|
|
Analysis Tech Art Repositories | |||||||
Author: Brad Cable
Illinois State University (IT497)
Are there more GitHub R repositories created in February than January?
These are the various libraries needed when executing this code.
library(ggplot2)
library(RCurl)
library(rjson)
library(stringr)
These are the variables that you'll need to put your GitHub API client ID and client secret into.
client.id <- "GITHUB_ID_HERE"
client.secret <- "GITHUB_SECRET_HERE"
This function defines a generic function for using the GitHub API. It will take in a search type, query, and page number (optional). The return is the decoded JSON as returned by the rjson::fromJSON() function.
search_type - One of “repositories”, “code”, “issues”, or “users”.
q - Query as defined on: https://help.github.com/articles/search-syntax/
page - Optional page number for query results as defined here: https://developer.github.com/v3/#pagination
gh_search <- function(search_type, q, page=1){
# this prevents issues with rate limiting
# (GitHub has 30 requests per minute limit)
Sys.sleep(5)
# now perform query
fromJSON(getURL(paste0(
"https://api.github.com/search/",
search_type,
"?client_id=", client.id,
"&client_secret=", client.secret,
"&q=", curlEscape(q),
"&page=", page
), httpheader=c("User-Agent"= "BCable")))
}
This function defines a search for a total number of results by language, year, and month. Returns the total number as a single number if only one month is provided or as a vector of results if multiple months are provided.
language - The language name to search for.
year - The year to search for.
month - The month or vector of months to search. If multiple months are provided, they will be looped through and returned as a vector accordingly.
gh_lang_date <- function(language, year, month){
ret <- NULL
# loop given months
for(begin_month in month){
end_month <- as.integer(begin_month)+1
end_date <- 1
# adjust end month if we are dealing with December
if(end_month > 12){
end_month <- 12
end_date <- 31
}
# pad values for string based search result
begin_month <- str_pad(begin_month, 2, "left", "0")
end_month <- str_pad(end_month, 2, "left", "0")
end_date <- str_pad(end_date, 2, "left", "0")
# construct search query and conduct search
result <- gh_search("repositories", paste0(
"language:", language,
' created:"',
year, "-", begin_month, "-01 .. ",
year, "-", end_month, "-", end_date,
'"'
))$total_count
# append result to return value
ret <- c(ret, result)
}
# return results
ret
}
This command searchs for the R programming language for 2014 between January and December.
data_2014 <- gh_lang_date("R", 2014, 1:12)
This command searchs for the R programming language for 2015 between January and December.
data_2015 <- gh_lang_date("R", 2015, 1:12)
This command searchs for the R programming language for 2016 between January and April. At time of writing, April 2016 is the last full month.
data_2016 <- gh_lang_date("R", 2016, 1:4)
This command produces a data frame that combines the counts together, then generates a set of POSIXlt dates associated with those counts.
final_data <- data.frame(
Count=c(data_2014, data_2015, data_2016),
Date=as.POSIXlt(paste0(
c(
rep(2014, length(data_2014)),
rep(2015, length(data_2015)),
rep(2016, length(data_2016))
), "-", str_pad(c(
seq(1, length(data_2014)),
seq(1, length(data_2015)),
seq(1, length(data_2016))
), 2, "left", "0"), "-01"
), format="%Y-%m-%d")
)
This command sorts the data by date.
final_data <- final_data[order(as.numeric(final_data$Date)),]
This command extracts the full month name from the POSIXlt date and converts it to a factor.
final_data$Month <- strftime(final_data$Date, format="%B")
final_data$Month <- factor(final_data$Month, levels=unique(final_data$Month))
This command extracts the year from the POSIXlt date and puts it in its own column, and converts it to a factor.
final_data$Year <- strftime(final_data$Date, format="%Y")
final_data$Year <- factor(final_data$Year, levels=unique(final_data$Year))
This command strips the POSIXlt date since we no longer need it since all we need is month and year. The only point of converting to POSIXlt in the first place was to get the full month names for the month field.
final_data$Date <- NULL
class(final_data)
## [1] "data.frame"
str(final_data)
## 'data.frame': 26 obs. of 3 variables:
## $ Count: num 1967 922 6206 3265 3580 ...
## $ Month: Factor w/ 11 levels "January","February",..: 1 2 3 4 5 6 7 8 9 10 ...
## $ Year : Factor w/ 3 levels "2014","2015",..: 1 1 1 1 1 1 1 1 1 1 ...
summary(final_data)
## Count Month Year
## Min. : 922 January : 3 2014:11
## 1st Qu.:4291 February: 3 2015:11
## Median :5248 March : 3 2016: 4
## Mean :4790 April : 3
## 3rd Qu.:5577 May : 2
## Max. :6549 June : 2
## (Other) :10
The data contains three columns, one for the year, one for the month, and one for the count of R repositories created in the associated year and month fields.
final_data
## Count Month Year
## 1 1967 January 2014
## 2 922 February 2014
## 3 6206 March 2014
## 4 3265 April 2014
## 5 3580 May 2014
## 6 4153 June 2014
## 7 4245 July 2014
## 8 4430 August 2014
## 9 4526 September 2014
## 10 4450 October 2014
## 11 3986 November 2014
## 12 5824 January 2015
## 13 5525 February 2015
## 14 6549 March 2015
## 15 5581 April 2015
## 16 5305 May 2015
## 17 5708 June 2015
## 18 5490 July 2015
## 19 5366 August 2015
## 20 5190 September 2015
## 21 4915 October 2015
## 22 4754 November 2015
## 23 5566 January 2016
## 24 5823 February 2016
## 25 5778 March 2016
## 26 5447 April 2016
Assuming 2016 is the year being compared, there were 5566 R repositories created in January and 5823 R repositories created in February, so there were more R repositories created in February 2016.
g <- ggplot(final_data, aes(x=Month, y=Count, group=Year, color=Year))
g + geom_line() + geom_point()
g <- ggplot(final_data, aes(x=Month, y=Count, group=Year, fill=Year))
g + geom_bar(stat="identity", position="dodge")
