[okfn-labs] Some further improvements to Recline Data Explorer

[Rufus Pollock]

On 18 March 2012 04:08, Tim McNamara <paperless at timmcnamara.co.nz> wrote:
> My 2c, and I may be biased because I have spent a lot of time with
> Refine. Here are a few things that I would really like:
>
>  - clone columns
>  - demonstrable rollback path (pref. sharing changeset format w/
> Google Refine to allow interoperability)
>
> Both of these things make it much easier to safely experiment with the data.

Indeed. Though this isn't completely trivial in that we need a
changeset format and a way to store and apply these. Want to open an
issue for discussing this at:

<https://github.com/okfn/recline/issues?milestone=&labels=&state=open>

@Max: do you have any thoughts on roll-back back. I'd guess on Couch
this would be relatively trivial because of the versioning (??)

> Maybe we could do some data introspection to produce aggregate stats,
> ala R's data frame, for each of the columns?

Great point. If you were using the ElasticSearch backend this would be
really easy due to its great stats functionality.

At the moment Recline is designed to work with multiple backends
including in-memory. I'm wondering if we should require for proper
data wrangling that you use just one -- and I'd propose ElasticSearch.
Otherwise we have to support all the relevant operations across all
the backends.

Rufus