Another approach for the load_data function would be to wrap the bigrquery library and essentially curate some bigrquery operations while also writing vignettes that would help users write SQL and more advanced queries.
Pros:
1) We already have the BQ project set up, it would be nice to use it in this way.
2) This would negate any questions about data storage raised in #2 (not only for the tabular files, but also geometry files) as these are all in big query
3) This would support both simple usage (show a variable for every state) but also more complex SQL queries against the data which would include spatial functions (show all tracts within ten miles of this ZCTA, for example)
Cons:
1) I think we would be billed for any queries to our BQ project, which is unlikely to be problem but perhaps could be.
2) As far as I can tell we would have to ship with some default auth credentials (doesn't seem like a good idea...), or alternatively, could create a signup process through which we distribute new read-only credentials as needed. Either option would require a good bit of research and follow-through.
Another approach for the load_data function would be to wrap the bigrquery library and essentially curate some bigrquery operations while also writing vignettes that would help users write SQL and more advanced queries.
Pros:
1) We already have the BQ project set up, it would be nice to use it in this way. 2) This would negate any questions about data storage raised in #2 (not only for the tabular files, but also geometry files) as these are all in big query 3) This would support both simple usage (show a variable for every state) but also more complex SQL queries against the data which would include spatial functions (show all tracts within ten miles of this ZCTA, for example)
Cons:
1) I think we would be billed for any queries to our BQ project, which is unlikely to be problem but perhaps could be. 2) As far as I can tell we would have to ship with some default auth credentials (doesn't seem like a good idea...), or alternatively, could create a signup process through which we distribute new read-only credentials as needed. Either option would require a good bit of research and follow-through.