# Automated congratulatory tweet to Twitter Friends

On the front page of my premiersoccerstats site, I have a Player Milestones table which highlights players who have reached certain levels in the Premier League’s latest round of games e.g. 100 Appearances

This requires comparing two datasets and subsetting the rows with differences in the variable of interest. To this end, I use the daff package which was the subject of a presentation at the recent R User 2017 conference

Coincidentally, an alternative package dataCompareR has just been released by Rob Noble-Eddy and others

As an aside I can’t help but seeing Rob leading a band called Eddy and the Rob-Nobs performing something like this

OK. Fast forward 40 years

One project I have had on the back-burner (yep it’s been there since this phrase was fashionable) is to send a congratulatory tweet whenever one of the tweeters I follow reaches a certain, significant, number of followers e.g 1000

Achieving this can be broken down into several stages. The packages I will rely on (other than the assumed tidyverse) are also listed

1. Acquire my ‘friends’ (rtweet)
2. Find out how many followers my friends have at a point in time, df.a (rtweet)
3. Find out how many followers they have at a later point in time,df.b (rtweet)
4. Compare df.a with df.b and filter to important milestones (dataCompareR)
5. Craft and issue relevant tweet messages (glue, rtweet, purrr)
6. Automate process on say, daily, basis

I have covered rtweet and purrr in previous posts e.g ‘Twitter Followers Collage’ but I am using different functions in those packages. This is my first experience of using dataCompareR and glue


library(tidyverse)
library(plotly)

library(rtweet)
library(dataCompareR)
library(glue)

N.B. If you want to try this for yourself, you will need to obtain an OAuth token and preferably store it for future use. Instructions here

First step is to use an rtweet function to garner the ids of those I am following. This takes the form of a data_set with a series of user_id’s
It changes rarely (and probably could do with some fine-tuning) but the data only takes a few seconds to return so there is no need to save the results in a local file unless you have significantly more than my, at the time of writing, total of 166.

I then need to get a summary of their data (which includes the number of followers they each have) This takes around 20 seconds - so reprise the video whilst waiting - and uses the map_df(), purrr function and lookup_users() from rtweet

I have added a plot for interest. Hover for further information


# obtain a list of accounts I follow
friends <- get_friends("pssguy", page = "-1", parse = TRUE, token = NULL) #default parameters

