Add to_dict SFrame method #2755
Conversation
| for key in self.column_names(): | ||
| sa = SArray(self[key]) | ||
| val_list = [] | ||
| for val in sa: |
There was a problem hiding this comment.
Hmm; Pandas actually seems to behave a bit differently. I'm not sure why but it gives dictionaries of dictionaries (where the inner dictionaries have numeric keys ranging from 0 to n for n rows), instead of dictionaries of lists. I suspect this is to better handle sparse dataframes? Example of this behavior:
In [23]: d = df.to_dict()
In [24]: type(d)
Out[24]: dict
In [25]: d.keys()
Out[25]: ['path', 'image']
In [26]: type(d['path'])
Out[26]: dict
In [27]: d['path'].keys()
Out[27]:
[0,
1,
2,
...In any case, we should probably preserve the same behavior here -- let's use a dict instead of a list for the inner structure, using row index as the key.
There was a problem hiding this comment.
I'm not sure I understand pandas' behavior here.
There was a problem hiding this comment.
Pandas always gives a sparse representation (even if the original data was dense) -- i.e. using dictionaries instead of lists, with row index as the dictionary key.
There was a problem hiding this comment.
I'm not sure I have an opinion on this one. I am not a fan of Pandas behavior here, but the goal of #1909 was to match Pandas behavior.
TobyRoseman
changed the title
This PR resolves #1909
Add a to_dict method on SFrame