R/feed.r

In atrrr: Wrapper for the 'AT' Protocol Behind 'Bluesky'

#' A view of an actor's skeets.
#'
#' @param actor user handle to retrieve feed for.
#' @param filter get only certain post/repost types. Possible values "posts_with_replies", "posts_no_replies", "posts_with_media", and "posts_and_author_threads".
#' @inheritParams get_followers
#'
#' @returns a data frame (or nested list) of posts
#' @export
#'
#' @examples
#' \dontrun{
#' andrews_posts <- get_skeets_authored_by("andrew.heiss.phd")
#' }
get_skeets_authored_by <- function(
  actor,
  limit = 25L,
  filter = NULL,
  cursor = NULL,
  parse = TRUE,
  verbose = NULL,
  .token = NULL
) {
  res <- list()
  req_limit <- ifelse(limit > 100, 100, limit)
  last_cursor <- NULL
  allowed_values <- c(
    "posts_with_replies",
    "posts_no_replies",
    "posts_with_media",
    "posts_and_author_threads"
  )
  if (!filter %in% allowed_values && !is.null(filter)) {
    cli::cli_abort("{.code filter} must be one of {.val {allowed_values}}")
  }

  if (verbosity(verbose)) {
    cli::cli_progress_bar(
      format = "{cli::pb_spin} Got {length(res)} skeets, but there is more.. [{cli::pb_elapsed}]",
      format_done = "Got {length(res)} records. All done! [{cli::pb_elapsed}]"
    )
  }

  while (length(res) < limit) {
    resp <- do.call(
      what = app_bsky_feed_get_author_feed,
      args = list(
        actor = actor,
        limit = req_limit,
        cursor = last_cursor,
        filter = filter,
        .token = .token,
        .return = "json"
      )
    )

    last_cursor <- resp$cursor
    res <- c(res, resp$feed)

    if (is.null(resp$cursor)) {
      break
    }
    if (verbosity(verbose)) cli::cli_progress_update(force = TRUE)
  }

  if (verbosity(verbose)) {
    cli::cli_progress_done()
  }

  if (parse) {
    if (verbosity(verbose)) {
      cli::cli_progress_step("Parsing {length(res)} results.")
    }
    out <- parse_feed(res)
    out$is_reskeet <- out$author_handle != actor
    if (verbosity(verbose)) {
      cli::cli_process_done(msg_done = "Got {nrow(out)} results. All done!")
    }
  } else {
    out <- res
  }

  attr(out, "last_cursor") <- last_cursor
  return(out)
}

#' A view of the feed created by an actor.
#'
#' @param actor user handle to retrieve feed from
#' @inheritParams get_followers
#'
#' @returns a data frame (or nested list) of feeds
#' @export
#'
#' @examples
#' \dontrun{
#' feed <- get_feeds_created_by("andrew.heiss.phd")
#' }
get_feeds_created_by <- function(
  actor,
  limit = 25L,
  cursor = NULL,
  parse = TRUE,
  verbose = NULL,
  .token = NULL
) {
  res <- list()
  req_limit <- ifelse(limit > 100, 100, limit)
  last_cursor <- NULL

  if (verbosity(verbose)) {
    cli::cli_progress_bar(
      format = "{cli::pb_spin} Got {length(res)} feeds, but there is more.. [{cli::pb_elapsed}]",
      format_done = "Got {length(res)} records. All done! [{cli::pb_elapsed}]"
    )
  }

  while (length(res) < limit) {
    resp <- do.call(
      what = app_bsky_feed_get_actor_feeds,
      args = list(
        actor = actor,
        limit = req_limit,
        cursor = last_cursor,
        .token = .token,
        .return = "json"
      )
    )

    last_cursor <- resp$cursor
    res <- c(res, resp$feed)

    if (is.null(resp$cursor)) {
      break
    }
    if (verbosity(verbose)) cli::cli_progress_update(force = TRUE)
  }

  if (verbosity(verbose)) {
    cli::cli_progress_done()
  }

  if (parse) {
    if (verbosity(verbose)) {
      cli::cli_progress_step("Parsing {length(res)} results.")
    }

    out <- parse_feeds_list(res)

    if (verbosity(verbose)) {
      cli::cli_process_done(msg_done = "Got {nrow(out)} results. All done!")
    }
  } else {
    out <- res
  }

  attr(out, "last_cursor") <- last_cursor
  return(out)
}

#' Get the skeets from a specific feed
#'
#' Get the skeets that would be shown when you open the given feed
#'
#' @param feed_url The url of the requested feed
#' @inheritParams search_user
#'
#' @returns a data frame (or nested list) of posts
#' @export
#'
#' @examples
#' \dontrun{
#' # use the URL of a feed
#' get_feed("https://bsky.app/profile/did:plc:2zcfjzyocp6kapg6jc4eacok/feed/aaaeckvqc3gzg")
#'
#' # or search for a feed by name
#' res <- search_feed("#rstats")
#' get_feed(res$uri[1])
#' }
get_feed <- function(
  feed_url,
  limit = 25L,
  cursor = NULL,
  parse = TRUE,
  verbose = NULL,
  .token = NULL
) {
  res <- list()
  req_limit <- ifelse(limit > 100, 100, limit)
  last_cursor <- NULL

