Add sheets_create()

[jennybc]

HERE WE GO 🚀

My next move is to work on the sheets argument .... taking names, taking datasets, etc. But there's already plenty of stuff to decide, so let's get started. I'm going to comment this PR heavily with discussion points.

Below I show a bit of usage.

---

devtools::load_all("~/rrr/googlesheets4")
#> Loading googlesheets4

sheets_create("sheets-create-demo-1")
#>   Spreadsheet name: sheets-create-demo-1
#>                 ID: 17A4RtHDeEi7L3zBX7FSAXtfhwa-GO3pCalLxCKnLzG4
#>             Locale: en_US
#>          Time zone: Etc/GMT
#>        # of sheets: 1
#> 
#> (Sheet name): (Nominal extent in rows x columns)
#>       Sheet1: 1000 x 26

sheets_create("sheets-create-demo-2", locale = "en_CA")
#>   Spreadsheet name: sheets-create-demo-2
#>                 ID: 1aZqeARCmHXAJcLWmZ5xb4-BhNVHOBxeZLU_tHa3CtEI
#>             Locale: en_CA
#>          Time zone: Etc/GMT
#>        # of sheets: 1
#> 
#> (Sheet name): (Nominal extent in rows x columns)
#>       Sheet1: 1000 x 26

sheets_create(
  "sheets-create-demo-3",
  locale = "fr_FR",
  timeZone = "Europe/Paris"
)
#>   Spreadsheet name: sheets-create-demo-3
#>                 ID: 1LoZnD-swifxuodEDcT4Ta-fVRgKLqVKCOQnTHRLPbgE
#>             Locale: fr_FR
#>          Time zone: Europe/Paris
#>        # of sheets: 1
#> 
#> (Sheet name): (Nominal extent in rows x columns)
#>    Feuille 1: 1000 x 26

^{Created on 2019-11-15 by the reprex package (v0.3.0.9001)}

[jennybc]

There are ~200 so-called schemas in the Sheets API. GridRange is an example.

Once you start writing and updating Sheets, you need to create a lot of bodies, in the REST sense, which was not true while we were only reading. And these bodies must be built up of (often nested) schemas. I have to create some sort of machinery around this.

What I do in this PR would not scale nicely to the full set of schemas, but is that necessary? I don't know. I definitely feel a bit like a schema-and-S3-creating robot, though.

But if I did something more automated re: class creation and validation, then it scales and, in fact, that would belong in gargle, for reuse across multiple APIs.

I don't know what to do.

[jennybc]

In addition to ~200 schemas there are lots of enums. They aren't particularly easy to pull out of the discovery document in a coherent way. But you could imagine doing so, in a fit of recursive fury, and create a way to declare something a enum, check supplied values against the valid ones, etc.

[jennybc]

Auto-generating an internal data structure and code around the enums sounds much easier to me (maybe even a clear "yes, do this"?) than doing same for the schemas.

[jennybc]

OK the constructors for S3 classes based on schemas are very minimalist now.

I must confess I don't see any reason to write these constructors. It feels like I could accomplish the same automatically based on the internal .schemas object, i.e. via a hypothetical new_from_schema(schema) function. All they do is construct a list with NULL elements with specific names. But I will trust in the process.

[hadley]

Maybe you shouldn’t then? Maybe it doesn’t make sense to use S3 here at all, and you could just constructor a validation function from the spec?