Newsbot.Collector/Newsbot.Collector.Services/Jobs/RssWatcherJob.cs

using System.ServiceModel.Syndication;
using System.Xml;
using Newsbot.Collector.Database.Repositories;
using Newsbot.Collector.Domain.Consts;
using Newsbot.Collector.Domain.Entities;
using Newsbot.Collector.Domain.Interfaces;
using Newsbot.Collector.Domain.Models;
using Newsbot.Collector.Domain.Models.Config;
using Newsbot.Collector.Services.HtmlParser;
using Serilog;

namespace Newsbot.Collector.Services.Jobs;

public class RssWatcherJobOptions
{
    public string? ConnectionString { get; init; }
    public string? OpenTelemetry { get; init; }
    public bool IsEnabled { get; init; }
}

// This class was made to work with Hangfire and it does not support constructors.
public class RssWatcherJob
{
    private const string JobName = "RssWatcherJob";

    private IArticlesRepository _articles;
    private ILogger _logger;
    private IDiscordQueueRepository _queue;
    private ISourcesRepository _source;

    public RssWatcherJob()
    {
        _articles = new ArticlesTable("");
        _queue = new DiscordQueueTable("");
        _source = new SourcesTable("");
        _logger = JobLogger.GetLogger("", JobName);
    }

    public void InitAndExecute(RssWatcherJobOptions options)
    {
        _articles = new ArticlesTable(options.ConnectionString ?? "");
        _queue = new DiscordQueueTable(options.ConnectionString ?? "");
        _source = new SourcesTable(options.ConnectionString ?? "");
        _logger = JobLogger.GetLogger(options.OpenTelemetry ?? "", JobName);

        _logger.Information($"{JobName} - Job was triggered.");
        if (!options.IsEnabled)
        {
            _logger.Information($"{JobName} - Going to exit because feature flag is off.");
            return;
        }

        _logger.Information($"{JobName} - Setting up the job.");

        Execute();
    }

    public void Execute()
    {
        var articles = new List<ArticlesEntity>();

        _logger.Information($"{JobName} - Requesting sources");
        var sources = _source.ListByType(SourceTypes.Rss);
        _logger.Information("{JobName} - Got {SourcesCount} back", JobName, sources.Count);

        foreach (var source in sources)
        {
            _logger.Information("{JobName} - Starting to process \'{SourceName}\'", JobName, source.Name);
            _logger.Information($"{JobName} - Starting to request feed to be processed");
            var results = Collect(source.Url, source.ID);

            _logger.Information($"{JobName} - Collected {results.Count} posts");
            articles.AddRange(results);
        }

        _logger.Information($"{JobName} - Sending posts over to the database");
        UpdateDatabase(articles);

        _logger.Information($"{JobName} - Done!");
    }

    public List<ArticlesEntity> Collect(string url, Guid sourceId, int sleep = 3000)
    {
        var collectedPosts = new List<ArticlesEntity>();

        using var reader = XmlReader.Create(url);
        var feed = SyndicationFeed.Load(reader);

        foreach (var post in feed.Items.ToList())
        {
            var articleUrl = post.Links[0].Uri.AbsoluteUri;

            // Check if we have seen the url before
            // If we have, skip and save the site bandwidth
            if (IsThisUrlKnown(articleUrl)) continue;

            var meta = new HtmlPageReader(new HtmlPageReaderOptions
            {
                Url = articleUrl
            });
            meta.Parse();

            var article = new ArticlesEntity
            {
                Title = post.Title.Text,
                Tags = FetchTags(post),
                Url = articleUrl,
                PubDate = post.PublishDate.DateTime,
                Thumbnail = meta.Data.Header.Image,
                Description = meta.Data.Header.Description,
                SourceId = sourceId
            };

            collectedPosts.Add(article);

            // try to not be too greedy
            Thread.Sleep(sleep);
        }

        return collectedPosts;
    }

    public void UpdateDatabase(List<ArticlesEntity> items)
    {
        foreach (var item in items)
        {
            if (item.Url is null)
            {
                Log.Warning("RSS Watcher collected a blank url and was skipped");
                continue;
            }

            var p = _articles.New(item);
            _queue.New(new DiscordQueueModel
            {
                ArticleID = p.Id
            });
        }
    }

    private bool IsThisUrlKnown(string url)
    {
        var isKnown = _articles.GetByUrl(url);
        if (isKnown.Url == url) return true;
        return false;
    }

    private string FetchTags(SyndicationItem post)
    {
        var result = "";
        foreach (var tag in post.Categories) result += $"{tag.Name},";
        return result;
    }
}