  if (verbosity(verbose)) {
    cli::cli_progress_bar(
      format = "{cli::pb_spin} Got {length(res)} skeets, but there is more.. [{cli::pb_elapsed}]",
      format_done = "Got {length(res)} records. All done! [{cli::pb_elapsed}]"
    )
  }

  feed_url <- convert_http_to_at(feed_url, .token = .token)

  while (length(res) < limit) {
    resp <- do.call(
      what = app_bsky_feed_get_feed,
      args = list(
        feed = feed_url,
        limit = req_limit,
        cursor = last_cursor,
        .token = .token,
        .return = "json"
      )
    )

    last_cursor <- resp$cursor
    res <- c(res, resp$feed)

    if (is.null(resp$cursor)) {
      break
    }
    if (verbosity(verbose)) cli::cli_progress_update(force = TRUE)
  }

  if (verbosity(verbose)) {
    cli::cli_progress_done()
  }

  if (parse) {
    if (verbosity(verbose)) {
      cli::cli_progress_step("Parsing {length(res)} results.")
    }
    out <- parse_feed(res)
    if (verbosity(verbose)) {
      cli::cli_process_done(msg_done = "Got {nrow(out)} results. All done!")
    }
  } else {
    out <- res
  }
  attr(out, "last_cursor") <- last_cursor
  return(out)
}

#' Search a specific feed
#'
#' Search the feed named after a given query
#'
#' @param query The term to be searched
#' @inheritParams search_user
#'
#' @returns a data frame (or nested list) of posts
#' @export
#'
#' @examples
#' \dontrun{
#' search_feed("rstats")
#' }
search_feed <- function(
  query,
  limit = 25L,
  cursor = NULL,
  parse = TRUE,
  verbose = NULL,
  .token = NULL
) {
  res <- list()
  req_limit <- ifelse(limit > 100, 100, limit)
  last_cursor <- NULL

  if (verbosity(verbose)) {
    cli::cli_progress_bar(
      format = "{cli::pb_spin} Got {length(res)} skeets, but there is more.. [{cli::pb_elapsed}]",
      format_done = "Got {length(res)} records. All done! [{cli::pb_elapsed}]"
    )
  }

  while (length(res) < limit) {
    resp <- do.call(
      what = app_bsky_unspecced_get_popular_feed_generators,
      args = list(
        query = query,
        limit = req_limit,
        cursor = last_cursor,
        .token = .token,
        .return = "json"
      )
    )

    last_cursor <- resp$cursor
    res <- c(res, resp$feed)

    if (is.null(resp$cursor)) {
      break
    }
    if (verbosity(verbose)) cli::cli_progress_update(force = TRUE)
  }

  if (verbosity(verbose)) {
    cli::cli_progress_done()
  }

  if (parse) {
    if (verbosity(verbose)) {
      cli::cli_progress_step("Parsing {length(res)} results.")
    }
    out <- parse_feeds_list(res)
    if (verbosity(verbose)) {
      cli::cli_process_done(msg_done = "Got {nrow(out)} results. All done!")
    }
  } else {
    out <- res
  }
  attr(out, "last_cursor") <- last_cursor
  return(out)
}


#' Get your own timeline
#'
#' Get the posts that would be shown when you open the Bluesky app or website.
#'
#' @param algorithm algorithm used to sort the posts
#' @inheritParams search_user
#'
#' @returns a data frame (or nested list) of posts
#' @export
#'
#' @examples
#' \dontrun{
#' get_own_timeline()
#' get_own_timeline(algorithm = "reverse-chronological")
#' }
get_own_timeline <- function(
  algorithm = NULL,
  limit = 25L,
  cursor = NULL,
  parse = TRUE,
  verbose = NULL,
  .token = NULL
) {
  res <- list()
  req_limit <- ifelse(limit > 100, 100, limit)
  last_cursor <- NULL

  if (verbosity(verbose)) {
    cli::cli_progress_bar(
      format = "{cli::pb_spin} Got {length(res)} skeets, but there is more.. [{cli::pb_elapsed}]",
      format_done = "Got {length(res)} records. All done! [{cli::pb_elapsed}]"
    )
  }

  while (length(res) < limit) {
    resp <- do.call(
      what = app_bsky_feed_get_timeline,
      args = list(
        algorithm = algorithm,
        limit = req_limit,
        cursor = last_cursor,
        .token = .token,
        .return = "json"
      )
    )

