lib
checking MACHDEP... darwin
[...]
gcc -c -fno-strict-aliasing -Wno-long-double -no-cpp-precomp -mno-fused-madd -DNDEBUG -g -O3 -Wall -Wstrict-prototypes -I. -I./Include -DPy_BUILD_CORE -o Python/mysnprintf.o Python/mysnprintf.c
gcc -c -fno-strict-aliasing -Wno-long-double -no-cpp-precomp -mno-fused-madd -DNDEBUG -g -O3 -Wall -Wstrict-prototypes -I. -I./Include -DPy_BUILD_CORE -o Parser/tokenizer_pgen.o Parser/tokenizer_pgen.c
gcc -c -fno-strict-aliasing -Wno-long-double -no-cpp-precomp -mno-fused-madd -DNDEBUG -g -O3 -Wall -Wstrict-prototypes -I. -I./Include -DPy_BUILD_CORE -o Parser/printgrammar.o Parser/printgrammar.c
gcc -c -fno-strict-aliasing -Wno-long-double -no-cpp-precomp -mno-fused-madd -DNDEBUG -g -O3 -Wall -Wstrict-prototypes -I. -I./Include -DPy_BUILD_CORE -o Parser/pgenmain.o Parser/pgenmain.c
gcc -DNDEBUG -g -O3 -Wall -Wstrict-prototypes -Wl,-rpath,/Users/ceball/topographica//lib Parser/acceler.o Parser/grammar1.o Parser/listnode.o Parser/node.o Parser/parser.o Parser/parsetok.o Parser/bitset.o Parser/metagrammar.o Parser/firstsets.o Parser/grammar.o Parser/pgen.o Objects/obmalloc.o Python/mysnprintf.o Parser/tokenizer_pgen.o Parser/printgrammar.o Parser/pgenmain.o -ldl -o Parser/pgen
ld: -rpath can only be used when targeting Mac OS X 10.5 or later

collect2: ld returned 1 exit status
make[2]: *** [Parser/pgen] Error 1
make[1]: *** [python] Error 2
make: *** [ext-packages] Error 2

With osx-patch:

sake:topographica ceball$ make
make -C external
echo /Users/ceball/topographica/
/Users/ceball/topographica/
tar -xozf Python-2.5.1.tgz
cd Python-2.5.1; ./configure --prefix=/Users/ceball/topographica/
checking MACHDEP... darwin
[...]
gcc -c -fno-strict-aliasing -Wno-long-double -no-cpp-precomp -mno-fused-madd -DNDEBUG -g -O3 -Wall -Wstrict-prototypes -I. -I./Include -DPy_BUILD_CORE -o Modules/main.o Modules/main.c
Modules/main.c:186: warning: function declaration isn't a prototype
gcc -c -fno-strict-aliasing -Wno-long-double -no-cpp-precomp -mno-fused-madd -DNDEBUG -g -O3 -Wall -Wstrict-prototypes -I. -I./Include -DPy_BUILD_CORE -o Modules/gcmodule.o Modules/gcmodule.c
gcc -fno-strict-aliasing -Wno-long-double -no-cpp-precomp -mno-fused-madd -DNDEBUG -g -O3 -Wall -Wstrict-prototypes -I. -I./Include -DPy_BUILD_CORE -c ./Modules/threadmodule.c -o Modules/threadmodule.o
gcc -fno-strict-aliasing -Wno-long-double -no-cpp-precomp -mno-fused-madd -DNDEBUG -g -O3 -Wall -Wstrict-prototypes -I. -I./Include -DPy_BUILD_CORE -c ./Modules/signalmodule.c -o Modules/signalmodule.o
gcc -fno-strict-aliasing -Wno-long-double -no-cpp-precomp -mno-fused-madd -DNDEBUG -g -O3 -Wall -Wstrict-prototypes -I. -I./Include -DPy_BUILD_CORE -c ./Modules/posixmodule.c -o Modules/posixmodule.o
./Modules/posixmodule.c: In function 'posix_setpgrp':
./Modules/posixmodule.c:3724: error: too few arguments to function 'setpgrp'
make[2]: *** [Modules/posixmodule.o] Error 1
make[1]: *** [python] Error 2
make: *** [ext-packages] Error 2

Converted from SourceForge issue 1465270, submitted by ceball
Submit Date: 2006-04-05 19:37 GMT

If I run Topographica without "-g" (i.e.
"./topographica") and later do something like "from
topo.commands.pylabplots import *", Topographica exits
and I get the following message:
"""
No module named pygtk
PyGTK version 1.99.16 or greater is required to run the
GTK Matplotlib backends
"""

I think just a warning should be printed.