# now create a function which gets the summary data
friendsSummary <- map_df(friends$user_id, function(x) lookup_users(x)) names(friendsSummary) ## [1] "user_id" "name" ## [3] "screen_name" "location" ## [5] "description" "url" ## [7] "protected" "followers_count" ## [9] "friends_count" "listed_count" ## [11] "statuses_count" "favourites_count" ## [13] "account_created_at" "verified" ## [15] "profile_url" "profile_expanded_url" ## [17] "account_lang" "profile_banner_url" ## [19] "profile_background_url" "profile_image_url" hadley <- friendsSummary %>% filter(screen_name=="hadleywickham") friendsSummary %>% plot_ly() %>% add_markers(x=~log10(friends_count), y=~log10(followers_count), hoverinfo="text", name="All Following", text=~paste0("Name:", name, "<br>Followers:",followers_count, "<br>Following:",friends_count) ) %>% add_markers(data=hadley,x=~log10(friends_count), y=~log10(followers_count), color=I("red"), hoverinfo="text",name="Hadley Wickham", text=~paste0("Name:", name, "<br>Followers:",followers_count, "<br>Following:",friends_count) ) %>% config(displayModeBar = F,showLink = F) I have a quite the range, with the people I am following having a median of just under 2000 followers, a figure I never expect to attain Hadley Wickham is the ‘biggest name’ I follow in the R community . Those with more adherents tend to be organizations So this friendsSummary data.frame can be saved for comparison at a later date. For the purposes of this tutorial, I will create a duplicate set and add a nominal 100 to each followers_count. Firstly, however I will add a couple of fields to categorize followers into specific ranges which make sense from the data  friendsSummary <- friendsSummary %>% mutate(category=case_when( followers_count > 9999 ~ 9, followers_count > 4999 ~ 8, followers_count > 2499 ~ 7, followers_count > 999 ~ 6, followers_count > 499 ~ 5, followers_count > 249 ~ 4, followers_count > 99 ~ 3, followers_count > 49 ~ 2, followers_count > 0 ~ 1) ) %>% mutate(minFollowers=case_when( followers_count > 9999 ~ 10000, followers_count > 4999 ~ 5000, followers_count > 2499 ~ 2500, followers_count > 999 ~ 1000, followers_count > 499 ~ 500, followers_count > 249 ~ 250, followers_count > 99 ~ 100, followers_count > 49 ~ 50, followers_count > 0 ~ 1) ) # save for subsequent retrieval, typically a day later in a script write_csv(friendsSummary,"data/friendsSummary.csv") # this will be our base data.frame df.a <- read_csv("data/friendsSummary.csv") # we now create a duplicate file, add 100 followers to each row and recategorize df.b <- df.a df.b[,"followers_count"] <- df.b[,"followers_count"] + 100L # adding L ensures the firls remains an integer df.b <- df.b %>% mutate(category=case_when( followers_count > 9999 ~ 9, followers_count > 4999 ~ 8, followers_count > 2499 ~ 7, followers_count > 999 ~ 6, followers_count > 499 ~ 5, followers_count > 249 ~ 4, followers_count > 99 ~ 3, followers_count > 49 ~ 2, followers_count > 0 ~ 1) ) %>% mutate(minFollowers=case_when( followers_count > 9999 ~ 10000, followers_count > 4999 ~ 5000, followers_count > 2499 ~ 2500, followers_count > 999 ~ 1000, followers_count > 499 ~ 500, followers_count > 249 ~ 250, followers_count > 99 ~ 100, followers_count > 49 ~ 50, followers_count > 0 ~ 1) ) # ensures comparison field are the same class i.e. integer df.b$category <- as.integer(df.b$category) OK, so now we can use the dataCompareR function. This results in a list including the desired data.frame (compFriends$mismatches$CATEGORY) showing changes, if any, in category We are only interested in the rows where the category has increased - letting people know they have less followers is not likely to win you friends  # Put newer data.frame first and add unique key you want to retain compFriends <- rCompare(df.b, df.a, keys = 'screen_name') # create a data.frame by joining to original data change_df <-compFriends$mismatches$CATEGORY %>% filter(diffAB>0) %>% # ensure only increases left_join(df.b,by=c("SCREEN_NAME"="screen_name")) change_df %>% select(SCREEN_NAME,valueA,valueB,diffAB) %>% DT::datatable(class='compact stripe hover row-border order-column',rownames=FALSE,options= list(paging = TRUE, searching = FALSE,info=FALSE)) So ,in this example, 23 have changed category. In real life, with the numbers I am following a daily change would be typically zero or one. Obviously, if you want to do a similar exercise it will depend where you set the catgories, how many followers you have and how often you run the code Now we get to the fun part of creating and issuing a tweet. I will use the glue package (as an alternative to paste) for the former and purrr and rtweet for the latter  ## add a tweet message change_df <-change_df %>% mutate(tweet=glue("Congrats @{SCREEN_NAME} you now have {minFollowers}+ Followers")) # adding media can be fun. token required if not saved tweetFunc <- function(x) post_tweet(x, media = "img/congrats.gif", token = NULL) # This will actually send tweets so I have commented it out #walk(change_df$tweet, tweetFunc)

You are notified that your tweet has been posted and it appears in your timeline with the clickable twitter handle of the user

Here is an example

library(blogdown) #v 0.0.54
shortcode("tweet", "889469685279752193")

rtweet, also offers a similar function to direct message, post_message(), which you can use for personal messages to people who follow you

All the above code can be adapted, as required, and collated into an R script which can be run routinely. How to do this will vary by operating system is outside the compass of this post. I plan on setting one up so, if I follow you, look out for your next milestone tweet!