Re: [Numpy-discussion] how to create an array of objects that are sequences?

[Robert Kern]

Christopher Barker wrote:
> Robert Kern wrote:
>> Chris Barker wrote:
>>> What if your "objects" were nested sequences, and you wanted to partly 
>>> flatten them -- which would you flatten?
>> I'm pretty sure that that is exactly the ambiguity that the depth option
>> resolves. Can you give me an example where it's still ambiguous with the 
>> depth
>> information provided?
> 
> I was imagining:
> 
> Say you have a bunch of nested sequences that, fully expanded, would 
> yield a (i,j,k) rank-3 array. If you wanted a rank-2 array, you could 
> have either:
> a (i,j) array with each element being a length-k sequence
> or:
> a (i,k) array, with each element being a length-j sequence.
> 
> This is quite trivial with numpy n-d arrays.
> 
> However, while you could build either of those from nested sequences, 
> there isn't, in fact, and object in there that is that length-j sequence.

And I think that's the key difference. I don't think array() should be 
responsible for *arbitrarily* flattening nested sequences in order to 
reinterpret structure. Instead, I think we really only need array() to be able 
to answer the question, "is this object an atom or a sequence I need to descend 
into?" Essentially, for all valid index-tuples (say, [i,j,k]) in the array:

   arr[i,j,k] == original[i,j,k]

> I guess this really points to the power of n-d arrays!

Yes. Since one can get the (i,k) array after getting the (i,j,k) array and 
reshaping, I don't think we need to support it in array().

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless enigma
  that is made terrible by our own mad attempt to interpret it as though it had
  an underlying truth."
   -- Umberto Eco
___
Numpy-discussion mailing list
Numpy-discussion@scipy.org
http://projects.scipy.org/mailman/listinfo/numpy-discussion

Re: [Numpy-discussion] how to create an array of objects that are sequences?

[Christopher Barker]

Robert Kern wrote:
> Chris Barker wrote:
>> What if your "objects" were nested sequences, and you wanted to partly 
>> flatten them -- which would you flatten?
> 
> I'm pretty sure that that is exactly the ambiguity that the depth option
> resolves. Can you give me an example where it's still ambiguous with the depth
> information provided?

I was imagining:

Say you have a bunch of nested sequences that, fully expanded, would 
yield a (i,j,k) rank-3 array. If you wanted a rank-2 array, you could 
have either:
a (i,j) array with each element being a length-k sequence
or:
a (i,k) array, with each element being a length-j sequence.

This is quite trivial with numpy n-d arrays.

However, while you could build either of those from nested sequences, 
there isn't, in fact, and object in there that is that length-j sequence.

I guess this really points to the power of n-d arrays!

Here's a simple example:

t = (((1,2),(3,4)),((5,6),(7,8)))

depth 1 -- (4,):
array([(1, 2), (3, 4), (5, 6), (7, 8)], dtype=object)
 >>> a[0]
(1, 2)

depth 1 -- (2,)
 >>> a
array([(1, 2, 3, 4), (5, 6, 7, 8)], dtype=object)
 >>> a[0]
(1, 2, 3, 4)

But since there was no object in the original sequence that actually had 
a 4-tuple, maybe no one would ever be looking for that.

-Chris







-- 
Christopher Barker, Ph.D.
Oceanographer

Emergency Response Division
NOAA/NOS/OR&R(206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115   (206) 526-6317   main reception

[EMAIL PROTECTED]
___
Numpy-discussion mailing list
Numpy-discussion@scipy.org
http://projects.scipy.org/mailman/listinfo/numpy-discussion

Re: [Numpy-discussion] how to create an array of objects that are sequences?

[Robert Kern]

Chris Barker wrote:
> Travis E. Oliphant wrote:
> 
>> I like the depth option, and that could be used with the current 
>> interface without too much difficulty, I think.
>>
>> This question does come up quite often and should probably be addressed.
> 
> Yes, it has, but the depth option was discussed too, and while it might 
> solve some problems, it would still be ambiguous:
> 
> What if your "objects" were nested sequences, and you wanted to partly 
> flatten them -- which would you flatten?

I'm pretty sure that that is exactly the ambiguity that the depth option
resolves. Can you give me an example where it's still ambiguous with the depth
information provided?

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless enigma
 that is made terrible by our own mad attempt to interpret it as though it had
 an underlying truth."
  -- Umberto Eco
___
Numpy-discussion mailing list
Numpy-discussion@scipy.org
http://projects.scipy.org/mailman/listinfo/numpy-discussion

Re: [Numpy-discussion] how to create an array of objects that are sequences?

