GithubDataScraper/GithubDataScraper.py

# To add a new cell, type '# %%'
# To add a new markdown cell, type '# %% [markdown]'
# %%
import requests
from requests_oauthlib import OAuth2Session
import os
import json
import pandas as pd
import time
from datetime import datetime


# %%
MAXTIME = 365
RESULT_FOLDER = "results"
# These two informations are needed so the OAuth-Session can be established
CLIENTID = ""
TOKEN = ""

FILE = "decentral_communication_protocols.csv"


# %%
def getGithubOAuthSession():
    token = {
        'access_token': TOKEN
    }
    return OAuth2Session(CLIENTID, token=token)    


# %%
def createCSV(pathToFile, list):
    df = pd.DataFrame(list)
    df.to_csv(pathToFile)


# %%
def getCommits(repoName, oAuthSession):
    commitList = []
    devList = []
    headers = {'Accept': 'application/vnd.github.mercy-preview+json'}
    commitsResponseJson = None
    i = 0
    commitsTooOld = False
    while commitsResponseJson is None or len(commitsResponseJson) > 0:
        commitsResponse = oAuthSession.get('https://api.github.com/repos/' + repoName + '/commits?page=' + str(i), headers=headers)
        if commitsResponse.status_code == 200:
            commitsResponseJson = commitsResponse.json()
            i = i + 1
            for commit in commitsResponseJson:
                if MAXTIME is not None and (datetime.now() - datetime.strptime(commit['commit']['author']['date'], '%Y-%m-%dT%H:%M:%SZ')).days > MAXTIME:
                    commitsTooOld = True
                    break
                committer = commit['committer']
                if committer is not None and "login" in committer.keys() and committer['login'] not in devList:
                    devList.append(committer['login'])
                commitList.append(commit)
            if commitsTooOld:
                break
        elif commitsResponse.status_code == 409: # empty repo
            break
        else:
            time.sleep(3600)
            print("Error in Http Request (commitsRequest):", commitsResponse.status_code, commitsResponse.text)
    return commitList, devList


# %%
def getIssues(repoName, oAuthSession):
    issueList = []
    devList = []
    headers = {'Accept': 'application/vnd.github.mercy-preview+json'}
    issuesResponseJson = None
    i = 0
    issuesTooOld = False
    while issuesResponseJson is None or len(issuesResponseJson) > 0:
        issuesResponse = oAuthSession.get('https://api.github.com/repos/' + repoName + '/issues?state=all&page=' + str(i), headers=headers)
        if issuesResponse.status_code == 200:
            i = i + 1
            issuesResponseJson = issuesResponse.json()
            for issue in issuesResponseJson:
                if MAXTIME is not None and (datetime.now() - datetime.strptime(issue['created_at'], '%Y-%m-%dT%H:%M:%SZ')).days > MAXTIME:
                    issuesTooOld = True
                    break
                assignee = issue['assignee']
                assignees = issue['assignees']
                if assignee is not None and 'login' in assignee.keys() and assignee['login'] not in devList:
                    devList.append(assignee['login'])
                for assignee in assignees:
                    if 'login' in assignee.keys() and assignee['login'] not in devList:
                        devList.append(assignee['login'])
                issueList.append(issue)
            if issuesTooOld:
                break
        elif issuesResponse.status_code == 409: # empty repo
            break
        else:
            print("Error in Http Request (issuesRequest):", issuesResponse.status_code, issuesResponse.text)
            time.sleep(3600)
    return issueList, devList


# %%
def getRepoFacts(repoName, oAuthSession):
    devList = []
    issueList = []
    commitList = []
    issueList, devListIssues = getIssues(repoName, oAuthSession)
    commitList, devListCommits = getCommits(repoName, oAuthSession)
    devList.append(devListIssues)
    for dev in devListCommits:
        if dev not in devList:
            devList.append(dev)
    return issueList, commitList, devList


# %%
def getTopicFacts(topicName):
    amountRepos = 0
    amountStars = 0
    amountIssues = 0
    amountCommits = 0
    amountDevs = 0

    repoNameList = []
    repoList = []
    issueList = []
    commitList = []
    devList = []

    oAuthSession = getGithubOAuthSession()
    headers = {'Accept': 'application/vnd.github.mercy-preview+json'}
    topicResponse = oAuthSession.get('https://api.github.com/search/repositories?q=topic:' + topicName, headers=headers)

    if topicResponse.status_code == 200:
        topicResponseJson = topicResponse.json()
        i = 0
        while "items" in topicResponseJson.keys() and len(topicResponseJson['items']) > 0:
            i = i + 1
            for repo in topicResponseJson['items']:
                print(repo['full_name'])
                if repo['full_name'] not in repoNameList:
                    issueListRepo, commitListRepo, devListRepo = getRepoFacts(repo['full_name'], oAuthSession)
                    for issue in issueListRepo:
                        issueList.append(issue)
                    for commit in commitListRepo:
                        commitList.append(commit)
                    for dev in devListRepo:
                        if dev not in devList:
                            devList.append(dev)
                    amountStars = amountStars + int(repo['stargazers_count'])
                    repoNameList.append(repo['full_name'])
                    repoList.append(repo)
            topicResponseJson = oAuthSession.get('https://api.github.com/search/repositories?q=topic:' + topicName + '&page=' + str(i), headers=headers).json()
        
        createCSV(topicName + "_repos.csv", repoList)
        createCSV(topicName + "_issues.csv", issueList)
        createCSV(topicName + "_commits.csv", commitList)

        amountRepos = len(repoList)
        amountDevs = len(devList)
        amountIssues = len(issueList)
        amountCommits = len(commitList)
        print("Amount Repos:", str(amountRepos))
        print("Amount Stars:", str(amountStars))
        print("Amount Issues:", str(amountIssues))
        print("Amount Commits:", str(amountCommits))
        print("Amount Devs:", str(amountDevs))
    else:
        print("Error in Http Request (topicRequest):", topicResponse.status_code, topicResponse.text)
    return amountRepos, amountStars, amountIssues, amountCommits, amountDevs


# %%
def main():
    amountReposList = []
    amountStarsList = []
    amountIssuesList = []
    amountCommitsList = []
    amountDevsList = []
    #result_path = os.path.join(os.path.dirname(os.path.abspath(__file__)), RESULT_FOLDER)
    #if not os.path.exists(result_path):
    #    os.makedirs(result_path)
    protocols = pd.read_csv(FILE)
    for i in range(len(protocols)):
        print("Protocol:", protocols["Github Tag"][i])
        amountRepos, amountStars, amountIssues, amountCommits, amountDevs = getTopicFacts(protocols["Github Tag"][i])

        amountReposList.append(amountRepos)
        amountStarsList.append(amountStars)
        amountIssuesList.append(amountIssues)
        amountCommitsList.append(amountCommits)
        amountDevsList.append(amountDevs)

    protocols.assign(**{
        'amountRepos':amountReposList,
        'amountStars':amountStarsList,
        'amountIssues':amountIssuesList,
        'amountCommits':amountCommitsList,
        'amountDevs':amountDevsList
    })
    protocols.to_csv("new.csv")

# %%
if __name__ == "__main__":
    print("Start")
    main()
    print("End")