private async Task LoadEventsAsync()
{
try
{
foreach (var name in await _store.ListAsync())
{
var payload = await _store.LoadAsync(name);
var headers = payload.Headers.ToDictionary(kv => kv.Key, kv => new StringValues(kv.Value.ToArray()));
var body = payload.Body;
var message = GitHubEventMessage.Parse(headers, body);
_processingService.Enqueue(message);
}
}
catch (Exception ex)
{
_logger.LogError(ex, "Couldn't load stored events");
}
}
private static async Task RunAsync(CrawledSubscriptionList subscriptionList, bool reindex, bool pullLatest, bool randomReindex, bool uploadToAzure, string startingRepoName, string outputPath)
{
var reindexIntervalInDays = 28;
var today = DateTime.Today;
var connectionString = GetAzureStorageConnectionString();
// TODO: We should avoid having to use a temp directory
var tempDirectory = Path.Combine(Path.GetTempPath(), "ghcrawler");
if (Directory.Exists(tempDirectory))
{
Directory.Delete(tempDirectory, recursive: true);
}
Directory.CreateDirectory(tempDirectory);
var cacheContainerName = "cache";
var cacheContainerClient = new BlobContainerClient(connectionString, cacheContainerName);
if (!reindex || startingRepoName is not null)
{
var startingBlobName = $"{startingRepoName}.crcache";
var reachedStartingBlob = false;
await foreach (var blob in cacheContainerClient.GetBlobsAsync())
{
if (!subscriptionList.Contains(blob.Name.Replace(".crcache", "")))
{
continue;
}
if (blob.Name == startingBlobName)
{
reachedStartingBlob = true;
}
if (reachedStartingBlob)
{
continue;
}
Console.WriteLine($"Downloading {blob.Name}...");
var localPath = Path.Combine(tempDirectory, blob.Name);
var localDirectory = Path.GetDirectoryName(localPath);
Directory.CreateDirectory(localDirectory);
var blobClient = new BlobClient(connectionString, cacheContainerName, blob.Name);
await blobClient.DownloadToAsync(localPath);
}
}
var factory = CreateGitHubClientFactory();
var client = await factory.CreateAsync();
var jsonOptions = new JsonSerializerOptions()
{
WriteIndented = true
};
var repos = new List <CrawledRepo>();
var reachedStartingRepo = reindex && startingRepoName is null;
foreach (var org in subscriptionList.Orgs)
{
var orgDirectory = Path.Join(tempDirectory, org);
Directory.CreateDirectory(orgDirectory);
var existingRepos = Directory.GetFiles(orgDirectory, "*.crcache")
.Select(p => Path.GetFileNameWithoutExtension(p));
if (!pullLatest)
{
Console.WriteLine($"Loading repos for {org}...");
foreach (var repoName in existingRepos)
{
var blobName = $"{repoName}.crcache";
var repoPath = Path.Join(orgDirectory, blobName);
var repo = await CrawledRepo.LoadAsync(repoPath);
if (repo is not null)
{
repos.Add(repo);
}
}
}
else
{
Console.WriteLine($"Requesting repos for {org}...");
var availableRepos = await RequestReposAsync(factory, client, org);
var deletedRepos = existingRepos.ToHashSet(StringComparer.OrdinalIgnoreCase);
deletedRepos.ExceptWith(availableRepos.Select(r => r.Name));
foreach (var deletedRepo in deletedRepos)
{
var blobName = $"{org}/{deletedRepo}.crcache";
var repoPath = Path.Join(tempDirectory, blobName);
Console.WriteLine($"Deleting local file {blobName}...");
File.Delete(repoPath);
if (uploadToAzure)
{
Console.WriteLine($"Deleting Azure blob {blobName}...");
await cacheContainerClient.DeleteBlobAsync(blobName);
}
}
foreach (var repo in availableRepos)
{
if (!subscriptionList.Contains(org, repo.Name))
{
continue;
}
var blobName = $"{org}/{repo.Name}.crcache";
var repoPath = Path.Join(tempDirectory, blobName);
if (string.Equals($"{org}/{repo.Name}", startingRepoName, StringComparison.OrdinalIgnoreCase))
{
reachedStartingRepo = true;
}
CrawledRepo crawledRepo;
try
{
crawledRepo = await CrawledRepo.LoadAsync(repoPath);
}
catch (JsonException)
{
Console.WriteLine($"WARNING: Couldn't parse {blobName}");
crawledRepo = null;
}
if (crawledRepo is null)
{
crawledRepo = new CrawledRepo
{
Id = repo.Id,
Org = org,
Name = repo.Name
};
}
crawledRepo.IsArchived = repo.Archived;
crawledRepo.Size = repo.Size;
repos.Add(crawledRepo);
var repoIsDueForReindexing = crawledRepo.LastReindex is null ||
crawledRepo.LastReindex?.AddDays(reindexIntervalInDays) <= today;
if (reachedStartingRepo)
{
Console.WriteLine($"Marking {repo.FullName} to be re-indexed because we reached the starting repo {startingRepoName}.");
}
if (repoIsDueForReindexing)
{
if (crawledRepo.LastReindex is null)
{
Console.WriteLine($"Marking {repo.FullName} to be re-indexed because it was never fully indexed.");
}
else
{
Console.WriteLine($"Marking {repo.FullName} to be re-indexed because it was more than {reindexIntervalInDays} days ago, on {crawledRepo.LastReindex}.");
}
}
if (reachedStartingRepo || repoIsDueForReindexing)
{
crawledRepo.Clear();
}
}
}
}
// We want to ensure that all repos are fully-reindexed at least once every four weeks.
