SkillAgentSearch skills...

Matft

Numpy-like library in swift. (Multi-dimensional Array, ndarray, matrix and vector library)

Install / Use

/learn @jjjkkkjjj/Matft

README

Matft

SwiftPM compatible CocoaPods compatible Carthage compatible license

Matft is Numpy-like library in Swift. Function name and usage is similar to Numpy.

INFO: Support Complex!!

<strike> Note: You can use [Protocol version(beta version)](https://github.com/jjjkkkjjj/Matft/tree/protocol) too. </strike>

Feature & Usage

  • Many types

  • Pretty print

  • Indexing

    • Positive
    • Negative
    • Boolean
    • Fancy
    • Complex
  • Slicing

    • Start / To / By
    • New Axis
  • View

    • Assignment
  • Conversion

    • Broadcast
    • Transpose
    • Reshape
    • Astype
  • Univarsal function reduction

  • Mathematic

    • Arithmetic
    • Statistic
    • Linear Algebra
  • Complex

  • Image Conversion

...etc.

See Function List for all functions.

Declaration

MfArray

  • The MfArray such like a numpy.ndarray

    let a = MfArray([[[ -8,  -7,  -6,  -5],
                      [ -4,  -3,  -2,  -1]],
            
                     [[ 0,  1,  2,  3],
                      [ 4,  5,  6,  7]]])
    let aa = Matft.arange(start: -8, to: 8, by: 1, shape: [2,2,4])
    print(a)
    print(aa)
    /*
    mfarray = 
    [[[	-8,		-7,		-6,		-5],
    [	-4,		-3,		-2,		-1]],
    
    [[	0,		1,		2,		3],
    [	4,		5,		6,		7]]], type=Int, shape=[2, 2, 4]
    mfarray = 
    [[[	-8,		-7,		-6,		-5],
    [	-4,		-3,		-2,		-1]],
    
    [[	0,		1,		2,		3],
    [	4,		5,		6,		7]]], type=Int, shape=[2, 2, 4]
    */
    

MfType

  • You can pass MfType as MfArray's argument mftype: .Hoge . It is similar to dtype.

    ※Note that stored data type will be Float or Double only even if you set MfType.Int. So, if you input big number to MfArray, it may be cause to overflow or strange results in any calculation (+, -, *, /,... etc.). But I believe this is not problem in practical use.

  • MfType's list is below

      public enum MfType: Int{
        case None // Unsupportted
        case Bool
        case UInt8
        case UInt16
        case UInt32
        case UInt64
        case UInt
        case Int8
        case Int16
        case Int32
        case Int64
        case Int
        case Float
        case Double
        case Object // Unsupported
    }
    
  • Also, you can convert MfType easily using astype

    let a = MfArray([[[ -8,  -7,  -6,  -5],
                      [ -4,  -3,  -2,  -1]],
            
                     [[ 0,  1,  2,  3],
                      [ 4,  5,  6,  7]]])
    print(a)//See above. if mftype is not passed, MfArray infer MfType. In this example, it's MfType.Int
    
    let a = MfArray([[[ -8,  -7,  -6,  -5],
                      [ -4,  -3,  -2,  -1]],
                
                     [[ 0,  1,  2,  3],
                      [ 4,  5,  6,  7]]], mftype: .Float)
    print(a)
    /*
    mfarray = 
    [[[	-8.0,		-7.0,		-6.0,		-5.0],
    [	-4.0,		-3.0,		-2.0,		-1.0]],
    
    [[	0.0,		1.0,		2.0,		3.0],
    [	4.0,		5.0,		6.0,		7.0]]], type=Float, shape=[2, 2, 4]
    */
    let aa = MfArray([[[ -8,  -7,  -6,  -5],
                      [ -4,  -3,  -2,  -1]],
                