    last_cursor <- resp$cursor
    res <- c(res, resp$feed)

    if (is.null(resp$cursor)) {
      break
    }
    if (verbosity(verbose)) cli::cli_progress_update(force = TRUE)
  }

  if (verbosity(verbose)) {
    cli::cli_progress_done()
  }

  if (parse) {
    if (verbosity(verbose)) {
      cli::cli_progress_step("Parsing {length(res)} results.")
    }
    out <- parse_timeline(res)
    if (verbosity(verbose)) {
      cli::cli_process_done(msg_done = "Got {nrow(out)} results. All done!")
    }
  } else {
    out <- res
  }
  attr(out, "last_cursor") <- last_cursor
  return(out)
}


#' Get likes/reposts of a skeet
#'
#' @param post_url the URL of a skeet for which to retrieve who liked/reposted it.
#' @inheritParams search_user
#'
#' @returns a data frame (or nested list) of likes/reposts
#'
#' @export
#'
#' @examples
#' \dontrun{
#' get_likes("https://bsky.app/profile/jbgruber.bsky.social/post/3kbi55xm6u62v")
#' get_reposts("https://bsky.app/profile/jbgruber.bsky.social/post/3kbi55xm6u62v")
#' }
get_likes <- function(
  post_url,
  limit = 25L,
  cursor = NULL,
  parse = TRUE,
  verbose = NULL,
  .token = NULL
) {
  uri <- convert_http_to_at(post_url, .token = .token)

  res <- list()
  req_limit <- ifelse(limit > 100, 100, limit)
  last_cursor <- NULL

  if (verbosity(verbose)) {
    cli::cli_progress_bar(
      format = "{cli::pb_spin} Got {length(res)} like entries, but there is more.. [{cli::pb_elapsed}]",
      format_done = "Got {length(res)} records. All done! [{cli::pb_elapsed}]"
    )
  }

  while (length(res) < limit) {
    resp <- do.call(
      what = app_bsky_feed_get_likes,
      args = list(
        uri = uri,
        cid = NULL,
        limit = req_limit,
        cursor = last_cursor,
        .token = .token,
        .return = "json"
      )
    )

    last_cursor <- resp$cursor
    res <- c(res, resp$likes)

    if (is.null(resp$cursor)) {
      break
    }
    if (verbosity(verbose)) cli::cli_progress_update(force = TRUE)
  }

  if (verbosity(verbose)) {
    cli::cli_progress_done()
  }

  if (parse) {
    if (verbosity(verbose)) {
      cli::cli_progress_step("Parsing {length(res)} results.")
    }

    out <- parse_likes(res)

    if (verbosity(verbose)) {
      cli::cli_process_done(msg_done = "Got {nrow(out)} results. All done!")
    }
  } else {
    out <- res
  }
  attr(out, "last_cursor") <- last_cursor
  return(out)
}


#' Get likes of a user
#'
#' @param actor user handle to retrieve likes for.
#' @inheritParams search_user
#'
#' @returns a data frame (or nested list) of likes/reposts
#'
#' @export
#'
#' @examples
#' \dontrun{
#' get_actor_likes("jbgruber.bsky.social")
#' }
get_actor_likes <- function(
  actor,
  limit = 25L,
  cursor = NULL,
  parse = TRUE,
  verbose = NULL,
  .token = NULL
) {
  res <- list()
  req_limit <- ifelse(limit > 100, 100, limit)
  last_cursor <- NULL

  if (verbosity(verbose)) {
    cli::cli_progress_bar(
      format = "{cli::pb_spin} Got {length(res)} like entries, but there is more.. [{cli::pb_elapsed}]",
      format_done = "Got {length(res)} records. All done! [{cli::pb_elapsed}]"
    )
  }

  while (length(res) < limit) {
    resp <- do.call(
      what = app_bsky_feed_get_actor_likes,
      args = list(
        actor = actor,
        limit = req_limit,
        cursor = last_cursor,
        .token = .token,
        .return = "json"
      )
    )

    last_cursor <- resp$cursor
    res <- c(res, resp$feed)

    if (is.null(resp$cursor)) {
      break
    }
    if (verbosity(verbose)) cli::cli_progress_update(force = TRUE)
  }

  if (verbosity(verbose)) {
    cli::cli_progress_done()
  }

  if (parse) {
    if (verbosity(verbose)) {
      cli::cli_progress_step("Parsing {length(res)} results.")
    }
    out <- parse_timeline(res)
    if (verbosity(verbose)) {
      cli::cli_process_done(msg_done = "Got {nrow(out)} results. All done!")
    }
  } else {
    out <- res
  }
  attr(out, "last_cursor") <- last_cursor
  return(out)
}


