Import model as space

[AB-Athene]

In case of broader adoption of modelx use by the organization, naturally different individuals/teams start to own different models (modelling logic, data sets, ...). Examples are:

• one team models endowment liability product, another - term
• one team models in-force business, another - new business
• one team models assets, another - liabilities, ...

In this situation ideal process would be

1. Each team produces modelx model object including all the data, parameters and formulas.
2. There is a way to merge all these models into one and build overarching logic on top.

To facilitate it, it seems relatively straightforward extension of the modelx framework to allow import of model as space.

[fumitoh]

This is not documented, but you can actually copy a space to another model, by UserSpace.copy. This usage is experimental, and it probably causes an issue when the elements of the copied space have refs associated with files, but it should work in simple cases as below.

import modelx as mx

# Create Model1:
#
# Model1---A---foo

m1 = mx.new_model()
m1.new_space('A')

@mx.defcells
def foo():
    return 1

# Create Model2:
#
# Model2---B---bar

m2 = mx.new_model()    
m2.new_space('B')

@mx.defcells
def bar():
    return 2

# Copy A to Model2:
#
# Model2---B---bar
#        +-A---foo
    
m1.A.copy(m2)

[AB-Athene]

File refs is one complication. Another is interdependency between spaces in the source model.

[fumitoh]

Yes. File refs can be fixed, but how to treat space interdependency upon copy requires some thoughts.

[AB-Athene]

What do you think about ability to import model as a space with model spaces becoming child spaces of the created space?

[fumitoh]

That’s a good idea. Let me think about how to implement it.