// That means we need to reindex at least #Repos / 28 per day.
//
// On top of that, we need to ensure that all repos which were never fully indexed (e.g.
// they are new or were forced to be reindexed) are also reindexed.
if (randomReindex)
{
var reposThatNeedReindexing = repos.Where(r => r.LastReindex is null).ToHashSet();
var minimumNumberOfReposToBeReindexed = (int)Math.Ceiling(repos.Count / (float)reindexIntervalInDays);
var numberOfReposThatNeedReindexing = reposThatNeedReindexing.Count;
if (numberOfReposThatNeedReindexing < minimumNumberOfReposToBeReindexed)
{
// OK, there are fewer repos that need reindexing than what we want to reindex
// per day. So let's randomly pick some repos to reindex.
var remainingRepos = repos.Except(reposThatNeedReindexing).ToList();
var choiceCount = minimumNumberOfReposToBeReindexed - numberOfReposThatNeedReindexing;
var random = new Random();
for (var choice = 0; choice < choiceCount; choice++)
{
var i = random.Next(0, remainingRepos.Count);
var repo = remainingRepos[i];
Console.WriteLine($"Marking {repo.FullName} to be re-indexed because it was randomly chosen.");
repo.Clear();
reposThatNeedReindexing.Add(repo);
remainingRepos.RemoveAt(i);
}
}
}
if (pullLatest)
{
Console.WriteLine($"Listing events...");
var eventStore = new GitHubEventStore(connectionString);
var events = await eventStore.ListAsync();
Console.WriteLine($"Crawling {repos.Count:N0} repos, fully reindexing {repos.Count(r => r.LastReindex is null):N0} repos...");
foreach (var crawledRepo in repos)
{
var blobName = $"{crawledRepo.FullName}.crcache";
var repoPath = Path.Join(tempDirectory, blobName);
var since = crawledRepo.IncrementalUpdateStart;
var messages = new List <GitHubEventMessage>();
if (since is null)
{
Console.WriteLine($"Crawling {crawledRepo.FullName}...");
}
else
{
var toBeDownloaded = events.Where(n => string.Equals(n.Org, crawledRepo.Org, StringComparison.OrdinalIgnoreCase) &&
string.Equals(n.Repo, crawledRepo.Name, StringComparison.OrdinalIgnoreCase))
.ToArray();
if (toBeDownloaded.Any())
{
Console.WriteLine($"Loading {toBeDownloaded.Length:N0} events for {crawledRepo.FullName}...");
var i = 0;
var lastPercent = 0;
foreach (var name in toBeDownloaded)
{
var percent = (int)Math.Ceiling((float)i / toBeDownloaded.Length * 100);
i++;
if (percent % 10 == 0)
{
if (percent != lastPercent)
{
Console.Write($"{percent}%...");
}
lastPercent = percent;
}
var payload = await eventStore.LoadAsync(name);
var headers = payload.Headers.ToDictionary(kv => kv.Key, kv => new StringValues(kv.Value.ToArray()));
var body = payload.Body;
var message = GitHubEventMessage.Parse(headers, body);
messages.Add(message);
}
Console.WriteLine("done.");
}
Console.WriteLine($"Crawling {crawledRepo.FullName} since {since}...");
}
if (crawledRepo.LastReindex is null)
{
crawledRepo.LastReindex = DateTimeOffset.UtcNow;
}
crawledRepo.AreaOwners = await GetAreaOwnersAsync(crawledRepo.Org, crawledRepo.Name);
var currentLabels = await RequestLabelsAsync(factory, client, crawledRepo.Org, crawledRepo.Name);
SyncLabels(crawledRepo, currentLabels, out var labelById);
var currentMilestones = await RequestMilestonesAsync(factory, client, crawledRepo.Org, crawledRepo.Name);
SyncMilestones(crawledRepo, currentMilestones, out var milestoneById);
// NOTE: GitHub's Issues.GetAllForeRepository() doesn't include issues that were transferred
//
// That's the good part. The bad part is that for the new repository where
// it shows up, we have no way of knowing which repo it came from and which
// number it used to have (even when looking at the issues timeline data),
// so we can't remove the issue from the source repo.
