1. Scala's real-world project repository data
With almost 30k commits and a history spanning over ten years, Scala is a mature programming language. It is a general-purpose programming language that has recently become another prominent language for data scientists.
Scala is also an open source project. Open source projects have the advantage that their entire development histories -- who made changes, what was changed, code reviews, etc. -- are publicly available.
We're going to read in, clean up, and visualize the real world project repository of Scala that spans data from a version control system (Git) as well as a project hosting site (GitHub). We will find out who has had the most influence on its development and who are the experts.
The dataset we will use, which has been previously mined and extracted from GitHub, is comprised of three files:
pulls_2011-2013.csv
contains the basic information about the pull requests, and spans from the end of 2011 up to (but not including) 2014.pulls_2014-2018.csv
contains identical information, and spans from 2014 up to 2018.pull_files.csv
contains the files that were modified by each pull request.
# Importing pandas
import pandas as pd
# Loading in the data
pulls_one = pd.read_csv('datasets/pulls_2011-2013.csv')
pulls_two = pd.read_csv('datasets/pulls_2014-2018.csv')
pull_files = pd.read_csv('datasets/pull_files.csv')
# view the first 5 rows of the pulls_one dataframe
pulls_one.head()
# view the first 5 rows of the pulls_two dataframe
pulls_two.head()
# view the first 5 rows of the pulls dataframe
pull_files.head()
2. Preparing and cleaning the data
First, we will need to combine the data from the two separate pull DataFrames.
Next, the raw data extracted from GitHub contains dates in the ISO8601 format. However, pandas
imports them as regular strings. To make our analysis easier, we need to convert the strings into Python's DateTime
objects. DateTime
objects have the important property that they can be compared and sorted.
The pull request times are all in UTC (also known as Coordinated Universal Time). The commit times, however, are in the local time of the author with time zone information (number of hours difference from UTC). To make comparisons easy, we should convert all times to UTC.
# Append pulls_one to pulls_two
pulls = pd.concat([pulls_one,pulls_two])
# Convert the date for the pulls object
pulls['date'] = pd.to_datetime(pulls['date'],utc=True)
# view the first rows of the new pulls dataframe with datetime column
pulls.head()
3. Merging the DataFrames
The data extracted comes in two separate files. Merging the two DataFrames will make it easier for us to analyze the data in the future tasks.
# Merge the two DataFrames
data = pd.merge(pulls,pull_files,on='pid')
# view the first rows of the merged data
data.head()
4. Is the project still actively maintained?
The activity in an open source project is not very consistent. Some projects might be active for many years after the initial release, while others can slowly taper out into oblivion. Before committing to contributing to a project, it is important to understand the state of the project. Is development going steadily, or is there a drop? Has the project been abandoned altogether?
The data used in this project was collected in January of 2018. We are interested in the evolution of the number of contributions up to that date.
For Scala, we will do this by plotting a chart of the project's activity. We will calculate the number of pull requests submitted each (calendar) month during the project's lifetime. We will then plot these numbers to see the trend of contributions.
A helpful reminder of how to access various components of a date can be found in this exercise of Data Manipulation with pandas
Additionally, recall that you can group by multiple variables by passing a list to
.groupby()
. This video from Data Manipulation with pandas should help!
%matplotlib inline
# Create a column that will store the month
data['month'] = data['date'].dt.month
# Create a column that will store the year
data['year'] = data['date'].dt.year
# Group by the month and year and count the pull requests
counts = data.groupby(['year','month'])['pid'].count()
# Plot the results
counts.plot(kind='bar', figsize = (12,4))
5. Is there camaraderie in the project?
The organizational structure varies from one project to another, and it can influence your success as a contributor. A project that has a very small community might not be the best one to start working on. The small community might indicate a high barrier of entry. This can be caused by several factors, including a community that is reluctant to accept pull requests from "outsiders," that the code base is hard to work with, etc. However, a large community can serve as an indicator that the project is regularly accepting pull requests from new contributors. Such a project would be a good place to start.
In order to evaluate the dynamics of the community, we will plot a histogram of the number of pull requests submitted by each user. A distribution that shows that there are few people that only contribute a small number of pull requests can be used as in indicator that the project is not welcoming of new contributors.
# Required for matplotlib
%matplotlib inline
# Group by the submitter
by_user = data.groupby('user').agg({'pid': 'count'})
# Plot the histogram
by_user.hist()
6. What files were changed in the last ten pull requests?
Choosing the right place to make a contribution is as important as choosing the project to contribute to. Some parts of the code might be stable, some might be dead. Contributing there might not have the most impact. Therefore it is important to understand the parts of the system that have been recently changed. This allows us to pinpoint the "hot" areas of the code where most of the activity is happening. Focusing on those parts might not the most effective use of our times.
# Identify the last 10 pull requests
last_10 = pulls.nlargest(10,'date')
# Join the two data sets
joined_pr = last_10.merge(pull_files,on='pid')
# Identify the unique files
files = set(joined_pr['file'])
# Print the results
files