Re: [ojAlgo-user] Interface analogous to Numpy?

[Bruno D. <ab...@ma...>]

Dear Anders Peterson,

> I looked at the ndarray "interface". Some things look a bit scary... Are all   
> the attributes and methods mentioned on that page public and non-optional?     

Regarding the methods, many of them can be implemented as stride-and-offset
shuffling and calculation, for example
 * ndarray.repeat
 * ndarray.diagonal
 * ndarray.reshape
 * ndarray.flatten
 * ndarray.swapaxes
 * ndarray.ravel
 * ndarray.view

Many others (min, argmin, round, sum, mean, var, prod, all, any ...) are easy
to add. I hope they cover most of the applications. We'll see.
 
Best regards
  Bruno Daniel

Bruno Daniel wrote:
> Dear Anders Peterson,
> 
>> I looked at the ndarray "interface". Some things look a bit scary... Are all
>> the attributes and methods mentioned on that page public and non-optional?
> 
> Ok, let's see:
> 
> http://docs.scipy.org/doc/numpy/reference/arrays.ndarray.html
> There are many fields which are all public. We could replace them by set- and
> get methods in the adapter. Most of them are used often. In consider only
> real, imag, __array_priority__ as optional:
> 
> "ndarray.flags 	Information about the memory layout of the array.
> ndarray.shape 	Tuple of array dimensions.
> ndarray.strides 	Tuple of bytes to step in each dimension when traversing an array.
> ndarray.ndim 	Number of array dimensions.
> ndarray.data 	Python buffer object pointing to the start of the array's data.
> ndarray.size 	Number of elements in the array.
> ndarray.itemsize 	Length of one array element in bytes.
> ndarray.nbytes 	Total bytes consumed by the elements of the array.
> ndarray.base 	Base object if memory is from some other object.
> ndarray.dtype 	Data-type of the array's elements.
> 
> ndarray.T 	Same as self.transpose(), except that self is returned if self.ndim < 2.
> ndarray.real 	The real part of the array.
> ndarray.imag 	The imaginary part of the array.
> ndarray.flat 	A 1-D iterator over the array.
> __array_priority__" 	
> 
> ndarray.strides and ndarray.shape are the most important for creating
> views. We would need an additional offset, because there are no pointers to
> array parts in Java.
> 
> For instance a vector-view of the third column of an m x n non-view,
> row-order matrix A would roughly look like this:
> 
> shape = tuple(m, 1)
> offset = 2 * m
> base = the same array as in A, no copy
> strides = tuple(n)
> size = m
> ndim = 1
> 
> The transpose of the same matrix would roughly look like this:
> 
> shape = tuple(m, n)
> offset = 0
> base = the same array as in A, no copy
> strides = the reverse of A.strides = tuple(1, n)
> size = the same as in A
> ndim = the same as in A = 2
> 
> If A is already a view there will be other values in offset and strides.
> 
> These are for interfacing C programs. I hope they are not used too often in
> common Numpy programs. Stefan Richthofer, what do you think?
> 
> "The Array Interface.
> __array_interface__ 	Python-side of the array interface
> __array_struct__ 	C-side of the array interface"
> ndarray.ctypes 	An object to simplify the interaction of the array
> with the ctypes module."
> 
>> Currently ojAlgo supports double, BigDecimal, ComplexNumber and
>> RationalNumber elements with the ArrayAnyD class. Are other types required?
> 
> Integers are quite common. The other types (there are many of them) could be
> added later:
> http://docs.scipy.org/doc/numpy/reference/arrays.dtypes.html#arrays-dtypes
> 
>> How about I create a sandbox version control system (for experimental code) over at http://code.google.com/p/ojalgo/
>> 
>> What do you prefer: Subversion, Git or Mercurial?
> Thanks, Git would be nice, Subversion is also ok.
> 
> Best regards
>   Bruno Daniel
> 
> Anders Peterson wrote:
>> Getting some small application that (only) depends on NumPy to run on Jython would be a proof of concept.
>> 
>> I looked at the ndarray "interface". Some things look a bit scary... Are all the attributes and methods mentioned on that page public and non-optional?
