-
Notifications
You must be signed in to change notification settings - Fork 235
idaes v2 migrate
This guide shows how to migrate IDAES client code developed using the IDAES v1 API so that it can be used with the IDAES v2 API.
The internal structure of the top-level idaes
Python package has undergone significant changes in the v1 -> v2 transition, which resulted in changes to the IDAES import paths:
- In most cases, importable objects (subpackages/modules/classes/functions/...) available in v1 can still be imported from a different location (i.e. the import path has been modified)
- In other cases, the importable object does not have a direct equivalent in the v2 API (i.e. the import path has been removed)
IN most cases, a deprecation warning is available through the v1 compatibility tool which will direct users to the new location where appropriate. For a list of the import path changes, users can refer to the specifications defined here:
In the specifications, the v1 import paths are mapped to the corresponding v2 import path if a direct replacement is available;
otherwise, they are be mapped to None
, or the specification is defined using the without_direct_replacement
method.
If the v1 import path has a direct replacement in v2, it is sufficient to modify the import paths from v1 to v2:
-from idaes.core.util import get_solver
+from idaes.core.solvers import get_solver
If the v1 import path has been removed, a non-direct alternative might still be available. If this is the case and the affected import path is not addressed in this guide, feel free to contact the IDAES team using one of the available support channels.
In IDAES v1, when creating instances of any class inheriting from ProcessBlock
, options had to be supplied as a dictionary object passed to the default
keyword argument (kwarg). In IDAES v2, the same options are are passed directly as keyword arguments.
If the dictionary passed to the default
argument was defined inline as a dict literal (i.e. default={"key1": ..., "key2": ..., ...}
),
then the same key-value pairs should be passed directly as keyword arguments:
-pb = ProcessBlock(default={"dynamic": False})
+pb = ProcessBlock(dynamic=False)
If a pre-defined dictionary was used, the "dict unpacking" operator **
may be used:
my_opts = {"dynamic": False}
-pb = ProcessBlock(default=my_opts)
+pb = ProcessBlock(**my_opts)
The HeatExchanger
API has been streamlined in IDAES v2. The most notable differences with respect to the v1 API are:
- Instead of
side_1
andside_2
, the two sides of the heat exchanger are calledhot_side
andcold_side
respectively - Custom aliases for the two sides are specified as keyword argument (
hot_side_name
andcold_side_name
resp.) when theHeatExchanger
object is created - Instead of
inlet_1
,outlet_1
,inlet_2
,outlet_2
, the inlet and outlet for each side will be accessible as<side-name>_inlet
and<side-name>_outlet
resp., so if e.g.hx = HeatExchanger(hot_side_name="shell", cold_side_name="tube")
, the following attributes will be set on thehx
object:shell_inlet
shell_outlet
tube_inlet
tube_outlet
For most cases, all that is required is to map the previous names for the sides of the heat exchanger and set the hot_side_name
and cold_side_name
attributes. This will allow you to continue to use the old names elsewhere in the flowsheet, including setting configuration arguments for each side of the heat exchanger. An example for a 0D heat exchanger is shown below:
-model.fs.heat_exchanger = HeatExchanger(default={
- "delta_temperature_callback":delta_temperature_amtd_callback,
- "shell":{"property_package": model.fs.properties},
- "tube":{"property_package": model.fs.properties}})
+model.fs.heat_exchanger = HeatExchanger(
+ delta_temperature_callback=delta_temperature_amtd_callback,
+ hot_side_name="shell",
+ cold_side_name="tube",
+ shell={"property_package": model.fs.properties},
+ tube={"property_package": model.fs.properties})
As part of the move to IDAES v2.0, a more flexible API for process costing has been deployed and the older API will be fully deprecated in v2.0.
An example of using the new tools can be found here.
Below is an example of using the inbuilt example costing library (base on the textbook "Process and Product Design Principles: Synthesis, Analysis, and Evaluation" by Seider, Seader, Lewin, and Windagdo, 3rd Ed.) using the old and new APIs:
-m.fs.hx.get_costing(**kwargs)
+from idaes.models.costing.SSLW import SSLWCosting
+m.fs.costing = SSLWCosting()
+m.fs.hx.costing = UnitModelCostingBlock(
+ flowsheet_costing_block=m.fs.costing,
+ costing_method=SSLWCosting.cost_heat_exchanger,
+ costing_method_arguments=**kwargs)
The key differences are:
- the need to import the desired costing package and explicitly declare the flowsheet level costing object,
- the need to explicitly declare a
UnitModelCostingBlock
associated with the unit to be costed, and - the ability to explicitly define the costing method to use when costing the unit (which does not need to come from the same costing package as was used for the flowsheet costing object). Costing packages may implement a default mapping of unit operation types to costing methods, in which case this argument is optional.
The new API supports custom costing correlations in a number of ways. Firstly, the costing_method
argument above can be any user-defined method which implements the correlations for costing a piece of equipment. Thus, any existing custom costing correlations can be adapted to work in the new API with a minimum of effort.
For those looking to build custom costing packages, the inbuilt example can be used as a starting point. The key elements are:
- a
build_global_params
method which defines any global parameters that are required which corresponds to the previousglobal_costing_parameters
function, and - and a library of costing methods for different pieces of equipment as discussed above.
- Set up pre-commit
- Run pytest with coverage report
- Run Pylint locally
- Update the Pyomo version
- Install Pyomo from a local Git clone
- Set up GitHub authentication with GCM
- Handle warnings in pytest