//
// However, since we're persisting GitHub events we received, we'll can look
// up which issues were transferred and remove them from the repo. This avoids
// having to wait until we fully reindex the repo.
//
// Note, we remove transferred issues before pulling issues in case the issues
// were being transferred back; it seems GitHub is reusing the numbers in that
// case.
foreach (var message in messages.Where(m => m.Body.Action == "transferred"))
{
Console.WriteLine($"Removing {message.Body?.Repository?.FullName}#{message.Body?.Issue?.Number}: {message.Body?.Issue?.Title}");
var number = message.Body?.Issue?.Number;
if (number is not null)
{
crawledRepo.Issues.Remove(number.Value);
}
}
foreach (var issue in await RequestIssuesAsync(factory, client, crawledRepo.Org, crawledRepo.Name, since))
{
var crawledIssue = ConvertIssue(crawledRepo, issue, labelById, milestoneById);
crawledRepo.Issues[issue.Number] = crawledIssue;
}
foreach (var pullRequest in await RequestPullRequestsAsync(factory, client, crawledRepo.Org, crawledRepo.Name, since))
{
if (crawledRepo.Issues.TryGetValue(pullRequest.Number, out var issue))
{
UpdateIssue(issue, pullRequest);
}
// TODO: Get PR reviews
// TODO: Get PR commits
// TODO: Get PR status
}
await crawledRepo.SaveAsync(repoPath);
if (uploadToAzure)
{
Console.WriteLine($"Uploading {blobName} to Azure...");
var repoClient = new BlobClient(connectionString, cacheContainerName, blobName);
await repoClient.UploadAsync(repoPath, overwrite : true);
if (since is null)
{
var eventsToBeDeleted = events.Where(e => string.Equals($"{e.Org}/{e.Repo}", crawledRepo.FullName, StringComparison.OrdinalIgnoreCase))
.ToArray();
Console.WriteLine($"Deleting {eventsToBeDeleted.Length:N0} events for {crawledRepo.FullName}...");
foreach (var e in eventsToBeDeleted)
{
await eventStore.DeleteAsync(e);
}
}
}
}
}
foreach (var repo in repos)
{
var milestones = repo.Milestones.ToHashSet();
var labels = repo.Labels.ToHashSet();
foreach (var issue in repo.Issues.Values)
{
foreach (var label in issue.Labels.Where(l => !labels.Contains(l)))
{
Console.Error.WriteLine($"error: {repo.FullName}#{issue.Number}: label '{label.Name}' doesn't exist");
}
if (issue.Milestone is not null && !milestones.Contains(issue.Milestone))
{
Console.Error.WriteLine($"error: {repo.FullName}#{issue.Number}: milestone '{issue.Milestone.Title}' doesn't exist");
}
}
}
Console.WriteLine("Creating trie...");
var trie = new CrawledTrie <CrawledIssue>();
foreach (var repo in repos)
{
foreach (var issue in repo.Issues.Values)
{
trie.Add(issue);
}
}
Console.WriteLine("Creating index...");
var index = new CrawledIndex()
{
Repos = repos.ToList(),
Trie = trie
};
var indexName = "index.cicache";
var indexPath = string.IsNullOrEmpty(outputPath)
? Path.Join(tempDirectory, indexName)
: outputPath;
await index.SaveAsync(indexPath);
if (uploadToAzure)
{
Console.WriteLine("Uploading index to Azure...");
var indexClient = new BlobClient(connectionString, "index", indexName);
await indexClient.UploadAsync(indexPath, overwrite : true);
}
Console.WriteLine("Deleting temp files...");
Directory.Delete(tempDirectory, recursive: true);
}