>> 
>> Currently ojAlgo supports double, BigDecimal, ComplexNumber and RationalNumber elements with the ArrayAnyD class. Are other types required?
>> 
>> How about I create a sandbox version control system (for experimental code) over at http://code.google.com/p/ojalgo/
>> 
>> What do you prefer: Subversion, Git or Mercurial?
>> 
>> /Anders
>> 
>> 
>> On 19 apr 2012, at 22:21, B.D. wrote:
>> 
>>> Dear Mr. Peterson,
>>> 
>>>> Don't talk too much - just do it. Start with the most simple case.
>>> 
>>> yes, you're right. I think this is a nice collection of simple unit tests for
>>> me: http://www.scipy.org/Numpy_Example_List
>>> 
>>> See you later,
>>>  B.D.
>>> 
>>> 
>>> Anders Peterson wrote:
>>>> Let's call it Jumpy...
>>>> 
>>>> Since I'm not a python/jython/numpy/scipy user I think someone else should take the lead here.
>>>> 
>>>> Don't talk too much - just do it. Start with the most simple case.
>>>> 
>>>> I'm definitely interested in ways to make ojAlgo more useful.
>>>> 
>>>> /Anders
>>>> 
>>>> 
>>>> On 19 apr 2012, at 15:39, Stefan Richthofer wrote:
>>>> 
>>>>> Hello together,
>>>>> I also often thought about the idea to have a numpy-substitute for java, but always feared the enormous effort related to it.
>>>>> 
>>>>> Be aware of the high Potential of the idea. 
>>>>> 
>>>>> Having a numpy substitute in java would enable a huge amount of scientific python-code to run directly in java via jython (maybe with a slight modification of the imports).
>>>>> 
>>>>> Jython's main issue keeping it from being very useful is, that it does not support external (native) libraries written for Python.
>>>>> 
>>>>> I regard numpy as one of the most important of these and I believe, that a significant amount of scientific python-code has numpy as the only non-python-core-dependency and thus could run via jython, if there was a numpy for java.
>>>>> 
>>>>> Maybe you could consider a cooperation with the jython-developers at http://www.jython.org ...?
>>>>> It should be highly in their interest to have a numpy implementation for java.
>>>>> 
>>>>> 
>>>>> Greetings
>>>>> 
>>>>> Stefan
>>>>> 
>>>>> 
>>>>> 
>>>>> -------- Original-Nachricht --------
>>>>>> Datum: Thu, 19 Apr 2012 13:56:49 +0200
>>>>>> Von: Anders Peterson <an...@op...>
>>>>>> An: oja...@li...
>>>>>> Betreff: Re: [ojAlgo-user] Interface analogous to Numpy?
>>>>> 
>>>>>> Short anser is yes, but you have to be more specific regarding what you
>>>>>> want/need and willing to do some of the work yourself.
>>>>>> 
>>>>>> ojAlgo has multi/any-dimensional arrays:
>>>>>> 
>>>>>> http://ojalgo.org/generated/org/ojalgo/array/ArrayAnyD.html
>>>>>> 
>>>>>> Frankly I don't really know what to do with them. I tend to only use 1- or
>>>>>> 2-dimensional data structures. I created them because they were a
>>>>>> generalization of what I needed that *could* become useful.
>>>>>> 
>>>>>> I guess the various MatrixStore implementations and the
>>>>>> MatrixStore.Builder does much of what you want, but only for 2-dim arrays (and with wrong
>>>>>> syntax).
>>>>>> 
>>>>>> http://ojalgo.org/generated/org/ojalgo/matrix/store/MatrixStore.html
>>>>>> http://ojalgo.org/generated/org/ojalgo/matrix/store/MatrixStore.Builder.html
>>>>>> 
>>>>>> In ojAlgo the description of the array shape/structure is fixed as defined
>>>>>> in AccessUtils
>>>>>> 
>>>>>> http://ojalgo.org/generated/org/ojalgo/access/AccessUtils.html
>>>>>> 
>>>>>> Would that need to be changed? (I hope not.)
>>>>>> 
>>>>>> /Anders
>>>>>> 
>>>>>> 
>>>>>> On 19 apr 2012, at 11:35, B.D. wrote:
>>>>>> 
>>>>>>> Dear developers,
>>>>>>> 
>>>>>>> I'm primarily interested in Machine Learning and Data analysis. Thanks
>>>>>> for the
>>>>>>> good work, I appreciate the ojAlgo library very much! However, most new
>>>>>>> algorithms and applications in this field are still written in Python
>>>>>> for
>>>>>>> Numpy/Scipy, not Java. I just had an idea:
>>>>>>> 
>>>>>>> If one could offer classes/interfaces in pure Java with similar
>>>>>> functionality
>>>>>>> and names as in
>>>>>>> 