#' @rdname get_likes
#' @export
get_reposts <- function(
  post_url,
  limit = 25L,
  cursor = NULL,
  parse = TRUE,
  verbose = NULL,
  .token = NULL
) {
  uri <- convert_http_to_at(post_url, .token = .token)

  res <- list()
  req_limit <- ifelse(limit > 100, 100, limit)
  last_cursor <- NULL

  if (verbosity(verbose)) {
    cli::cli_progress_bar(
      format = "{cli::pb_spin} Got {length(res)} reposts, but there is more.. [{cli::pb_elapsed}]",
      format_done = "Got {length(res)} reposts. All done! [{cli::pb_elapsed}]"
    )
  }

  while (length(res) < limit) {
    resp <- do.call(
      what = app_bsky_feed_get_reposted_by,
      args = list(
        uri = uri,
        cid = NULL,
        limit = req_limit,
        cursor = last_cursor,
        .token = .token,
        .return = "json"
      )
    )

    last_cursor <- resp$cursor
    res <- c(res, resp$repostedBy)

    if (is.null(resp$cursor)) {
      break
    }
    if (verbosity(verbose)) cli::cli_progress_update(force = TRUE)
  }

  if (verbosity(verbose)) {
    cli::cli_progress_done()
  }

  if (parse) {
    if (verbosity(verbose)) {
      cli::cli_progress_step("Parsing {length(res)} results.")
    }

    out <- parse_actors(res)

    if (verbosity(verbose)) {
      cli::cli_process_done(msg_done = "Got {nrow(out)} results. All done!")
    }
  } else {
    out <- res
  }
  attr(out, "last_cursor") <- last_cursor
  return(out)
}


#' Get likes of a feed
#'
#' @param feed_url the URL of a feed for which to retrieve who liked it.
#' @inheritParams search_user
#'
#' @returns a data frame (or nested list) of likes/reposts
#'
#' @examples
#' \dontrun{
#' # use the URL of a feed
#' get_feed_likes("https://bsky.app/profile/did:plc:2zcfjzyocp6kapg6jc4eacok/feed/aaaeckvqc3gzg")
#'
#' # or search for a feed by name
#' res <- search_feed("#rstats")
#' get_feed_likes(res$uri[1])
#' }
#' @export
get_feed_likes <- function(
  feed_url,
  limit = 25L,
  cursor = NULL,
  parse = TRUE,
  verbose = NULL,
  .token = NULL
) {
  uri <- convert_http_to_at(feed_url, .token = .token)

  res <- list()
  req_limit <- ifelse(limit > 100, 100, limit)
  last_cursor <- NULL

  if (verbosity(verbose)) {
    cli::cli_progress_bar(
      format = "{cli::pb_spin} Got {length(res)} like entries, but there is more.. [{cli::pb_elapsed}]",
      format_done = "Got {length(res)} records. All done! [{cli::pb_elapsed}]"
    )
  }

  while (length(res) < limit) {
    resp <- do.call(
      what = app_bsky_feed_get_likes,
      args = list(
        uri = uri,
        cid = NULL,
        limit = req_limit,
        cursor = last_cursor,
        .token = .token,
        .return = "json"
      )
    )

    last_cursor <- resp$cursor
    res <- c(res, resp$likes)

    if (is.null(resp$cursor)) {
      break
    }
    if (verbosity(verbose)) cli::cli_progress_update(force = TRUE)
  }

  if (verbosity(verbose)) {
    cli::cli_progress_done()
  }

  if (parse) {
    if (verbosity(verbose)) {
      cli::cli_progress_step("Parsing {length(res)} results.")
    }
    out <- parse_likes(res)
    if (verbosity(verbose)) {
      cli::cli_process_done(msg_done = "Got {nrow(out)} results. All done!")
    }
  } else {
    out <- res
  }
  attr(out, "last_cursor") <- last_cursor
  return(out)
}


#' Get all skeets in a thread
#'
#' Retrieve all skeets in a thread (all replies to an original skeet by any
#' author). It does not matter if you use the original skeet or any reply as
#' \code{post_url}.
#'
#' @param post_url the URL of any skeet in a thread.
#' @inheritParams search_user
#'
#' @returns a data frame of skeets
#' @export
#'
#' @examples
#' \dontrun{
#' get_thread("https://bsky.app/profile/jbgruber.bsky.social/post/3kbi57u4sys2l")
#' }
get_thread <- function(post_url, parse = TRUE, .token = NULL) {
  post_uri <- convert_http_to_at(post_url, .token = .token)
  root <- do.call(
    app_bsky_feed_get_post_thread,
    list(post_uri, .token = .token)
  ) |>
    get_thread_root()
  thread <- do.call(
    app_bsky_feed_get_post_thread,
    list(purrr::pluck(root, "post", "uri"), .token = .token)
  )

