Introduction to {pocketapi}

Create a Pocket Application

You need to create a Pocket application in Pocket’s developer portal to access your Pocket data. Don’t worry: this app will only be visible to you and only serves the purpose of acquiring the credentials for pocketapi.

1. Log in to your Pocket account and go to https://getpocket.com/developer/apps/new.
2. Click “Create New Application”. You’ll be presented with a form. Choose a sensible application name and a description.
3. Give your app permissions: Depending on the permissions of your app, you’ll be able to use all or just a subset of the pocketapi functions:

4. Check any platform for your app - this does not really matter but you need to check at least one box.
5. Accept the terms of service and click “Create Application”.

Get consumer key and token

Pocketapi uses the OAuth2 flow provided by the Pocket Authentication API to get an access token for your App. Because Pocket does not closely follow the OAuth standard, we are not able to provide as smooth an experience as other packages do (e.g. googlesheets4). Instead, the user has to follow the following instructions once to obtain an access token:

1. Request a request token: req_token <- get_request_token(consumer_key)

2. Authorize your app by entering the URL created by create_authorize_url in your browser:

create_authorize_url(req_token)

This step is critical: Even if you have authorized your app before and you want to get a new access token, you need to do the authorization in your browser again. Otherwise, the request token will not be authorized to generate an access token!

3. Get access token using the now authorized request token:

access_token <- get_access_token(consumer_key, request_token)

Important: Never make your consumer_key and access_token publicly available – or anyone will be able to access your Pocket!

Add the consumer key and access token to your environment

It is common practice to set API keys in your R environment file so that every time you start R the key is loaded.

All pocketapi functions access your consumer_key and access_token automatically by executing Sys.getenv("POCKET_CONSUMER_KEY") respectively Sys.getenv("POCKET_ACCESS_TOKEN"). Alternatively, you can provide an explicit definition of your consumer_key and access_token with each function call.

In order to add your key to your environment file, you can use the function edit_r_environ from the usethis package:

usethis::edit_r_environ()

This will open your .Renviron file in the RStudio editor. Now, you can add the following lines to it:

POCKET_CONSUMER_KEY="yourkeygoeshere"
POCKET_ACCESS_TOKEN="youraccesstokengoeshere"

Save the file and restart R for the changes to take effect.

If your .Renviron lives at a non-conventional place, you can also edit it manually using RStudio or your favorite text editor.

If you don’t want to clutter your .Renviron file, you can also use an .env file in your project directory together with the dotenv package. In this case, make sure to never share your .env file.

Input for the pocket_get() function

The main arguments of pocket_get() are also explained in the function’s help file, but below you find a quick overview of the main function arguments for adjusting which content is retrieved from Pocket.

• favorite allows to filter on favorited items (favorite = TRUE) or on unfavorited items (favorite = FALSE). Using the default value (favorite = NULL) means that items are retrieved regardless if they are favorited or not.
• item_type filters based on the format of the Pocket content, allowing to filter on “article”, “image” or “video”. This format classification is done automatically by Pocket and cannot be modified (and there are some instances of misclassification). Keeping the default (item_type = NULL) retrieves all items regardless of type.
• tag selects only the items that have a certain tag, also making it possible to filter by any untagged items. At the moment, only one tag can be used for filtering at the same time.
• state can select items depending on their reading status. It’s possible to retrieve only items that have not yet been read and are still in the Pocket queue (state = "unread") or to filter on items that are already read and archived (state = "archive"). By default, the function returns all items (state = "all"), regardless whether they are read or unread. Depending on your goals, you might prefer another option besides this default.

The function arguments “consumer_key” and “access_token” are the mandatory account credentials for the API. Please see the previous section for an explanation how to set up your account.

By default, the Pocket API has a limit of 5000 items per API call. This means that the output of pocket_get() is a data frame with up to 5000 rows, even if there might actually be more items in your Pocket. If you have more items saved in Pocket, you may use some tricks to combine multiple API calls to get a more complete picture (e.g., using the “state” parameter or playing around with favorited/tagged items), but it may not always be possible to retrieve all items if you have an extremely high number of items saved in your Pocket account.

Like many other APIs, Pocket allows only a certain number of API calls per hour (find details here), but usually this limit should not be an issue for pocket_get(). Note that all API calls, including adding or modifying Pocket items, count towards this hourly limit.

