Add tags to metadata

[juliasilge]

Closes #570

This PR adds a tags field to pin metadata. It is used like this:

library(pins)
b <- board_rsconnect()
#> Connecting to RSC 2022.09.0 at <https://colorado.rstudio.com/rsc>
b %>% pin_write(
  1:10, 
  "julia.silge/great-numbers", 
  tags = c("i-love-integers", "numbers-are-fun")
)
#> Guessing `type = 'rds'`
#> Writing to pin 'julia.silge/great-numbers'

^{Created on 2022-11-16 with reprex v2.0.2}

You can see how this is stored on Connect here (click "Raw metadata" to see the tags stored in YAML).

This adds a new argument in pin_write() between description and metadata, which could cause problems for any folks who have used metadata by position only. It does not fail silently, though:

library(pins)
b <- board_temp()
b %>% pin_write(1:10, "great-numbers", "json", "My beautiful numbers", "Some numbers for you to enjoy", list(a = "a"))
#> Error in `check_tags()` at pins-r/R/pin-read-write.R:88:2:
#> ! `tags` must be a character vector

#> Backtrace:
#>     ▆
#>  1. ├─b %>% ...
#>  2. └─pins::pin_write(...)
#>  3.   └─pins:::check_tags(tags) at pins-r/R/pin-read-write.R:88:2
#>  4.     └─rlang::abort("`tags` must be a character vector") at pins-r/R/pin-read-write.R:256:4

^{Created on 2022-11-16 with reprex v2.0.2}

I think the number of folks who have used metadata like this (by position) must be very small, though, because there are so many arguments before it. Is this a breaking change? Probably?

For future extensions, on Connect we could try to store these tags as Connect tags. Publishers cannot create tags, so the plan would be to check if the tag exists and then add it to the pin.

[juliasilge]

This change is backwards compatible for reading pins that were stored without tags:

library(pins)
b <- board_rsconnect()
#> Connecting to RSC 2022.09.0 at <https://colorado.rstudio.com/rsc>
b %>% pin_meta("julia.silge/wow-another-mtcars-model")
#> List of 11
#>  $ file       : chr "wow-another-mtcars-model.rds"
#>  $ file_size  : 'fs_bytes' int 5.07K
#>  $ pin_hash   : chr "ebc45bb5da89fbcc"
#>  $ type       : chr "rds"
#>  $ title      : chr "wow-another-mtcars-model: a pinned list"
#>  $ description: chr "An OLS linear regression model"
#>  $ created    : POSIXct[1:1], format: "2022-10-07 14:45:38"
#>  $ api_version: num 1
#>  $ user       : list()
#>  $ name       : chr "julia.silge/wow-another-mtcars-model"
#>  $ local      :List of 4
#>   ..$ dir       : 'fs_path' chr "~/Library/Caches/pins/rsc-e62371cfd77db754024f9c5ed3556a73/3ff71b31-f257-480a-a87b-fc6dc9a7e4dc/63410"
#>   ..$ url       : chr "https://colorado.rstudio.com/rsc/content/3ff71b31-f257-480a-a87b-fc6dc9a7e4dc/_rev63410/"
#>   ..$ version   : chr "63410"
#>   ..$ content_id: chr "3ff71b31-f257-480a-a87b-fc6dc9a7e4dc"

^{Created on 2022-11-16 with reprex v2.0.2}

[juliasilge]

@machow do you mind taking a look, with an eye to a) whether this change would be a problem for pins for Python as it exists today and b) whether you think this approach would work for Python (later, when adding it may be appropriate)?

[machow]

This LGTM, and should be pretty quick to add to python pins!

I left a quick question on whether tags should be added as the last argument to pin_write. It seems like "tags" would be pretty quick to type for kw args, and will let this be backwards compatible (but I get wanting it to come before rarely used arguments 😭)

[juliasilge]

I think you are right @machow that we should just put this at the end. I still kept it before the dots ... which means it could possibly still be a breaking change for folks who have used all existing args by position only (not by name) and then passed something through the dots; I think this must be very rare since there are so many arguments ahead of the dots.

[hadley]

If I was to design this function today, I would put ... after x.

[github-actions]

This pull request has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex: https://reprex.tidyverse.org) and link to this issue.