Creating arrays in Python without pre-populating with data

Question

enter code hereIs there a way to create (multidimensional) arrays in Python, without pre-populating these arrays with data?

Specifically I'm looking for something equivalent to the Ironpython solution I found here :

import clr

from System import Array,Boolean,Double

from System.Activator import CreateInstance

a=clr.Reference[Array[str]](Array[str](('')))

b=clr.Reference[Array[float]](Array.CreateInstance(Double,0))

I tryed with ctypes (e.g. a=(c_double*2*3)() ) but this way arrays are prepopulated with zeros. Any idea?

The problem is that I have a COM object with an API which functions are something like:

ret=function(var1,var2,var3,...)

where var1,var2,var3 are multidimensional arrays to be filled by values. If I use Ironpython I can communicate with the COM object via the System module like:

from System.Type import GetTypeFromProgID
from System.Activator import CreateInstance

# Create Sap2000 object
sap = CreateInstance(GetTypeFromProgID("Sap2000.SapObject"))

and then I solve my problem by building var1, var2, var3 with clr.Reference...

With CPython I used the win32com.client module to communicate with the COM object but now I don't know how to build the var1, var2, var3 so that they work in the function.

Thanks,

HERE'S THE CODE:

import win32com.client 
from ctypes import*


def Sap2000():
    sap,SapModel=OpenSap2000(Visible = True)
    NewModel(SapModel)

    #define material property
    ret = SapModel.PropMaterial.SetMaterial("CONC", 2) #MATERIAL_CONCRETE)

    #assign isotropic mechanical properties to material
    ret = SapModel.PropMaterial.SetMPIsotropic("CONC", 3600, 0.2, 0.0000055)

    #create model from template
    ret = SapModel.File.New2DFrame(0, 3, 124, 3, 200) # 0=PortalFrame

    #run analysis
    ret = SapModel.File.Save("C:\SapAPI\xy.sdb")
    ret = SapModel.Analyze.RunAnalysis

    #clear all case and combo output selections
    ret = SapModel.Results.Setup.DeselectAllCasesAndCombosForOutput

    #set case and combo output selections
    ret = SapModel.Results.Setup.SetCaseSelectedForOutput("DEAD")

    #get point displacements

    # Arrays passed by ref
    #Obj=clr.Reference[Array[str]](Array[str](('')))
    Obj=[[""]]
    #Elm=clr.Reference[Array[str]](Array[str](('')))
    Elm=[[""]]
    #LoadCase=clr.Reference[Array[str]](Array[str](('')))
    LoadCase=[[""]]
    #StepType=clr.Reference[Array[str]](Array[str](('')))
    StepType=[[""]]

    #StepNum=clr.Reference[Array[float]](Array.CreateInstance(Double,0))
    StepNum=(c_double*3*2)()
    #U1=clr.Reference[Array[float]](Array.CreateInstance(Double,0))
    U1=(c_double*3*2)()
    #U2=clr.Reference[Array[float]](Array.CreateInstance(Double,0))
    U2=(c_double*3*2)()
    #U3=clr.Reference[Array[float]](Array.CreateInstance(Double,0))
    U3=(c_double*3*2)()
    #R1=clr.Reference[Array[float]](Array.CreateInstance(Double,0))
    R1=(c_double*3*2)()
    #R2=clr.Reference[Array[float]](Array.CreateInstance(Double,0))
    R2=(c_double*3*2)()
    #R3=clr.Reference[Array[float]](Array.CreateInstance(Double,0))
    R3=(c_double*3*2)()
    ret = SapModel.Results.JointDispl("ALL", GroupElm, NumberResults, Obj, Elm, LoadCase, StepType, StepNum, U1, U2, U3, R1, R2, R3)
    print ret




def OpenSap2000(Units='kN_m_C',Visible = False,FileName=''):
    # Create Sap2000 object
    sap = win32com.client.Dispatch("Sap2000.SapObject")

    # Start application
    ret=sap.ApplicationStart(6,Visible,FileName)
    print ret
    SapModel=sap.SapModel

    return sap,SapModel


def NewModel(SapModel,Units='kN_m_C'):
    # Initialize new model
    #SapModel=sap.SapModel
    u=SetUnits(Units)
    ret = SapModel.InitializeNewModel(u)
    # New blank model
    ret = SapModel.File.NewBlank()
    return


if __name__=="__main__":
    Sap2000()

It fails calling the SapModel.Results.JointDispl(..) function.