[Chris Barker]

Travis E. Oliphant wrote:

> I like the depth option, and that could be used with the current 
> interface without too much difficulty, I think.
> 
> This question does come up quite often and should probably be addressed.

Yes, it has, but the depth option was discussed too, and while it might 
solve some problems, it would still be ambiguous:

What if your "objects" were nested sequences, and you wanted to partly 
flatten them -- which would you flatten? Maybe that's a weird enough 
case that the depth option would work most of the time, but I'm not 
sure. I think a "shape" option might cover this better.

It seems the case at hand is how to make a custom object that can tell 
numpy how it should be put into an object array -- that may be an even 
rarer use case, but it seems like a good idea to support it.

-Chris


-- 
Christopher Barker, Ph.D.
Oceanographer

Emergency Response Division
NOAA/NOS/OR&R(206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115   (206) 526-6317   main reception

[EMAIL PROTECTED]
___
Numpy-discussion mailing list
Numpy-discussion@scipy.org
http://projects.scipy.org/mailman/listinfo/numpy-discussion

Re: [Numpy-discussion] how to create an array of objects that are sequences?

[Robert Kern]

Travis E. Oliphant wrote:

> I like the depth option, and that could be used with the current 
> interface without too much difficulty, I think.
> 
> This question does come up quite often and should probably be addressed.

Alternatively, a shape= argument could be used. You can use -1 entries if you
want array() to just figure out the exact dimensions like reshape() does. The
shape= argument lets you give array() a bit more information, which might be 
useful.

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless enigma
 that is made terrible by our own mad attempt to interpret it as though it had
 an underlying truth."
  -- Umberto Eco
___
Numpy-discussion mailing list
Numpy-discussion@scipy.org
http://projects.scipy.org/mailman/listinfo/numpy-discussion

Re: [Numpy-discussion] how to create an array of objects that are sequences?

[Pearu Peterson]

Just ignore this solution. It was not quite working
and I was able to get a segfault from it.

Pearu

On Fri, January 4, 2008 8:58 pm, Pearu Peterson wrote:
> On Fri, January 4, 2008 8:00 pm, Pearu Peterson wrote:
>> On Fri, January 4, 2008 7:33 pm, Travis E. Oliphant wrote:
>>> Pearu Peterson wrote:
 Hi,

 Say, one defines

 class A(tuple):
   def __repr__(self):
 return 'A(%s)' % (tuple.__repr__(self))

 and I'd like to create an array of A instances.
>
>>> So, create an empty object array and insert the entries the way you
>>> want
>>> them:
>>>
>>> a = np.empty(1,dtype=object)
>>> a[0] = A((1,2))
>>
>> Meantime I was reading arrayobject.c and it seems that
>> before objects are checked for being sequences, their
>> __array_interface__ is accessed (eg in Array_FromSequence,
>> discover_depth).
>>
>> Would this provide a solution if the class A would define
>> a property __array_interface__? I just don't know what
>> the data field should be for an object.
>
> Ok, I found a partial solution:
>
>
> class A(tuple):
>def __repr__(self):
>  return 'A(%s)' % (tuple.__repr__(self))
> @property
> def __array_interface__(self):
> import numpy
> obj = numpy.empty(1,dtype=object)
> obj[0] = self
> return obj.__array_interface__.copy()
>
> ___
> Numpy-discussion mailing list
> Numpy-discussion@scipy.org
> http://projects.scipy.org/mailman/listinfo/numpy-discussion
>


___
Numpy-discussion mailing list
Numpy-discussion@scipy.org
http://projects.scipy.org/mailman/listinfo/numpy-discussion

Re: [Numpy-discussion] how to create an array of objects that are sequences?

[Travis E. Oliphant]

Charles R Harris wrote:
>
>
> On Jan 4, 2008 11:13 AM, Pearu Peterson <[EMAIL PROTECTED] 
> > wrote:
>
> On Fri, January 4, 2008 7:33 pm, Travis E. Oliphant wrote:
>
> > So, create an empty object array and insert the entries the way
> you want
> > them:
> >
> > a = np.empty(1,dtype=object)
> > a[0] = A((1,2))
>
> Actually this is still an option if to put the above
> to a A.as_array method something like
>
> class A(tuple):
>def as_array(self):
>import numpy
>obj = numpy.empty(1,dtype=object)
>obj[0] = self
>return obj
>
> but it would be nice if the __array__ method also worked.
> Could this be something that should be put to issues or
> would it just get a WontFix flag? 
>
>  
> IIRC the previous discussions, there simply isn't enough information 
> in the array interface to decide what needs to be done with regards to 
> object arrays. I think a general solution would require a specialized 
> creation function with more arguments, say, a depth argument which 
> would create objects at a certain depth of nesting. In your case that 
> would be at level 0(1?). I think that would solve most problems, and I 
> suppose it could be added to the current array interface with a 
> default value implying current behavior, i.e., as deep as possible.
>
I like the depth option, and that could be used with the current 
interface without too much difficulty, I think.

