pyclips
pyclips copied to clipboard
Error when setting a clips class instance multislot to [ ]
import clips
C = clips.BuildClass("C", "(is-a USER)(multislot s)")
i = clips.BuildInstance("test1", C)
#i.Slots['s'] = clips.Multifield([]) # error
i.Slots['s'] = [] # error
#i.Slots['s'] = clips.Multifield(["1", "4"]) # no error
#i.Slots['s'] = ["1", "4"] # no error
print i.Slots["s"]
I am getting this error:
_clips.ClipsError: C09: unable to understand argument
Stacktrace:
File "\\VBOXSVR\vagrant\return_empty_list_bug.py", line 6, in <module>
i.Slots['s'] = [] # error
File "c:\Python27\Lib\site-packages\clips\_clips_wrap.py", line 175, in _WRAPPER
return f(self, *args)
File "c:\Python27\Lib\site-packages\clips\_clips_wrap.py", line 269, in _WRAPPER
return f(self, *newargs)
File "c:\Python27\Lib\site-packages\clips\_clips_wrap.py", line 2506, in __setitem__
_c.directPutSlot(self.__instance, name, _py2cl(v))
_clips.ClipsError: C09: unable to understand argument
The problem seems to be in file clipsmodule.c
line 1788:
case MULTIFIELD:
if(!PyList_Check(value))
goto fail;
n = PyList_Size(value);
if(n == 0) goto fail; // <---- This seems to cause the error.
if(!(do_value = EnvCreateMultifield(env, n)))
goto fail;
for(i2 = 0; i2 < n; i2++) {
item = PyList_GetItem(value, i2);
if(!item)
goto fail;
if(!i_py2do_mfhelp_e(env, item, do_value, (int)(i2 + 1)))
goto fail;
}
break; /* n is still good for below */
Why did you add this check? What will happen if I uncomment it?
In fact I don't really remember why at the moment: the APG states that multifields can be of zero length - in which case the end index is lower than the begin index! Probably some specific code should be written to handle this case in the clips_to_python and python_to_clips multifield conversion helpers, indeed because of the peculiarity of this indexing method.
Which helper functions do you mean? _py2cl()
and _cl2py()
in _clips_wrap.py
?
No, i mean the lower-level i_do2py...()
and i_py2do...()
function groups in _clipsmodule.c starting at line 1476.
Any hint what this "specific code" to handle this case should do? It seems to work if I simply comment out the check.
The 0
value was used in that code as an initializer for Py_ssize_t
(line 1756), so I used that value to determine that something went wrong. Probably I shouldn't have used 0
and maybe Py_ssize_t
in the first place, and initialize the two sizes in some different way - for instance -1
. Maybe it would be safe to switch back to plain int
s, use -1
for an initializer, check whether n < 0
and in such case goto fail
.