File "C:\Python26\Lib\site-packages\win32com\client__init__.py", line 456, in ApplyTypes self.oleobj.InvokeTypes(dispid, 0, wFlags, retType, argTypes, *args), TypeError: Objects for SAFEARRAYS must be sequences (of sequences), or a buffer object.

Why is it necessary for there to be no data there? Is it because you think it's faster that way? — Cosmologicon
– Cosmologicon, Commented Jun 17, 2011 at 14:00
I added some specs to the first post about the problem I'm trying to face. @Cosmologicon: the performance is not the problem @lazyr: I'm looking at your link — Paolo
– Paolo, Commented Jun 17, 2011 at 14:18
Why don't you show us the code using win32com.client up to the point it doesn't work. For all the guessing below no one understands your problem. — Mark Tolonen
– Mark Tolonen, Commented Jun 17, 2011 at 14:58

David Webb · Accepted Answer · 2011-06-17 13:43:56Z

The list entries have so have some value in but you could use None.

For a one dimension "array" it's as simple as:

[None] * 3

For a two-dimensional array you have to more careful, as you don't want all your nested lists to the same list, so you can do something like:

[[None] * 2 for _ in range(3)]

Another option is to use a dict with tuples as keys but this doesn't have a fixed size either.

>>> d = {}
>>> d[(1,1)] = 2
>>> d[(1,1)]
2
>>> d[(1,2)]
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
KeyError: (1, 2)

The array module offers typed numeric arrays but these are like a list in that they don't have a fixed size. However, these do need to be populated with a value.

>>> from array import array
>>> a = [array('f',[0] * 2) for _ in range(3)]
>>> a
[array('f', [0.0, 0.0]), array('f', [0.0, 0.0]), array('f', [0.0, 0.0])]
>>> a[1][1] = 3.4
>>> a[1][1] = 'string'
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: a float is required

Sven Marnach · Accepted Answer · 2011-06-17 13:25:20Z

1

If you don't mind using NumPy, you can use numpy.empty(). NumPy is probably the best available implementation of a multidimensional array.

answered Jun 17, 2011 at 13:25

Sven Marnach

608k123 gold badges966 silver badges865 bronze badges

3 Comments

Paolo Over a year ago

It doesn't solve the problem cause numpy.empty(n) returns a numpy array like --> array([ 1.25350215e-163, 9.98394023e-317, 1.58456325e+029], ...)

Mark Tolonen Over a year ago

@Paolo, an array has to contain something. So unpopulated doesn't really have any meaning. Do you mean uninitialized? An array is created faster if it doesn't have to be filled with a known value, and numpy.empty() is an uninitialized array.

Paolo Over a year ago

Please look at the Ironpython topic I linked at the beginning. I just used the same words I found there.

Daren Thomas · Accepted Answer · 2011-06-17 14:09:46Z

1

What should it be pre-populated with? What is your use case?

I think the quickest solution for an array of dimensions (x, y) would be:

a = [[None] * x for i in range(y)]

This will give you a multidimensional array with None values denoting where you haven't set any value yet.

edited Jun 17, 2011 at 14:09

answered Jun 17, 2011 at 13:28

Daren Thomas

70.8k42 gold badges156 silver badges205 bronze badges

6 Comments

David Webb Over a year ago

Inside a you've got the same list three times so, for example, when you change a[1][1] you'll change a[0][1] and a[0][2] at the same time.

Paolo Over a year ago

I should assign a type but not the dimension nor the values. Let's say I want a multidimensional array (2 layers) of doubles or strings (as in the Ironpython example I posted).

Ned Batchelder Over a year ago

You're both wrong: Daren: it need to end with "for i in range(y)", and Dave, this is the right way to avoid the shared list problem.

David Webb Over a year ago

@Ned - Daren has edited his answer since I added my comment. It originally said [[None] * x] * y.

Paolo Over a year ago

Assigning None values doeasn't solve my problem. It's like I need to declare a variable in python...maybe using Cython??

|

Collectives™ on Stack Overflow

Creating arrays in Python without pre-populating with data

3 Answers 3

Comments

3 Comments

6 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

Comments

3 Comments

6 Comments

Your Answer

Sign up or log in

Post as a guest

Related