  if (parse) {
    return(parse_threads(thread))
  } else {
    thread
  }
}


#' Get all replies
#'
#' Get all replies and replies on replies of a skeet.
#'
#' @param post_url the URL of a skeet.
#' @inheritParams search_user
#'
#' @returns a data frame of skeets
#' @export
#'
#' @examples
#' \dontrun{
#' get_replies("https://bsky.app/profile/jbgruber.bsky.social/post/3kbi57u4sys2l")
#' }
get_replies <- function(post_url, .token = NULL) {
  post_uri <- convert_http_to_at(post_url, .token = .token)
  replies <- do.call(
    app_bsky_feed_get_post_thread,
    list(post_uri, .token = .token)
  )
  return(parse_threads(replies))
}


#' Post a skeet
#'
#' @param text Text to post
#' @param in_reply_to URL or URI of a skeet this should reply to.
#' @param quote URL or URI of a skeet this should quote.
#' @param image,video path to an image or video to post.
#' @param image_alt alt text for the image.
#' @param created_at time stamp of the post.
#' @param labels can be used to label a post, for example "!no-unauthenticated",
#'   "porn", "sexual", "nudity", or "graphic-media".
#' @param langs indicates human language(s) (up to 3) of post's primary text
#'   content.
#' @param tags additional hashtags, in addition to any included in post text and
#'   facets.
#' @param link instead of adding a link in text (gets parsed automatically),
#'   it's also possible to add a link directly (and save some characters).
#' @param preview_card logical. Display a preview card for links included in the
#'   `text` or `link` (if images or videos are included, they take precedence).
#'   alternatively, fetch a card with [fetch_preview()] and supply the object
#'   here.
#' @param post_url URL or URI of post to delete.
#' @param .reply a pre-built reply list with `root` and `parent` elements (each
#'   containing `uri` and `cid`). Used for [post_thread()] not really important
#'   for users.
#' @inheritParams search_user
#'
#' @returns list of the URI and CID of the post (invisible)
#' @export
#'
#' @examples
#' \dontrun{
#' post("Hello from #rstats with {atrrr}")
#' }
post <- function(
  text,
  in_reply_to = NULL,
  quote = NULL,
  image = NULL,
  image_alt = NULL,
  video = NULL,
  link = NULL,
  created_at = Sys.time(),
  labels = NULL,
  langs = NULL,
  tags = NULL,
  preview_card = TRUE,
  verbose = NULL,
  .token = NULL,
  .reply = NULL
) {
  cli::cli_progress_step(
    msg = "Request to post {.emph {text}}",
    msg_done = "Posted {.emph {text}}",
    msg_failed = "Something went wrong"
  )

  repo <- get_token()[["handle"]]
  collection <- "app.bsky.feed.post"

  record <- list(
    "$type" = "app.bsky.feed.post",
    "text" = text,
    "createdAt" = format(
      as.POSIXct(created_at, tz = "UTC"),
      "%Y-%m-%dT%H:%M:%OS6Z"
    ),
    langs = as.list(langs),
    tags = as.list(tags)
  )

  record$labels <- list(
    "$type" = "com.atproto.label.defs#selfLabels",
    values = lapply(labels, function(l) {
      list("$type" = "com.atproto.label.defs#selfLabel", val = l)
    })
  )

  if (!is.null(.reply)) {
    record[["reply"]] <- .reply
  } else if (!is.null(in_reply_to)) {
    in_reply_to <- ifelse(
      grepl("^http", in_reply_to),
      convert_http_to_at(in_reply_to, .token = .token),
      in_reply_to
    )

    thread <- do.call(
      app_bsky_feed_get_post_thread,
      list(in_reply_to, .token = .token)
    )
    thread_root <- get_thread_root(thread)
    record[["reply"]] <- list(
      root = list(
        "uri" = thread_root$post$uri,
        "cid" = thread_root$post$cid
      ),
      parent = list(
        "uri" = thread$thread$post$uri,
        "cid" = thread$thread$post$cid
      )
    )
  }

  if (!is.null(image) && !identical(image, "")) {
    image <- purrr::map_chr(image, from_ggplot)
    rlang::check_installed("magick")
    image_alt[utils::tail(length(image_alt):length(image), -1)] <- ""
    images <- purrr::map2(image, image_alt, function(i, alt) {
      ar <- magick::image_info(magick::image_read(i))
      blob <- com_atproto_repo_upload_blob2(i, .token = .token)
      list(
        alt = alt,
        image = blob[["blob"]],
        aspectRatio = list(height = ar$height, width = ar$width)
      )
    })
    record[["embed"]] <- list(
      "$type" = "app.bsky.embed.images",
      images = images
    )
  }

  if (!is.null(video) && !identical(video, "")) {
    if (!file.exists(video)) {
      # av can't deal with remote files
      vid_temp <- tempfile()
      httr2::request(video) |>
        httr2::req_perform(path = vid_temp)
      video <- vid_temp
    }
    rlang::check_installed("av")
    ar <- av::av_video_info(vid_temp)
    blob <- com_atproto_repo_upload_blob2(video, .token = .token)
    record[["embed"]] <- list(
      "$type" = "app.bsky.embed.video",
      video = blob$blob,
      aspectRatio = list(height = ar$video$height, width = ar$video$width)
    )
  }