                     [[ 0,  1,  2,  3],
                      [ 4,  5,  6,  7]]], mftype: .UInt)
    print(aa)
    /*
    mfarray = 
    [[[	4294967288,		4294967289,		4294967290,		4294967291],
    [	4294967292,		4294967293,		4294967294,		4294967295]],
    
    [[	0,		1,		2,		3],
    [	4,		5,		6,		7]]], type=UInt, shape=[2, 2, 4]
    */
    //Above output is same as numpy!
    /*
    >>> np.arange(-8, 8, dtype=np.uint32).reshape(2,2,4)
    array([[[4294967288, 4294967289, 4294967290, 4294967291],
            [4294967292, 4294967293, 4294967294, 4294967295]],
    
           [[         0,          1,          2,          3],
            [         4,          5,          6,          7]]], dtype=uint32)
    */
    
    print(aa.astype(.Float))
    /*
    mfarray = 
    [[[	-8.0,		-7.0,		-6.0,		-5.0],
    [	-4.0,		-3.0,		-2.0,		-1.0]],
    
    [[	0.0,		1.0,		2.0,		3.0],
    [	4.0,		5.0,		6.0,		7.0]]], type=Float, shape=[2, 2, 4]
    */
    

Subscription

MfSlice

  • You can access specific data using subscript.

You can set MfSlice (see below's list) to subscript.

  • MfSlice(start: Int? = nil, to: Int? = nil, by: Int = 1)
    
  • Matft.newaxis
    
  • ~< //this is prefix, postfix and infix operator. same as python's slice, ":"
    
  • Matft.all // same as python's slice :, matft's 0~<
    
  • Matft.reverse // same as python's slice ::-1, matft's ~<<-1
    

(Positive) Indexing

  • Normal indexing

    let a = Matft.arange(start: 0, to: 27, by: 1, shape: [3,3,3])
    print(a)
    /*
    mfarray = 
    [[[	0,		1,		2],
    [	3,		4,		5],
    [	6,		7,		8]],
    
    [[	9,		10,		11],
    [	12,		13,		14],
    [	15,		16,		17]],
    
    [[	18,		19,		20],
    [	21,		22,		23],
    [	24,		25,		26]]], type=Int, shape=[3, 3, 3]
    */
    print(a[2,1,0])
    // 21
    

    Caution

    MfArray conforms to Collection protocol, so 1D MfArray returns MfArray! Use item as workaround instead of subscription such like

    let a = Matft.arange(start: 0, to: 27, by: 1, shape: [27])
    print(a[0])
    /*
    0 // a[0] is MfArray!! Nevertheless, The scalar is printed (I don't know why...)
    */
    print(a[0] + 4)
    /*
    mfarray = 
    [    4], type=Int, shape=[1]
    */
    
    // Workaround
    print(a.item(index: 0, type: Int.self))
    /*
    0
    */
    print(a.item(index: 0, type: Int.self) + 4)
    /*
    4
    */
    

    Slicing

  • If you replace : with ~<, you can get sliced mfarray. Note that use a[0~<] instead of a[:] to get all elements along axis.

    print(a[~<1])  //same as a[:1] for numpy
    /*
    mfarray = 
    [[[	9,		10,		11],
    [	12,		13,		14],
    [	15,		16,		17]]], type=Int, shape=[1, 3, 3]
    */
    print(a[1~<3]) //same as a[1:3] for numpy
    /*
    mfarray = 
    [[[	9,		10,		11],
    [	12,		13,		14],
    [	15,		16,		17]],
    
    [[	18,		19,		20],
    [	21,		22,		23],
    [	24,		25,		26]]], type=Int, shape=[2, 3, 3]
    */
    print(a[~<~<2]) //same as a[::2] for numpy
    //print(a[~<<2]) //alias
    /*
    mfarray = 
    [[[	0,		1,		2],
    [	3,		4,		5],
    [	6,		7,		8]],
    