Converted from SourceForge issue 1864835, submitted by tcf
Submit Date: 2008-01-06 01:19 GMT

Topographica sounds great, but I am not able to get it to perform as advertised. I downloaded version 0.9.4 to Ubuntu Linux, selecting the tar.gz version and unpacked it. That produces 867 objects in one directory. Most files are .html with some images; there do not appear to be any source code files. I am not a Linux command line guru; I prefer the graphical interface. I do not know how to follow the instructions about needing to specify that some standard libraries be installed.

I tried the self-extracting Windows version. It seems to install properly and produced lots of subdirectories. I try to follow the LISSOM Orientation Map tutorial. I get a Topographica GUI similar to that shown in the tutorial. When I try to do Simulation | Load Snapshot, there initially appear to be no files available. The default is to just show files of type *.typ and there are none. When I ask for all files, there are several of type .ty, but there is no file that looks like "lissom_oo_or_10000.". The closest match seems to be "lissom_oo_or_noshrinking.ty". When I try to load that I get an error window from Tk that complains about an UnpicklingError Exception. When I try to load "lissom_oo_or_homeomaxent.ty" I get the same thing. Same for "lissom_oo_or.ty".

Next I try to follow the SOM Retininotopic Map tutorial. From the Windows GUI, I try the Simulation | Model Editor. It brings up a blank window and it is not clear how to get the graphic displayed in the tutorial. I try Plots | Projection. It does nothing.

All this was based on running Topographica when it was installed. Double clicking on the Topogrphica icon on my desktop flashes something for a few ms and does nothing. Launching topographica.bat gives some messages, then the prompt "Topographica_t0>>>".

Converted from SourceForge issue 1372767, submitted by ceball
Submit Date: 2005-12-04 10:46 GMT

- Entering an expression like:
1,
into one of the boxes produces an annoying error that
won't go away.

- Strings get evaluated to numbers. So if someone
specifies an orientation of, say,
hello
orientation gets set to zero.

Really the GUI should alert the user to bad input
before passing it to the simulator. Or, if it is to
pass on bad input to the simulator, it mustn't corrupt
it (e.g. string to zero). Then the Parameter itself can
detect the problem and raise an error as it would do
for commandline use.

Converted from SourceForge issue 1860837, submitted by ceball
Submit Date: 2007-12-30 10:06 GMT

Working through the LISSOM tutorial, I see that if I open a projection window and ask to see the LateralInhibitory connections, I get some strange coloring.

Haven't investigated, but I guess this is related to changes around the global plotgroups dictionary. Probably only happens after loading a snapshot?

Converted from SourceForge issue 1822514, submitted by ceball
Submit Date: 2007-10-30 04:55 GMT

Also, probably related to failing test in testtkparameterizedobject (only fails on Windows).

Converted from SourceForge issue 1815911, submitted by jbednar
Submit Date: 2007-10-18 16:05 GMT

The pattern_present command used by the various map measurement commands in topo.commands.* saves the state of all of the Sheets, but it does not appear to save the state of the projection activity. (E.g. try measuring an orientation map, then looking at the projection activity.) With the provided examples this is not very important, because maps are measured only in between interesting activity, but in general it would be best if the state of the projections was saved and restored by pattern_present. This will require Sheet.state_push() to ask its projections to save their state, and will require projections to add code to do so.

Converted from SourceForge issue 1827304, submitted by ceball
Submit Date: 2007-11-07 03:48 GMT

The command used to start topographica, as stored in the stdout file, is missing quotes.

It's tricky because one can't just add quotes around things that might have quotes themselves. We'd have to process each string somehow, e.g. to expand existing quotes using ". I wonder if there is already something somewhere in Python that would do that...

Converted from SourceForge issue 1451399, submitted by ceball
Submit Date: 2006-03-16 16:02 GMT

Occasionally when I type Ctrl-z at the command prompt
to quit Topographica, the program exits but I get this
note:

Fatal Python error: PyEval_RestoreThread: NULL tstate

This does not appear to happen when I choose 'Quit'
from the menu, but I haven't tested this extensively.

Converted from SourceForge issue 1760226, submitted by ceball
Submit Date: 2007-07-25 11:26 GMT

Panels such as the activity window and test pattern window are slow to open. This is very noticeable with lissom_oo_dr, for instance.

Converted from SourceForge issue 1869796, submitted by nobody
Submit Date: 2008-01-12 02:50 GMT

The unit tests and the "make compare_oo_or" test are fine, but I
can't run the other, smaller tests of all the networks because the
checked in _DATA files don't seem to work on Windows. I get
"ImportError: no module named fixedpoint" during unpickling. Importing
fixedpoint works in Windows, and I can see it there in
site-packages. So there's some confusion somewhere, and it could be
difficult to solve. I have no idea whatever about this problem.

Converted from SourceForge issue 1351532, submitted by ceball
Submit Date: 2005-11-08 17:30 GMT

Some window managers place borders around e.g. the
weights in the projection window; there is a
one-pixel-wide gap between the edge of each plot and
its border.