  # link is only added when no image or video exist, but takes precedence over
  # links in text
  if (
    !is.null(link) &&
      !purrr::pluck_exists(record, "embed") &&
      isTRUE(preview_card)
  ) {
    record$embed <- fetch_preview(link)
    # record$embed$uri can't be empty, but the preview endpoint returns empty
    # uris sometimes. Fixing it here
    if (purrr::pluck(record, "embed", "external", "uri") == "") {
      purrr::pluck(record, "embed", "external", "uri") <- link
    }
  } else if (is.list(preview_card) && !purrr::pluck_exists(record, "embed")) {
    record$embed <- preview_card
  }

  if (!is.null(quote)) {
    quote <- ifelse(
      grepl("^http", quote),
      convert_http_to_at(quote, .token = .token),
      quote
    )

    quote_post <- do.call(app_bsky_feed_get_posts, list(quote, .token = .token))
    embed_record <- list(
      "$type" = "app.bsky.embed.record",
      "record" = list(
        "uri" = quote_post$posts[[1]]$uri,
        "cid" = quote_post$posts[[1]]$cid
      )
    )
    if (purrr::pluck_exists(record, "embed")) {
      record[["embed"]] <- list(
        "$type" = "app.bsky.embed.recordWithMedia",
        "record" = embed_record,
        "media" = record[["embed"]]
      )
    } else {
      record[["embed"]] <- embed_record
    }
  }

  # https://atproto.com/blog/create-post#mentions-and-links
  parsed_richtext <- parse_facets(text)
  if (!any(is.na(unlist(parsed_richtext)))) {
    record[["facets"]] <- parsed_richtext
    if (!purrr::pluck_exists(record, "embed") && isTRUE(preview_card)) {
      uri <- purrr::map_chr(parsed_richtext, function(f) {
        purrr::pluck(f, "features", 1, "uri", .default = NA_character_)
      }) |>
        stats::na.omit() |>
        utils::head(1L) # only one link can be previewed

      if (length(uri) > 0L) {
        # preview card
        record$embed <- fetch_preview(uri)
      }
    }
  }

  invisible(do.call(
    what = com_atproto_repo_create_record,
    args = list(repo, collection, record, .token = .token)
  ))
}


#' @rdname post
#' @export
post_skeet <- post


#' Fetch link preview
#'
#' @param uri URL or URI to fetch preview for.
#'
#' @returns list strutured for use as preview_card
#' @export
#'
#' @examples
#' \dontrun{
#' wiki_preview <- fetch_preview("https://en.wikipedia.org/wiki/AT_Protocol")
#' post_skeet("Do you know the AT Protocol?", preview_card = wiki_preview)
#' }
fetch_preview <- function(uri) {
  # this is the API bsky.app is using. Not sure how robust it is
  resp <- httr2::request("https://cardyb.bsky.app/v1/extract") |>
    httr2::req_url_query(url = uri) |>
    httr2::req_error(is_error = function(resp) FALSE) |>
    httr2::req_perform()

  if (httr2::resp_status(resp) < 400L) {
    preview <- resp |>
      httr2::resp_body_json()
    embed <- list(
      `$type` = "app.bsky.embed.external",
      external = list(
        uri = preview$url,
        title = preview$title,
        description = preview$description
      )
    )
    if (purrr::pluck_exists(preview, "image")) {
      embed$external$thumb <-
        com_atproto_repo_upload_blob2(purrr::pluck(preview, "image"))$blob
    }
  } else {
    embed <- list(
      `$type` = "app.bsky.embed.external",
      external = list(uri = uri, title = "", description = "")
    )
  }
  return(embed)
}


#' @rdname post
#' @export
delete_skeet <- function(post_url, verbose = NULL, .token = NULL) {
  id <- basename(post_url)
  if (verbosity(verbose)) {
    cli::cli_progress_step(
      msg = "Request to delete post {.emph {id}}",
      msg_done = "Deleted {.emph {id}}",
      msg_failed = "Something went wrong"
    )
  }

  invisible(purrr::map(post_url, function(u) {
    post_info <- ifelse(
      grepl("^http", u),
      convert_http_to_at(u, .token = .token),
      u
    ) |>
      parse_at_uri()

    do.call(
      what = com_atproto_repo_delete_record,
      args = list(
        repo = post_info$repo,
        collection = post_info$collection,
        rkey = post_info$rkey,
        .token = .token,
        .return = "json"
      )
    )
  }))
}


