jbgruber / atrrr Goto Github PK

Search skeet-endpoint + Pagination

• Keywords-based search
• Pagination example, how to deal with cursor?

Get profiles + follow graph

• Take an account
• Get description
• Get all followers/follows
• Get description of all followers/follows

Create a post

• support for media

Get list of members in a feed

• Get the list of people in the rstats feed
• Top poster from a specific feed
• Top popular poster from a specific feed

Get likes from a post

• Most famous rstats post
• Like anlaysis
• Eventual temporal structure

If no github creds are available, users are asked to create and enter them. So I'm reversing the install command to remotes.

Since there was at some point a package called ATR on CRAN, the name is blocked and we can't submit the package under this name. We thus decided to rename the package to atrrr. You can easily move your old tokens to work in the new package with:

R.utils::copyDirectory(tools::R_user_dir("atr", "cache"), 
                       tools::R_user_dir("atrrr", "cache"))

Nothing else changes, except that you have to change library(atr) and atr:: call to the new atrrr.

Would be cool if people could use Markdown syntax like [test](https://github.com/JBGruber/atr) to add custom link texts to links.

It looks like there is a limit how many actors can go into the get_user_info function. Below 26 is fine, but then:

library(atrrr)
df <- search_post("Amsterdam", limit = 200)
#> ℹ Found an *unknown total* of posts that fit the query
#> ⠙ Got 97 posts, but there is more.. [1.2s]
#> ⠹ Got 192 posts, but there is more.. [1.6s]
#> ⠸ Got 285 posts, but there is more.. [1.9s]
#> ℹ Parsing 285 results.
#> ✔ Got 285 results. All done!
#> 
user_info <- get_user_info(actor = df$author_handle)
#> Error in `httr2::req_perform()` at atrrr/R/utils.r:57:5:
#> ! HTTP 400 Bad Request.
#> • error type: InvalidRequest
#> • message: Error: actors must not have more than 25 elements

^{Created on 2024-02-13 with reprex v2.1.0}

We should fix this, since we take care of pagination elsewhere (e.g., in search_post). So why not here.

Once #1 and #2 are done (maybe they are already done, I haven't checked all todos), we should aim to push this to CRAN. From my experience, any tests make it rather harder to pass CRAN checks, so I probably wouldn't bother finishing #3 just now. So what's left to do is:

• resolve note about the regexs data. I think it would make more sense to have the regular expressions as internal data anyway, but I wasn't aware that option existed at the time
• we should take another stab at pre-computing the vignettes (see this) otherwise cran tests will get mad
• CRAN has a weird policy that tests are not allowed to leave behind any files outside the package directory, except in tempdir. This includes rendering the examples, readme and vignettes. We need to check if anything produces a file (auth() does for sure).

Anything else?

Hi,

It seems that for me adding hashtags to skeets does not integrate fully into Bluesky's framework... i.e. they show up in the skeet, but do not hyperlink. Am I the only one noticing that?

Thanks, Philipp

Not sure if this is an actual issue or user error. I can only find my handle when I add .bsky.social

library(atr)
search_user("schochastics")
#> ℹ Parsing 0 results.
#> ✔ Got 0 results. All done!
#> 
#> # A tibble: 0 × 0
search_user("schochastics.bsky.social")
#> ℹ Parsing 1 results.
#> ✔ Got 1 results. All done!
#> 
#> # A tibble: 1 × 8
#>   did             handle display_name description avatar indexed_at viewer_muted
#>   <chr>           <chr>  <chr>        <chr>       <chr>  <chr>      <lgl>       
#> 1 did:plc:sv2aev… schoc… David Schoch "Team Lead… https… 2023-11-0… FALSE       
#> # ℹ 1 more variable: viewer_blocked_by <lgl>

^{Created on 2023-11-10 with reprex v2.0.2}

I would suggest to implement these user-facing functions first:

Un/follow functions might all need to use follow.json, not sure. Couldn't find the endpoint to look up feed followers.

Running into issues on a virtual Windows server with R (4.1.3). The atr package installs without issues, and authentication works fine. Retrieving skeets yields the expected results (e.g. your example feeds <- get_feeds_created_by(actor = "andrew.heiss.phd" works without errors).

But, I’m running into an issue when posting:

Although the response claims that the message was posted, it never shows up on the feed.

If helpful, here’s the output from sessionInfo:

Given that the posts at your end 'react' to mine, they cannot possibly have happened six hours earlier:

Is a simpler explanation that maybe the timestamp is a mix of UTC and a TZ? Why would my desktop here in the America/Central timezone show a difference of six hours (especially given that Amsterdam to Chicago should be seven hours). On the other hand, UTC is six hours from me. Maybe (and I am just guessing here, based on many scars from fighting similar problems on the parsing and formatting side) you format it such that it seen as UTC but does properly account for current UTC?

Tests are a little tricky. As the protocol might change rapidly, it does not make much sense to mock server replies. Some of the endpoints also don't behave exactly like advertised, so testing them against the described output might be a pain for now. I would suggest to only test the exported functions against the API and the auto-generated ones against the reply defined in the lexicon. We won't learn a lot from the latter, but at least all the code is covered and we can play around with the generator without causing issues.