Row should be dict not list

[saulshanabrook]

Would it make sense for rows to behave like dictionaries instead of lists? I would love to be able to do key based field retrieval, instead of index based.

[iurisilvio]

I like this idea, but it requires some internal changes. Row doesn't know anything about headers, but it is a reasonable change.

We can't remove index based support, so it should work both ways.

[DanielDisisto]

PR #268 looks like a great solution, and for me at least would make tablib much more useful. Is this being considered for merge? Given the change to Dataset.__getitem__ return type, I understand this is not a simple decision to make, but I think it would be beneficial.

If the change is considered too breaking, how about leaving __getitem__ return a tuple as currently does, but adding a Dataset.get_row(i) which returns the Row instance (just to prevent having to access Dataset._data directly)? It's a bit clunky, but avoids all backward-compatibility issues while still providing a way to access row data by column name.

EDIT: I'd somehow missed the fact that getting a specific item on the table by row index & column name can already be achieved with Dataset.dict[row][column_name]. The only issue I have with that is that it is quite expensive to do this repeatedly, running the _package() method each time it's accessed to rebuild the entire dict.

[ms8r]

Good point - I think we could get around the breaking issue in #268 by placing the logic of #268 behind the Dataset.dict API (less expensive than running _package each time) while continuing to return tuples for plain vanilla __getitem__ access. Will be a few weeks until I have some time for this again but will look into amending #268.