>>>>>>> http://docs.scipy.org/doc/numpy/reference/ ,
>>>>>>> 
>>>>>>> in particular ndarray
>>>>>>> 
>>>>>>> http://docs.scipy.org/doc/numpy/reference/arrays.ndarray.html
>>>>>>> 
>>>>>>> (e.g. strides in matrices and vectors, transposition, reshaping of
>>>>>> matrices as
>>>>>>> a view without copying, many views on the same matrix data), then one
>>>>>> could
>>>>>>> rather easily translate Python programs using Numpy to pure Java without
>>>>>>> reinventing the wheel and offer the excellent Python software for Java
>>>>>> under
>>>>>>> the same license as the original.
>>>>>>> 
>>>>>>> There are syntax limitations in Java, so
>>>>>>> 
>>>>>>> A[2:5, :]
>>>>>>> 
>>>>>>> would have to be translated to some intutive name or even a crazy
>>>>>> mnemonic like
>>>>>>> 
>>>>>>> A.viewI_c_icI(2, 5)
>>>>>>> 
>>>>>>> (I = [ or ],
>>>>>>> i = comma,
>>>>>>> c = colon,
>>>>>>> _ = slots for the arguments).
>>>>>>> 
>>>>>>> What do you think?
>>>>>>> 
>>>>>>> Best regards
>>>>>>> B.D.
>>>>>>> 
>>>>>>> 
>>>>>> ------------------------------------------------------------------------------
>>>>>>> For Developers, A Lot Can Happen In A Second.
>>>>>>> Boundary is the first to Know...and Tell You.
>>>>>>> Monitor Your Applications in Ultra-Fine Resolution. Try it FREE!
>>>>>>> http://p.sf.net/sfu/Boundary-d2dvs2
>>>>>>> _______________________________________________
>>>>>>> ojAlgo-user mailing list
>>>>>>> ojA...@li...
>>>>>>> https://lists.sourceforge.net/lists/listinfo/ojalgo-user
>>>>>> 
>>>>>> ------------------------------------------------------------------------------
>>>>>> For Developers, A Lot Can Happen In A Second.
>>>>>> Boundary is the first to Know...and Tell You.
>>>>>> Monitor Your Applications in Ultra-Fine Resolution. Try it FREE!
>>>>>> http://p.sf.net/sfu/Boundary-d2dvs2
>>>>>> _______________________________________________
>>>>>> ojAlgo-user mailing list
>>>>>> ojA...@li...
>>>>>> https://lists.sourceforge.net/lists/listinfo/ojalgo-user
>>>>> 
>>>>> -- 
>>>>> NEU: FreePhone 3-fach-Flat mit kostenlosem Smartphone!                                  
>>>>> Jetzt informieren: http://mobile.1und1.de/?ac=OM.PW.PW003K20328T7073a
>>>>> 
>>>>> ------------------------------------------------------------------------------
>>>>> For Developers, A Lot Can Happen In A Second.
>>>>> Boundary is the first to Know...and Tell You.
>>>>> Monitor Your Applications in Ultra-Fine Resolution. Try it FREE!
>>>>> http://p.sf.net/sfu/Boundary-d2dvs2
>>>>> _______________________________________________
>>>>> ojAlgo-user mailing list
>>>>> ojA...@li...
>>>>> https://lists.sourceforge.net/lists/listinfo/ojalgo-user
>>>>> 
>>>>> 
>>>> 
>>>> 
>>>> ------------------------------------------------------------------------------
>>>> For Developers, A Lot Can Happen In A Second.
>>>> Boundary is the first to Know...and Tell You.
>>>> Monitor Your Applications in Ultra-Fine Resolution. Try it FREE!
>>>> http://p.sf.net/sfu/Boundary-d2dvs2
>>>> _______________________________________________
>>>> ojAlgo-user mailing list
>>>> ojA...@li...
>>>> https://lists.sourceforge.net/lists/listinfo/ojalgo-user
>>> 
>>> ------------------------------------------------------------------------------
>>> For Developers, A Lot Can Happen In A Second.
>>> Boundary is the first to Know...and Tell You.
>>> Monitor Your Applications in Ultra-Fine Resolution. Try it FREE!
>>> http://p.sf.net/sfu/Boundary-d2dvs2
>>> _______________________________________________
>>> ojAlgo-user mailing list
>>> ojA...@li...
>>> https://lists.sourceforge.net/lists/listinfo/ojalgo-user
>>> 
>>> 
>> 
>> 
>> ------------------------------------------------------------------------------
>> For Developers, A Lot Can Happen In A Second.
>> Boundary is the first to Know...and Tell You.
>> Monitor Your Applications in Ultra-Fine Resolution. Try it FREE!
>> http://p.sf.net/sfu/Boundary-d2dvs2
>> _______________________________________________
>> ojAlgo-user mailing list
>> ojA...@li...
>> https://lists.sourceforge.net/lists/listinfo/ojalgo-user