mverzilli/crystalla

Crystal library for Numerical Methods. It binds to LAPACK and is unashamedly inspired by Numpy.

crystalla's Issues

Test, ignore this

Add rows to matrix / Add element to Ndarray: shouldn't it add it to the existing object instead of returning a new one ?

I'm experimenting with a new Ndarray class and while creating a class method add_element, I took a look at how the add_rows method was coded for the Matrix class, and I noticed that it wasn't modifying the object, but was returning a new one.

Instinctively, I would expect that I could do:

matrix = Crystalla::Matrix.new([1.0, 2.0, 2.0, 3.0], 2, 2)
p matrix
>> Matrix[[ 1.0, 2.0 ],
       [ 2.0, 3.0 ]]
matrix.add_rows(1, Crystalla::Matrix.new([4.5, 3.0], 1,2))
p matrix 
>> Matrix[[ 1.0, 2.0 ],
       [ 4.5, 3.0 ],
       [ 2.0, 3.0 ]]

Instead, I need to assign it to a new instance to get the modified matrix:
matrix2 = matrix.add_rows(1, Crystalla::Matrix.new([4.5, 3.0], 1,2))

Is there a particular reason why you made this choice ?
For Ndarrays, it would seem more natural to do:

array = Crystalla::Ndarray.new([1.0, 2.0, 2.0, 3.0])
array.add_element(6.0, 1)
p array 
>> Ndarray [1.0, 6.0, 2.0, 2.0, 3.0]

Let me know what you think !

Vector support

Hi, I have started to play with Crystal. For now, I'm trying to migrate some scripts I had in ruby with some algorithms implementations (about all ML algorithms). I could migrate some matrices using this library, but I couldn't find support for Vector class in Crystal and neither crystalla.

http://ruby-doc.org/stdlib-2.2.1/libdoc/matrix/rdoc/Vector.html

Can you help me? I am willing to contribute with these kind of libraries when I have enough knowledge and experience with Crystal , I think it has an enormous potential for Data Science, and I eventually want to use it in some real data oriented projects 😄

Load/Save to .npy files

The .npy file format has become standard to serialize arrays, so I think crystalla could really use it. It will open up possibility of interoperability with other systems such as tensorflow/python. WDYT?

Installation issues

Hi !

I'm trying to require crystalla in a small script but it crashes at the require.

Error in eig.cr:1: while requiring "crystalla"

require "crystalla"
^

in lib/crystalla/src/crystalla.cr:1: while requiring "./crystalla/*"

require "./crystalla/*"
^

Syntax error in lib/crystalla/src/crystalla/lib_blas.cr:4: expecting identifier 'end', not 'else'

  else
  ^

Since there isn't a lot of users yet, it's hard to find answers elsewhere. Is this a common issue ?

My shard.yml:

dependencies:
  crystalla:
    github: mverzilli/crystalla

Any idea ?

Thanks a lot