    [[	18,		19,		20],
    [	21,		22,		23],
    [	24,		25,		26]]], type=Int, shape=[2, 3, 3]
    */
    
    print(a[Matft.all, 0]) //same as a[:, 0] for numpy
    /*
    mfarray = 
    [[    0,      1,      2],
    [ 9,      10,     11],
    [18,      19,     20]], type=Int, shape=[3, 3]
    */
    

Negative Indexing

  • Negative indexing is also available That's implementation was hardest for me...

    print(a[~<-1])
    /*
    mfarray = 
    [[[	0,		1,		2],
    [	3,		4,		5],
    [	6,		7,		8]],
    
    [[	9,		10,		11],
    [	12,		13,		14],
    [	15,		16,		17]]], type=Int, shape=[2, 3, 3]
    */
    print(a[-1~<-3])
    /*
    mfarray = 
    	[], type=Int, shape=[0, 3, 3]
    */
    print(a[Matft.reverse])
    //print(a[~<~<-1]) //alias
    //print(a[~<<-1]) //alias
    /*
    mfarray = 
    [[[	18,		19,		20],
    [	21,		22,		23],
    [	24,		25,		26]],
    
    [[	9,		10,		11],
    [	12,		13,		14],
    [	15,		16,		17]],
    
    [[	0,		1,		2],
    [	3,		4,		5],
    [	6,		7,		8]]], type=Int, shape=[3, 3, 3]*/
    

Boolean Indexing

  • You can use boolean indexing.

    <strike>Caution! I don't check performance, so this boolean indexing may be slow</strike>

    Unfortunately, Matft is too slower than numpy...

    (numpy is 1ms, Matft is 7ms...)

    let img = MfArray([[1, 2, 3],
                                   [4, 5, 6],
                                   [7, 8, 9]], mftype: .UInt8)
    img[img > 3] = MfArray([10], mftype: .UInt8)
    print(img)
    /*
    mfarray = 
    [[	1,		2,		3],
    [	10,		10,		10],
    [	10,		10,		10]], type=UInt8, shape=[3, 3]
    */
    

Fancy Indexing

  • You can use fancy indexing!!!

    let a = MfArray([[1, 2], [3, 4], [5, 6]])
                
    a[MfArray([0, 1, 2]), MfArray([0, -1, 0])] = MfArray([999,888,777])
    print(a)
    /*
    mfarray = 
    [[	999,		2],
    [	3,		888],
    [	777,		6]], type=Int, shape=[3, 2]
    */
                
    a.T[MfArray([0, 1, -1]), MfArray([0, 1, 0])] = MfArray([-999,-888,-777])
    print(a)
    /*
    mfarray = 
    [[	-999,		-777],
    [	3,		-888],
    [	777,		6]], type=Int, shape=[3, 2]
    */
    

View

  • Note that returned subscripted mfarray will have base property (is similar to view in Numpy). See numpy doc in detail.

    let a = Matft.arange(start: 0, to: 4*4*2, by: 1, shape: [4,4,2])
                
    let b = a[0~<, 1]
    b[~<<-1] = MfArray([9999]) // cannot pass Int directly such like 9999
    
    print(a)
    /*
    mfarray = 
    [[[	0,		1],
    [	9999,		9999],
    [	4,		5],
    [	6,		7]],
    
    [[	8,		9],
    [	9999,		9999],
    [	12,		13],
    [	14,		15]],
    
    [[	16,		17],
    [	9999,		9999],
    [	20,		21],
    [	22,		23]],
    
    [[	24,		25],
    [	9999,		9999],
    [	28,		29],
    [	30,		31]]], type=Int, shape=[4, 4, 2]
    */
    

Complex

Matft supports Complex!!

But this is beta version. so, any bug may be ocurred.

Please report me by issue

Related Skills

View on GitHub
GitHub Stars145
CategoryDevelopment
Updated26d ago
Forks23

Languages

Swift

Security Score

100/100

Audited on Mar 2, 2026

No findings