MAINT: save data and rebuild model

[eigenfoo]

I forgot to save the data as part of the trace. This does that and also allows us to rebuild_model.

[aseyboldt]

These dimension names should match the ones used in the model (line 59)

[aseyboldt]

That way xarray knows that they are the same dimension, and doesn't store the coordinates twice.

[eigenfoo]

Ah, so then:

author -> meta_user_id
algo -> meta_algorithm_id
backtest -> meta_code_id

Right?

[aseyboldt]

yes

[eigenfoo]

Done!

[aseyboldt]

thanks

[eigenfoo]

@aseyboldt latest Jenkins build failed because of the same dim names

E           ValueError: conflicting MultiIndex level name(s):
E           'meta_code_id' (dim_0), (meta_code_id)
E           'meta_user_id' (dim_0), (meta_user_id)
E           'meta_algorithm_id' (dim_0), (meta_algorithm_id)

[eigenfoo]

I think the easiest solution to this would be to change the dim names for _data to be meta_code_id__, meta_user_id__, etc. The underscores can be added or deleted when saving and loading the data.

I'm not sure if this is the most elegant solution though, since the user may wait to access the data by itself, and the underscores would make it a pain.

@aseyboldt what do you think?

[aseyboldt]

It should be possible to fix that. We somehow need to tell xarray to reuse those coords. Give me a couple of minutes.

[aseyboldt]

This seems to be more difficult than I thought: pydata/xarray#1603

The dimension meta_user_id in _data is not the same as the other meta_user_id. The first one has repeated entries, so that the length is equal len(data), while the second has len num_authors. So you could use new names like data_meta_user_id. This isn't all that nice, but I think it makes some sense.

[eigenfoo]

data storage is necromancy

[aseyboldt]

I created pydata/xarray#2299.

I think we should just add a

d.index.names = ['data_meta_user_id', 'data_meta_...']

just before the assignment to the xarray dataset.