This question does come up quite often and should probably be addressed.

-Travis


___
Numpy-discussion mailing list
Numpy-discussion@scipy.org
http://projects.scipy.org/mailman/listinfo/numpy-discussion

Re: [Numpy-discussion] how to create an array of objects that are sequences?

[Pearu Peterson]

On Fri, January 4, 2008 8:00 pm, Pearu Peterson wrote:
> On Fri, January 4, 2008 7:33 pm, Travis E. Oliphant wrote:
>> Pearu Peterson wrote:
>>> Hi,
>>>
>>> Say, one defines
>>>
>>> class A(tuple):
>>>   def __repr__(self):
>>> return 'A(%s)' % (tuple.__repr__(self))
>>>
>>> and I'd like to create an array of A instances.
>>
>> The array function was designed a long time ago to inspect sequences and
>> flatten them.
>>
>> Arguably, there should be more intelligence when an "object" array is
>> requested, but there is ambiguity about what the "right" thing to do is.
>>
>> Thus, the current situation is that if you are creating object arrays,
>> the advice is to populate it after the fact.
>>
>> So, create an empty object array and insert the entries the way you want
>> them:
>>
>> a = np.empty(1,dtype=object)
>> a[0] = A((1,2))

> Meantime I was reading arrayobject.c and it seems that
> before objects are checked for being sequences, their
> __array_interface__ is accessed (eg in Array_FromSequence,
> discover_depth).
>
> Would this provide a solution if the class A would define
> a property __array_interface__? I just don't know what
> the data field should be for an object.

Ok, I found a partial solution:

class A(tuple):
def __repr__(self):
return 'A('+tuple.__repr__(self)+')'
@property
def __array_interface__(self):
import numpy
obj = numpy.empty(1,dtype=object)
obj[0] = self
return obj.__array_interface__.copy()

>>> from numpy import *
>>> array([A((1,2))])
array([[1, 2]], dtype=object)

but

>>> array(A((1,2)))
array([None], dtype=object)

Pearu
PS: sorry about previous mail, Send was pressed accidentaly.


___
Numpy-discussion mailing list
Numpy-discussion@scipy.org
http://projects.scipy.org/mailman/listinfo/numpy-discussion

Re: [Numpy-discussion] how to create an array of objects that are sequences?

[Pearu Peterson]

On Fri, January 4, 2008 8:00 pm, Pearu Peterson wrote:
> On Fri, January 4, 2008 7:33 pm, Travis E. Oliphant wrote:
>> Pearu Peterson wrote:
>>> Hi,
>>>
>>> Say, one defines
>>>
>>> class A(tuple):
>>>   def __repr__(self):
>>> return 'A(%s)' % (tuple.__repr__(self))
>>>
>>> and I'd like to create an array of A instances.

>> So, create an empty object array and insert the entries the way you want
>> them:
>>
>> a = np.empty(1,dtype=object)
>> a[0] = A((1,2))
>
> Meantime I was reading arrayobject.c and it seems that
> before objects are checked for being sequences, their
> __array_interface__ is accessed (eg in Array_FromSequence,
> discover_depth).
>
> Would this provide a solution if the class A would define
> a property __array_interface__? I just don't know what
> the data field should be for an object.

Ok, I found a partial solution:


class A(tuple):
   def __repr__(self):
 return 'A(%s)' % (tuple.__repr__(self))
@property
def __array_interface__(self):
import numpy
obj = numpy.empty(1,dtype=object)
obj[0] = self
return obj.__array_interface__.copy()

___
Numpy-discussion mailing list
Numpy-discussion@scipy.org
http://projects.scipy.org/mailman/listinfo/numpy-discussion

Re: [Numpy-discussion] how to create an array of objects that are sequences?

[Charles R Harris]

On Jan 4, 2008 11:13 AM, Pearu Peterson <[EMAIL PROTECTED]> wrote:

> On Fri, January 4, 2008 7:33 pm, Travis E. Oliphant wrote:
>
> > So, create an empty object array and insert the entries the way you want
> > them:
> >
> > a = np.empty(1,dtype=object)
> > a[0] = A((1,2))
>
> Actually this is still an option if to put the above
> to a A.as_array method something like
>
> class A(tuple):
>def as_array(self):
>import numpy
>obj = numpy.empty(1,dtype=object)
>obj[0] = self
>return obj
>
> but it would be nice if the __array__ method also worked.
> Could this be something that should be put to issues or
> would it just get a WontFix flag?


