The SkimDataset structure requires every variable to have a unique name. It also merges OMX variables based on time period, so that e.g. BIKETIME__AM and BIKETIME__PM, which would be 2-d arrays in the OMX file, become just two different parts of a 3-d array called BIKETIME in the SkimDataset.
This is problematic when the skims also contain a 2-d array called BIKETIME, as that has no temporal dimension, and it gets loaded into a 2-d array in the SkimDataset, with the same name as the 3-d array, and thus one is overwritten and lost.
A workaround has been added via the omx_ignore_patterns setting in #867, which allows the user to not load certain skims, but this relies on the model developer or user to set this up correctly.
Describe the solution you'd like
We need the input checker to identify this overwriting condition and raise an error if it is happening, so that the user can correct the condition via (1) the omx_ignore_patterns setting, (2) revising the skim generation process to not create the overlapping named skims in the file in the first place, or (3) renaming one or both skims if the users actually wants both skims variables in the model.
The SkimDataset structure requires every variable to have a unique name. It also merges OMX variables based on time period, so that e.g.
BIKETIME__AMandBIKETIME__PM, which would be 2-d arrays in the OMX file, become just two different parts of a 3-d array calledBIKETIMEin the SkimDataset.This is problematic when the skims also contain a 2-d array called
BIKETIME, as that has no temporal dimension, and it gets loaded into a 2-d array in the SkimDataset, with the same name as the 3-d array, and thus one is overwritten and lost.A workaround has been added via the
omx_ignore_patternssetting in #867, which allows the user to not load certain skims, but this relies on the model developer or user to set this up correctly.Describe the solution you'd like We need the input checker to identify this overwriting condition and raise an error if it is happening, so that the user can correct the condition via (1) the
omx_ignore_patternssetting, (2) revising the skim generation process to not create the overlapping named skims in the file in the first place, or (3) renaming one or both skims if the users actually wants both skims variables in the model.