Modellen vergelijken

[gijsber]

Zou in de Ribasim-NL Model klasse kunnen, maar Ribasim Python is misschien beter.

model.equals --> True/False
model.diff/compare --> lijst met tabellen die niet gelijk zijn?
model.diff/compare -->objecten die niet gelijk zijn?

Zie ook https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.compare.html

[evetion]

@rbruijnshkv @DanielTollenaar Can you comment here on your proposed/expected API?

def equals(modela, modelb): boolean  # what should this compare

We can compare on the network topology (same nodes, same links), and if so (if topology is not equal I expect a comparison on individual node tables doesn't make sense?) we can print the difference in tables? Do we consider spatial differences to count, or is it topology only? It can be quite a list, is that wanted/expected? Do we want compare to work on individual node(tables)?

[Huite]

My two cents:

• equals should return a bool
• A list of difference seems useful, makes sense to implement it per table.
• Compare with https://pandas.pydata.org/docs/reference/api/pandas.testing.assert_frame_equal.html#pandas.testing.assert_frame_equal
• The assert_frame_equal has a bunch of options; I can imagine analogous options are imaginable for ribasim input
• What to return: printing or raising an error isn't ideal (since you're expected to parse by eye). I think returning a dataframe of differences makes sense. At that stage you're free to print it, or write it to csv, etc.
• For a method: maybe difference isn't too bad (see set methods). Equals can call this and return True if length of difference dataframe is 0.

[DanielTollenaar]

As a model developer I want to be able to efficiently check if two models are equal, that basically covers all in database.gpkg (and theoretically toml-file, but let's skip that for now). What would be convenient:

Something that could work: result = model.equals(other_model), results is a list (or tuple) with tables that are not equal, so e.g. ["ManningResistance.Static", "Pump.Static"] means these tables are not the same. If the list is empty I know models are the same.

Than I can use model.manning_resistance.static.df.diff(oter_model.manning_resistance.static.df) to check what has changed in the table using native diff of Pandas (not working in GeoPandas though!): https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.diff.html

@rbruijnshkv anything to add?

[rbruijnshkv]

Great suggestions so far—I agree.

It would be helpful to have functionality that allows the "meta_" columns to be ignored during difference comparisons. I often add additional information (sometimes temporarily) in these columns, which is later used to fill i.e. the control of connection nodes. Having the ability to exclude them would make it easier to identify where the model starts to deviate, which would be very helpful for debugging.