one problem with caching

[nick]

this problem is hard to track as it seems to disappear when you simplify the problem.

I will explain the situation

Model 1 (low res mesh, res_factor=2) see \inundation\data\western_australia\dampier_tsunami_scenario_2006\anuga\outputs\20070416_231316_run

the interesting part is the number of triangle in the mesh see "screen_output_0_2.txt" for details, there is 254859 triangles

Model 2(high res mesh, res_factor=1) see J:\inundation\data\western_australia\dampier_tsunami_scenario_2006\anuga\outputs\20070417_000545_run

there is 499366 triangles

only difference between the 2 models is the "res_factor" in the project.py file

Now running Model 1 again using the exactly same files as before. see J:\inundation\data\western_australia\dampier_tsunami_scenario_2006\anuga\outputs\20070417_102354_run

The model retrieves cached results of the wrong mesh resolution, producing 499366 triangles

I have tried to simplify the problem but it seems to change.

This is not the first time this has occurred and is very frustrating.

Can someone please try and replicate this.

[ole]

This is correct.

Caching create_mesh_from_regions will bypass the function whenever all input parameters hash to a cached value.

Domain depends on the msh file but will always take the latest which has been produced (as dictated by dependencies and input parameters).

A work around is *not* to cache Domain (which only takes few minutes anyway) or to assign a unique name to the stored mesh and possible for create_mesh_from_regions to return that name.

[ole]

Sorry - the work around is to disable caching *both* for create_mesh and Domain.