Constructing Models

With a few exceptions, most models are constructed by describing the model in JSON format, and passing the JSON-formatted information to the make_model function. There are some convenience functions exposed for backwards compatibility, but as of version 0.14.0, all model construction should go via this route.

At the C++ level, the returned value from the make_model function is a shared_ptr that wraps a pointer to an AbstractModel class. The AbstractModel class is an abstract class which defines the public C++ interface.

In Python, construction is in two parts. First, the model is constructed, which only includes the common methods. Then, the model-specific attributes and methods are attached with the attach_model_specific_methods method.

The JSON structure is of two parts, the kind field is a case-sensitive string defining which model kind is being constructed, and the model field contains all the information needed to build the model. In the case of hard-coded models, nothing is provided in the model field, but it must still be provided.

Also, the argument to make_model must be valid JSON. So if you are working with numpy array datatypes, make sure to convert them to a list (which is convertible to JSON). Example below.

import teqp, numpy as np
teqp.__version__

'0.22.0'

teqp.make_model({'kind': 'vdW1', 'model': {'a': 1, 'b': 2}})

<teqp.teqp.AbstractModel at 0x7f4fd90385f0>

# Fields are case-sensitive
teqp.make_model({'kind': 'vdW1', 'model': {'a': 1, 'B': 2}})

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
Cell In[3], line 2
      1 # Fields are case-sensitive
----> 2 teqp.make_model({'kind': 'vdW1', 'model': {'a': 1, 'B': 2}})

File /opt/conda/lib/python3.11/site-packages/teqp/__init__.py:47, in make_model(*args, **kwargs)
     42 def make_model(*args, **kwargs):
     43     """
     44     This function is in two parts; first the make_model function (renamed to _make_model in the Python interface)
     45     is used to make the model and then the model-specific methods are attached to the instance
     46     """
---> 47     AS = _make_model(*args, **kwargs)
     48     attach_model_specific_methods(AS)
     49     return AS

RuntimeError: :{"B":2,"a":1}': required property 'b' not found in object
|/|\|:{"B":2,"a":1}': validation failed for additional property 'B': instance invalid as per false-schema

# A hard-coded model
teqp.make_model({
    'kind': 'AmmoniaWaterTillnerRoth',
    'model': {}
})

<teqp.teqp.AbstractModel at 0x7f4fd02e30b0>

# Show what to do with numpy array
Tc_K = np.array([100,200])
pc_Pa = np.array([3e6, 4e6])
teqp.make_model({
    "kind": "vdW",
    "model": {
        "Tcrit / K": Tc_K.tolist(),
        "pcrit / Pa": pc_Pa.tolist()
    }
})

<teqp.teqp.AbstractModel at 0x7f4fd02e31d0>

# methane with conventional PC-SAFT
j = {
    'kind': 'PCSAFT',
    'model': {
        'coeffs': [{
            'name': 'methane',
             'BibTeXKey': 'Gross-IECR-2001',
             'm': 1.00,
             'sigma_Angstrom': 3.7039,
             'epsilon_over_k': 150.03,
        }]
    }
}
model = teqp.make_model(j)