Assign column types (instead of considering everything is a string)

[niconoe]

[niconoe]

(idea coming from ropensci-archive/finch#25)

That seemed the simplest sensible approach at the time, since the standard doesn't allow to specify data types in the Metafile (which would be great). That means we have to:

• either rely on a on standardized fixed list of "expected type per column" like @damianooldoni did
• either rely on some sort of guess algorithm

(or a combination of both)

It seemed to me that the former might be a bit rigid (different people may use the standard in a different way) while the latter can be difficult to implement properly.

Anyway, implementing it as an optional interpretation layer on top of the current behavior seems helpful to end user, so I'll consider it!

(would the Pandas integration help/can be used for the search algorithm?)

[stijnvanhoey]

@niconoe I do understand the benefits of this, but I would keep this indeed as an optional setting as for some use cases (e.g. validation of inputs pywhip) knowing the data will enter as string maks things easier to build on. Certainly the latter algorithm based approach will be tricky to further build a on (and python-dwca-reader is a typical building block in other implementations).