IIRC the previous discussions, there simply isn't enough information in the
array interface to decide what needs to be done with regards to object
arrays. I think a general solution would require a specialized creation
function with more arguments, say, a depth argument which would create
objects at a certain depth of nesting. In your case that would be at level
0(1?). I think that would solve most problems, and I suppose it could be
added to the current array interface with a default value implying current
behavior, i.e., as deep as possible.

Chuck
___
Numpy-discussion mailing list
Numpy-discussion@scipy.org
http://projects.scipy.org/mailman/listinfo/numpy-discussion

Re: [Numpy-discussion] how to create an array of objects that are sequences?

[Pearu Peterson]

On Fri, January 4, 2008 7:33 pm, Travis E. Oliphant wrote:

> So, create an empty object array and insert the entries the way you want
> them:
>
> a = np.empty(1,dtype=object)
> a[0] = A((1,2))

Actually this is still an option if to put the above
to a A.as_array method something like

class A(tuple):
def as_array(self):
import numpy
obj = numpy.empty(1,dtype=object)
obj[0] = self
return obj

but it would be nice if the __array__ method also worked.
Could this be something that should be put to issues or
would it just get a WontFix flag?

Thanks again,
Pearu

___
Numpy-discussion mailing list
Numpy-discussion@scipy.org
http://projects.scipy.org/mailman/listinfo/numpy-discussion

Re: [Numpy-discussion] how to create an array of objects that are sequences?

[Pearu Peterson]

On Fri, January 4, 2008 7:33 pm, Travis E. Oliphant wrote:
> Pearu Peterson wrote:
>> Hi,
>>
>> Say, one defines
>>
>> class A(tuple):
>>   def __repr__(self):
>> return 'A(%s)' % (tuple.__repr__(self))
>>
>> and I'd like to create an array of A instances.
>
> The array function was designed a long time ago to inspect sequences and
> flatten them.
>
> Arguably, there should be more intelligence when an "object" array is
> requested, but there is ambiguity about what the "right" thing to do is.
>
> Thus, the current situation is that if you are creating object arrays,
> the advice is to populate it after the fact.
>
> So, create an empty object array and insert the entries the way you want
> them:
>
> a = np.empty(1,dtype=object)
> a[0] = A((1,2))

Thanks Travis for the hint!
The solution is too verbose for an end-user, though.

Meantime I was reading arrayobject.c and it seems that
before objects are checked for being sequences, their
__array_interface__ is accessed (eg in Array_FromSequence,
discover_depth).

Would this provide a solution if the class A would define
a property __array_interface__? I just don't know what
the data field should be for an object.

Thanks,
Pearu

___
Numpy-discussion mailing list
Numpy-discussion@scipy.org
http://projects.scipy.org/mailman/listinfo/numpy-discussion

Re: [Numpy-discussion] how to create an array of objects that are sequences?

[Travis E. Oliphant]

Pearu Peterson wrote:
> Hi,
>
> Say, one defines
>
> class A(tuple):
>   def __repr__(self):
> return 'A(%s)' % (tuple.__repr__(self))
>
> and I'd like to create an array of A instances.
>
> Currently I get
>
>   

The array function was designed a long time ago to inspect sequences and 
flatten them.   

Arguably, there should be more intelligence when an "object" array is 
requested, but there is ambiguity about what the "right" thing to do is. 

Thus, the current situation is that if you are creating object arrays, 
the advice is to populate it after the fact. 

So, create an empty object array and insert the entries the way you want 
them:

a = np.empty(1,dtype=object)
a[0] = A((1,2))


-Travis

___
Numpy-discussion mailing list
Numpy-discussion@scipy.org
http://projects.scipy.org/mailman/listinfo/numpy-discussion

[Numpy-discussion] how to create an array of objects that are sequences?

[Pearu Peterson]

Hi,

Say, one defines

class A(tuple):
  def __repr__(self):
return 'A(%s)' % (tuple.__repr__(self))

and I'd like to create an array of A instances.

Currently I get

>>> array([A((1,2))], dtype=object)
array([[1, 2]], dtype=object)

but what I'd like to get is

array([A((1, 2))], dtype=object)

I tried defining A.__array__ method but that does not have
effect when A instance is inside a sequence to array call.

Thanks for any information for how to resolve this problem,
PEaru

___
Numpy-discussion mailing list
Numpy-discussion@scipy.org
http://projects.scipy.org/mailman/listinfo/numpy-discussion