Output of the pocket_get() function

pocket_get() returns a data frame where each row is an item that has been saved in Pocket. For each item, the following information is saved:

• item_id and resolved_id: internal Pocket item identifiers. If you want to modify an item (e.g. via pocket_archive() or pocket_delete()), you need to specify the item_id as an input for this function.
• given_url: the web address of the item that has been saved.
• given_title and resolved_title: the title of the item. “given_title” is the title that was saved along with the item (but can be blank), whereas “resolved_title” is coming from Pocket’s attempt to parse it from the website (which is often more complete).
• favorite: logical variable, indicating TRUE if an item is favorited.
• status: numeric variable with three possible values: 0 if the item is unread, 1 if the item is archived, 2 if the item is about to be deleted.
• is_article: logical variable, TRUE if Pocket has detected the item to be an article (and not an image or a video). The classification is provided by Pocket and not always 100 percent precise.
• has_image and has_video: numeric variable, with three possible values: 0 if the item does not contain any image/video, 1 if the item contains an image/video, 2 if the item is an image/video.
• word_count: numeric variable, indicating the number of words in the item.
• tags: the user-defined tags associated with each item. If an item has several tags, they are separated via comma (e.g. “tag 1,tag 2”).
• authors: information about the author of the item (generated by Pocket via parsing the website), consisting for each item of a list with four elements: “item_id” (should be identical to the information from the variable with the same name mentioned above), “author_id” (an identifier created by Pocket), “name” (the author’s name in plain text) and “url” (the web address of the author at the website, usually showing all posts/articles written by the same author).
• images and image: image contains some identifier for the image. images is a list with seven elements: “item_id” (should be identical to the information from the variable with the same name mentioned above), “image_id”, “src” (the images’ web address), “width” and “height” (supposedly showing the image’s dimensions, but often mistakenly displaying “0” as a value even if the picture was larger than 0x0 pixels), “credit” (the person/organization credited for the picture, parsed by Pocket) and “caption” (the picture’s caption as parsed by Pocket).

If you need a more technical description of the data that is returned by the Pocket API, the overview in the developer documentation might help.

Applied example: obtaining data via pocket_get() and using it

The following applied example illustrates how to extract data from Pocket via pocket_get() and use it thereafter in R. Please note that the example assumes that you have already connected a new Pocket application in your developer settings, created the necessary credentials and saved them in your .renviron file. All these steps are explained in the previous sections of this vignette.

In our first code snippet, we call pocket_get(), keeping all options to the defaults, and assign its result to a new object pocket_items. Technically, state = "all" is also the default, but we specify it to make it easier to understand that we get data for both read and unread items.

library(pocketapi)

pocket_items <- pocket_get(state = "all")

The resulting object pocket_items is now part of our R environment and can be used just like any standard data frame from other sources. It’s a data.frame / tibble, with the latter allowing it to be integrated nicely into workflows based on the tidyverse packages ^[For more information about the tibble class, see here. We can also see that we have 53 rows in our data frame, one for each item saved in Pocket. Tabulating the “status” variable shows us the reading status: There are 20 items which are unread (status == 0) and 33 items which that are archived (status == 1). (Note: The data is coming from an account that has been set up specifically for the development of this package. It’s very likely that “real” users have many more items in their Pocket accounts.)

class(pocket_items)
#> [1] "tbl_df"     "tbl"        "data.frame"

nrow(pocket_items)
#> [1] 53

table(pocket_items$status) # reading status
#> 
#>  0  1 
#> 20 33

The data can be used nicely as input for other R functions. In the code example below, we first show how the data is used as input for tidyverse packages: We use the dplyr package to calculate the mean and the standard deviation of the number of words per item, separately for unread and archived items. Then, we demonstrate how the data is used in a base R function - in this case, a t-test that allows us to see that there is no significant difference in the length of unread vs. archived articles.

library(dplyr)

# quick re-code: new "reading_status" variable with labels that are intuitive to understand
pocket_items$reading_status <- recode(pocket_items$status, `0` = "unread", `1` = "archived")

# using dplyr functions on the data: grouping the data and then summarizing it
pocket_items %>%
  group_by(reading_status) %>%
  summarise(word_count_avg = mean(word_count),
            word_count_sd = sd(word_count),
            base_group = n())
#> # A tibble: 2 x 4
#>   reading_status word_count_avg word_count_sd base_group
#>   <chr>                   <dbl>         <dbl>      <int>
#> 1 archived                 967.          626.         33
#> 2 unread                  1167.          834.         20

# using base-R functions on the data: t-test
t.test(word_count ~ reading_status, data = pocket_items)
#> 
#>  Welch Two Sample t-test
#> 
#> data:  word_count by reading_status
#> t = -0.92555, df = 31.96, p-value = 0.3616
#> alternative hypothesis: true difference in means is not equal to 0
#> 95 percent confidence interval:
#>  -639.9889  240.1071
#> sample estimates:
#> mean in group archived   mean in group unread 
#>               966.9091              1166.8500

Of course, the data can also serve as input for plotting functions. Below, we use the ggplot2 package to show the distribution of the number of words per item for both unread and archived articles. As already shown in the t-test, the mean for both groups is very similar and the distributions largely overlap, but it seems that the distribution of the unread articles has slightly wider tails.

library(ggplot2)

ggplot(data = pocket_items) + 
  geom_density(aes(x = word_count, group = reading_status, fill = reading_status), alpha = 0.25) +
  labs(title = "Article Length of Read / Unread Items in Pocket",
       x = "Word Count",
       y = "Density", 
       fill = "Reading Status",
       caption = "Data: Articles from an example account.
       \nn=20 unread and n=33 read/archived articles") + 
  theme_minimal() + 
  theme(legend.position = "top")