public async Task <PageRankingProcessorResponse> PageRankingProcessorHandler(PageRankingProcessorRequest request)
{
string targetPath = PathGenerator.Generate(request.SearchEngine, request.Keywords, request.TargetDate);
StorageClientListFolderResponse files = await _storageClient.ListFolder(new StorageClientListFolderRequest { FolderPath = targetPath });
List <Task <StorageClientDownloadResponse> > downloadTasks = files.FilePaths
.Select(f => _storageClient.Download(new StorageClientDownloadRequest {
DownloadPath = f
}))
.ToList();
List <PageRankingProcessorWithSequenceResponse> processedResponses = new List <PageRankingProcessorWithSequenceResponse>();
while (downloadTasks.Any())
{
Task <StorageClientDownloadResponse> finishedTask = await Task.WhenAny(downloadTasks);
downloadTasks.Remove(finishedTask);
StorageClientDownloadResponse downloadedData = await finishedTask;
processedResponses.Add(await GenerateResponse(request.SearchEngine, downloadedData.FilePath, downloadedData.FileData));
}
//combine
List <PageRank> pageRanks = new List <PageRank>();
int rank = 1;
foreach (var r in processedResponses.OrderBy(c => c.Sequence))
{
foreach (var p in r.Pages.OrderBy(c => c.Rank))
{
pageRanks.Add(new PageRank {
Rank = rank++, Url = p.Url
});
}
}
return(new PageRankingProcessorResponse
{
Date = request.TargetDate,
Pages = pageRanks
});
}
public async Task <int?> Resolve(string keywords, DateTime targetDate)
{
//fetch files from S3 and check for the next sequence to fetch
string targetPath = PathGenerator.Generate(SearchEngine.Google, keywords, targetDate);
StorageClientListFolderResponse files = await _storageClient.ListFolder(new StorageClientListFolderRequest { FolderPath = targetPath });
//expected format: sympli/SearchRankings/Google/e-settlements/2021-04-24/1.html
// sympli/SearchRankings/Google/e-settlements/2021-04-24/2.html
// ...
if (files == null || files.FilePaths == null || files.FilePaths.Count == 0)
{
return(1);
}
//duplicate! TODO: move and refractor
Regex regex = new Regex("(?i)/(\\d+).htm", RegexOptions.Singleline);
List <int> sequencesFromFilePaths = files.FilePaths.Select(filePath =>
{
var matches = regex.Matches(filePath);
if (matches.Any())
{
bool numberFound = int.TryParse(matches.First().Groups[1].Value, out int sequence);
return(numberFound ? sequence : 0);
}
return(0);
}).ToList();
if (sequencesFromFilePaths.Max() + 1 <= MaxSequence)
{
return(sequencesFromFilePaths.Max() + 1);
}
else
{
return(null); //max sequence exceeded, return NULL to terminate process.
}
}
