Add pb_read and pb_write functions

[tanho63]

Closes #97.

I thought briefly about making this a wrapper around pb_download_url + read function that accepts URLs, but I don't think it had the flexibility I wanted plus I ran into issues downloading from private repositories that I later learned was around not being able to pass an auth token to it.

I think this is the most flexible approach to the problem but would love to hear any thoughts

[cboettig]

this strategy (specifically the guess_read_function()) always makes me nervous. While all abstractions are a bit leaky, I've always found attempts to write a generic "read" function as an abstract method particularly leaky. There's just too much going on in read_ methods to abstract away (what about other data serializations, like spatial formats? what about lazy reads / remote reads etc).

I'm fine with a convenience wrapper that makes common use patterns more concise. Maybe all that's needed is a bit more documentation saying something like 'for common formats such as ..., along with advice about how to bypass this convenience if a user prefers their own read function (say, from readr` package, or for some other format).

[tanho63]

too much going on in read_ methods to abstract away (what about other data serializations, like spatial formats?

I believe this will fail on spatial formats since it would not be one of csv, tsv, rds, parquet? (unfamiliar with how geoparquet works and whether arrow::read_parquet will process geoparquet, but I assume yes?).

what about lazy reads / remote reads etc?

Yep, this will read eagerly by design/default, and maybe it's a bad thing for folks who should be thinking about optimizing - however uninformed users would currently do pb_download anyways so it's not necessarily much different than that?

I agree with improving the docs, e.g.

• explaining that cloud native is best performance (more clearly linking to vignette)
• demonstrating ways to read from URL (maybe adding to getting started vignette?)
• better demonstrating examples of passing in a different read function and explaining what is supported ?

[tanho63]

Flow state hit me like a bus, many apologies for this PR running away from me. Since your last review (diff), I:

• gitignored + got rid of the stored docs/ folder because ropensci builds pkgdown externally + I've been using it to test out how the vignette looks after generating it
• updated DESCRIPTION with the newly shortened description from the readme (because I realized it hadn't been updated)
• fleshed out man files for read_function and write_function args, added man files for guess_read_function and guess_write_function that fully explain how it gets mapped
• added various \dontshow blocks to hide the interactive() and try() blocks as inspired by the examplesIf
• updated README to not evaluate any chunks except for regenerating codemeta (improves syntax highlighting in RStudio to use {r} instead of just r)
• rewrote vignette/piggyback.Rmd again to try and address your various feedback points - this is probably the main thing that needs looked at

[cboettig]

looks good!