#' @rdname post
#' @export
delete_post <- delete_skeet


#' Post a thread
#'
#' @param texts a vector of skeet (post) texts
#' @param images paths to images to be included in each post. This may be a character vector, or a list of character vectors if multiple images per post are required.
#' @param image_alts alt texts for the images to be included in each post. If images is a list of character vectors, this should also be a list of character vectors and have the same shape.
#' @param thread_df instead of defining texts, images and image_alts, you can
#'   also create a data frame with the information in columns `text`, `image`, and `image_alt`.
#' @inheritParams search_user
#'
#' @return list of the URIs and CIDs of the posts (invisible)
#' @export
#'
#' @examples
#' \dontrun{
#' # post three messages in a thread
#' thread <- post_thread(c("Post 1", "Post 2", "Post 3"))
#'
#' # delete the thread
#' delete_post(thread$uri)
#' }
post_thread <- function(
  texts,
  images = NULL,
  image_alts = NULL,
  thread_df = NULL,
  verbose = NULL,
  .token = NULL
) {
  if (is.null(thread_df)) {
    images <- images %||% rep("", length(texts))
    image_alts <- image_alts %||% lapply(images, \(x) rep("", length(x)))
    if (length(unique(lengths(list(texts, images, image_alts)))) != 1L) {
      cli::cli_abort(
        "texts, images, image_alts must all have the same length or be NULL."
      )
    }

    thread_df <- tibble::tibble(
      text = texts,
      image = images,
      image_alt = image_alts
    )
  }

  root_resp <- NULL
  prev_resp <- NULL
  refs <- data.frame()

  for (i in seq_along(thread_df$text)) {
    reply_arg <- if (!is.null(root_resp)) {
      list(
        root = list(uri = root_resp$uri, cid = root_resp$cid),
        parent = list(uri = prev_resp$uri, cid = prev_resp$cid)
      )
    }

    resp <- do.call(
      what = post_skeet,
      args = list(
        text = thread_df$text[[i]],
        image = thread_df$image[[i]],
        image_alt = thread_df$image_alt[[i]],
        .reply = reply_arg,
        verbose = verbose,
        .token = .token
      )
    )

    if (is.null(root_resp)) {
      root_resp <- resp
    }
    prev_resp <- resp
    refs <- rbind(refs, as.data.frame(resp))
  }
  return(refs)
}


#' Search Posts
#'
#' @param q search query. See Details.
#' @param sort string. Specifies the ranking order of results. Possible values
#'   are "top" or "latest". Defaults to "latest".
#' @param since string. Filter results for posts after the specified datetime
#'   (inclusive). Can be a date or datetime object or a string that can be
#'   parsed either.
#' @param until string. Filter results for posts before the specified datetime
#'   (not inclusive). Can be a date or datetime object or a string that can be
#'   parsed either.
#' @param mentions string. Filter to posts that mention the given account. Only
#'   matches rich-text facet mentions.
#' @param author string. Filter to posts authored by the specified account.
#' @param lang string. Filter results to posts in the specified language.
#'   Language detection is expected to use the post's language field, though the
#'   server may override detection.
#' @param domain string. Filter results to posts containing URLs (links or
#'   embeds) pointing to the specified domain. Hostname normalization may apply.
#' @param url string. Filter results to posts containing links or embeds
#'   matching the specified URL. URL normalization or fuzzy matching may apply.
#' @param tag string. Filter results to posts containing the specified tag
#'   (hashtag). Do not include the hash (#) prefix. Multiple tags can be
#'   specified, with results matching all specified tags (logical AND).
#' @inheritParams search_user
#'
#' @details The [API
#'   docs](https://docs.bsky.app/docs/api/app-bsky-feed-search-posts) claim that
#'   Lucene query syntax is supported (Boolean operators and brackets for
#'   complex queries). But only a small subset is actually
#'   implemented:
#'
#'   - Whitespace is treated as implicit AND, so all words in a query must occur,
#'   but the word order and proximity are ignored.
#'   - Double quotes indicate exact phrases.
#'   - `from:<handle>` will filter to results from that account.
#'   - `-` excludes terms.
#'   - `OR` and parentheses for grouping (e.g. `(rstats OR rstat) bluesky`) are
#'   **not** supported and are treated literally (i.e. will only find posts
#'   with parentheses and the word OR in it).
#'
#'   Note that matches can occur anywhere in the skeet, not just the text. For
#'   example, a term can be in the link preview, or alt text of an image.
#'
#'
#' @returns a data frame (or nested list) of posts
#'
#' @examples
#' \dontrun{
#' search_post("rstats")
#' # finds post with the hashtag rstats AND the word Bluesky somewhere in the
#' # skeet (ignoring capitalisaion)
#' search_post("#rstats Bluesky")
#'
#' # search for the exact phrase "new #rstats package"
#' search_post("\"new #rstats package\"")
#' # Use single quotes so you do not need to escape double quotes
#' search_post('"new #rstats package"')
#'
#' # only search for skeets from one user
#' search_post("from:jbgruber.bsky.social #rstats")
#'
#' # narrow down the search with more parameters
#' search_post("{atrrr}",
#'             sort = "top",
#'             since = "2024-12-05",
#'             until = "2024-12-07 10:00:00",
#'             mentions = NULL,
#'             author = "jbgruber.bsky.social",
#'             domain = "jbgruber.github.io",
#'             url = "https://jbgruber.github.io/atrrr",
#'             tag = "rstats")
#' }
#' @export
search_post <- function(
  q,
  limit = 100L,
  sort = NULL,
  since = NULL,
  until = NULL,
  mentions = NULL,
  author = NULL,
  lang = NULL,
  domain = NULL,
  url = NULL,
  tag = NULL,
  parse = TRUE,
  verbose = NULL,
  .token = NULL
) {
  res <- list()
  req_limit <- ifelse(limit > 100, 100, limit)
  last_cursor <- NULL
  if (!is.null(since)) {
    since <- as_iso_date(since)
  }
  if (!is.null(until)) {
    until <- as_iso_date(until)
  }

