<br><br><div class="gmail_quote">On Mon, Apr 21, 2008 at 9:09 PM, Travis E. Oliphant <<a href="mailto:oliphant@enthought.com">oliphant@enthought.com</a>> wrote:<br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<div class="Ih2E3d">Charles R Harris wrote:<br>
><br>
><br>
> On Mon, Apr 21, 2008 at 8:53 PM, Charles R Harris<br>
</div><div class="Ih2E3d">> <<a href="mailto:charlesr.harris@gmail.com">charlesr.harris@gmail.com</a> <mailto:<a href="mailto:charlesr.harris@gmail.com">charlesr.harris@gmail.com</a>>> wrote:<br>
><br>
><br>
><br>
> On Mon, Apr 21, 2008 at 8:28 PM, Travis E. Oliphant<br>
</div><div><div></div><div class="Wj3C7c">> <<a href="mailto:oliphant@enthought.com">oliphant@enthought.com</a> <mailto:<a href="mailto:oliphant@enthought.com">oliphant@enthought.com</a>>> wrote:<br>
><br>
> Charles R Harris wrote:<br>
> > I've gotten my own python class with a logical_not method to<br>
> work<br>
> > correctly if I goto<br>
> numpy/core/code_generators/generate_umath.py and<br>
> > change<br>
> I need more context for this. Why does the umath generator<br>
> matter for<br>
> your python class?<br>
> ><br>
> > 'logical_not' :<br>
> > Ufunc(1, 1, None,<br>
> > 'returns not x elementwise.',<br>
> > TD(noobj, out='?'),<br>
> > TD(M, f='logical_not', out='?'),<br>
> > ),<br>
> ><br>
> > to<br>
> ><br>
> > 'logical_not' :<br>
> > Ufunc(1, 1, None,<br>
> > 'returns not x elementwise.',<br>
> > TD(noobj, out='?'),<br>
> > TD(M, f='logical_not'),<br>
> > ),<br>
> ><br>
> Why is this an error? Is the difference only removing the out<br>
> variable? It's been a while since I reviewed this code, so<br>
> what does<br>
> removing the out variable do functionally (What is the<br>
> difference in the<br>
> ufunc that is generated)?<br>
><br>
><br>
> The way it is, it passes a boolean array to the PyUFunc_O_O_method<br>
> loop where the loop is expecting an object array. I checked the<br>
> step size to see this, and that's also how the generated signature<br>
> reads. Consequently, when the reference count is decremented bad<br>
> things happen. I suspect this hasn't been seen before because it<br>
> hadn't been tried. I wrote loop tests before cleaning up the loop<br>
> code and the bug turned up then.<br>
><br>
> My guess is that here M means object called through non-Python<br>
> method (logical_not), and omitting out means the output type is<br>
> the same as the input. I suspect that '?' should do the same thing<br>
> and that there might be a bug in the function dispatcher or the<br>
> signature generator, but I'm not clear on that yet.<br>
><br>
><br>
> If out is omitted, the output types matches the input types. Here's<br>
> where the output types are generated:<br>
><br>
> def finish_signature(self, nin, nout):<br>
> if self.in_ is None:<br>
> self.in_ = self.type * nin<br>
> assert len(self.in_) == nin<br>
> if self.out is None:<br>
> self.out = self.type * nout<br>
> assert len(self.out) == nout<br>
><br>
> So that seems like the right thing to do. I still don't know what '?'<br>
> means, though.<br>
<br>
</div></div>Thanks Chuck,<br>
<br>
I get it now (I should have spent a minute looking at the code). The<br>
'?' is the character code for "boolean"<br>
<br>
So, I think you are right about the bug.<br>
<div><div></div><div class="Wj3C7c"></div></div></blockquote><div><br>OK,<br><br>So I quess that everywhere I see TD(M... I should remove the "out='?' ". Will do.<br><br>Chuck<br></div><br></div><br>