  if (verbosity(verbose)) {
    cli::cli_progress_bar(
      format = "{cli::pb_spin} Got {length(res)} posts, but there is more.. [{cli::pb_elapsed}]",
      format_done = "Retrieved {length(res)} posts from {resp$hitsTotal} total hits. All done! [{cli::pb_elapsed}]"
    )
  }

  while (length(res) < limit) {
    resp <- do.call(
      what = app_bsky_feed_search_posts,
      args = list(
        q = q,
        limit = req_limit,
        cursor = last_cursor,
        sort = sort,
        since = since,
        until = until,
        mentions = mentions,
        author = author,
        lang = lang,
        domain = domain,
        url = url,
        tag = tag,
        .token = .token,
        .return = "json"
      )
    )

    if (is.null(resp$hitsTotal)) {
      resp$hitsTotal <- "an *unknown total* of"
    }
    if (is.null(last_cursor) && verbosity(verbose)) {
      cli::cli_alert_info("Found {resp$hitsTotal} posts that fit the query")
    }

    last_cursor <- resp$cursor
    res <- c(res, resp$posts)

    if (is.null(resp$cursor)) {
      break
    }
    if (verbosity(verbose)) cli::cli_progress_update(force = TRUE)
  }

  if (verbosity(verbose)) {
    cli::cli_progress_done()
  }

  if (parse) {
    if (verbosity(verbose)) {
      cli::cli_progress_step("Parsing {length(res)} results.")
    }

    out <- parse_post_list(res)

    if (verbosity(verbose)) {
      cli::cli_process_done(msg_done = "Got {nrow(out)} results. All done!")
    }
  } else {
    out <- res
  }
  attr(out, "last_cursor") <- last_cursor
  return(out)
}


#' @rdname search_post
#' @export
search_skeet <- search_post


#' Like a skeet
#'
#' @inheritParams post
#'
#' @returns invisible record information from the API
#' @export
#'
#' @examples
#' \dontrun{
#' # like a post
#' like_skeet("https://bsky.app/profile/jbgruber.bsky.social/post/3lcmymlgxwa2t")
#'
#' # or feed in the result of some search
#' johannes_posts <- get_skeets_authored_by("jbgruber.bsky.social")
#' like_skeet(johannes_posts$uri)
#' }
like_skeet <- function(post_url, verbose = NULL, .token = NULL) {
  id <- basename(post_url)
  if (verbosity(verbose)) {
    cli::cli_progress_step(
      msg = "Request to like post {.emph {id}}",
      msg_done = "Liked {.emph {id}}",
      msg_failed = "Something went wrong"
    )
  }

  invisible(purrr::map(post_url, function(u) {
    uri <- convert_http_to_at(u, .token = .token)
    post_info <- do.call(
      app_bsky_feed_get_posts,
      args = list(uris = uri, .token = .token)
    ) |>
      purrr::pluck("posts", 1L)

    do.call(
      what = com_atproto_repo_create_record,
      args = list(
        repo = get_token()$did,
        collection = "app.bsky.feed.like",
        record = list(
          subject = list(
            uri = post_info$uri,
            cid = post_info$cid
          ),
          createdAt = as_iso_date(Sys.time()),
          `$type` = "app.bsky.feed.like"
        ),
        .token = .token,
        .return = "json"
      )
    )
  }))
}


#' @rdname like_skeet
#' @export
like_post <- like_skeet