Saffron | News & announcements aggregation framework.

Table of Contents

• What is Saffron?
• Installation
• Initialization
• Configuration
• Parsers
  • WordPress V2
  • RSS
  • HTML
  • JSON / XML
  • Dynamic
  • Which to choose
• Article
• Source files
  • What is a source file?
  • Creating a source file
• Extensions
  • Register a middleware
  • Format article
  • Articles
• Standalone

What is Saffron?

Saffron stands for Simple Abstract Framework For the Retrieval Of News.

As said, saffron is a framework. It is an abstraction engine that helps you collect news and announcements from websites in a uniform way.

It supports different ways of data collection, such as API endpoints and web-scraping. It tries to ease the process of integrating all data sources by abstracting data collection into a few simple and powerful functions.

Installation

To install the latest release:

npm install @jexsrs/saffron

To install a specific version:

npm install @jexsrs/saffron@version

Initialization

Once you have installed the library:

import {initializeApp, getApp} from "@jexsrs/saffron";

const app = initializeApp();
app.start();

Configuration

scheduler

A function that will resolve the scheduler to be used. The scheduler is responsible for tracking which source needs to be scraped and when.

Default value is (app: Saffron) => new LinearScheduler(app).

To implement a custom scheduler, you can create a class that extends the Scheduler class.

import {Scheduler, initializeApp, getApp} from "@jexsrs/saffron";

class MyScheduler extends Scheduler {
  constructor(saffron: Saffron) {
      super(saffron);
  }

  // Return if the scheduler is running or not.
  get running(): boolean {
    return false;
  }

  // Starts the scheduler
  async start(reset: boolean): Promise<void> {
    this._running = true;
    // ...
    
    
    // When we need to call the scraping
    try {
      const result = await this.saffron.scrape(job.source);
      this.fire('success', job.source, result, null);
      
      // Reschedule source ...
    } catch (e) {
      this.fire('error', job.source, [], e);
      
      // Reschedule source ...
    }
  }

  // Stops the scheduler
  stop(): void {
    this._running = false;
    // ...
  }
}

const app = initializeApp({
  scheduler: (app: Saffron) => new MyScheduler(app)
});
app.start();

Available schedulers:

• LinearScheduler: will run all the sources in a linear fashion
• ExternalScheduler: will run all the sources based on a request function

sources

ignore_certificates

When true, the request will ignore all SSL certificates.

Default value is false.

delay_between_requests

The delay between requests on the same source in milliseconds.

Default value is 100.

axios

The axios configuration to be used for the requests. Can be JSON or function (source: Source<any>) => Promise<AxiosRequestConfig>.`

preprocessor

A preprocessor function that will be called between the request and the article parsing. Type is (responses: any, source: Source<I>) => Promise<any>.

articles.amount

The amount of articles to be retrieved from the source.

Default value is 30.

articles.extract_attachments_from_content

When true, any attachments from the content of the article will be extracted.

Default value is true.

on_request_fail

When fail, the source will be skipped if the request of one of the URLs fails. When skip, the source will continue to test the other URLs even if one of them fails.

Default value is fail.

dynamic_sources

An array of dynamic sources that will be used from the dynamic parser.

encoding

The encoding to be used for the requests. It will be passed in new TextDecoder(encoding) and be applied to the response.

Parsers

To retrieve the desired information from the websites we use parsers. There are four available parser types: wordpress, rss, html, xml, json and dynamic.

WordPress V2

Parser type: wordpress-v2

By default, WordPress based websites has an open API for news retrieval. We make use of that to get access on the articles and categories of the website.

To quickly check if a website supports the WordPress API simply open your browser and type <website-root-link>/wp-json/wp/v2/posts/. If a valid JSON file is displayed on the browser (or downloaded on your computer) which contains the website's articles, then you can safely use the wordpress parser.

RSS

Parser type: rss

Many websites support RSS feed. RSS allows users and applications to access updates to websites in a standardized, computer-readable format. You can check if a website supports RSS if you can see this icon .

JSON / XML

Parser type: json (or xml)

This parser is best to be used when it comes to pages that are loading data using API requests (e.g. lazy loading). The only prerequisite for this parser is that the response of the API requests is in a structured JSON or XML format.

HTML

Parser type: html

This parser uses scrapping tools like CheerioJS to scrape the website content and receive the displayed news. This parser is best to be used when the HTML in the website is structured. Websites where the HTML and CSS are not structured will be very difficult to scrape.

Dynamic

Parser type: dynamic

Unlike the other parsers, this parser uses typescript code to parse a website. The user decides all the logic for the scraping by extending the class Parser<I>.

import {Parser} from "@jexsrs/saffron";

type Instructions = {
    my_option: string;
}

export class Dynamic1 extends Parser<Instructions> {

  /**
   * The name of the dynamic parser. Must be unique
   */  
  get name(): string {
    return this.source.name;
  }

  /**
   * A function that will validate if the instruction field is correct.
   * It must throw an error if the data are incorrect.
   * @param data
   */
  public validate(data?: any): void {
    if(data.my_option && data.my_option !== 'hello') {
        throw new Error('invalid my_option');
    } 
  }

  /**
   * A function that will take the validated scrape configuration and return the parsed output.
   * Defaults to return the data as is.
   * @param data
   */
  public preprocess(data: any): Instructions {
    data.my_option ??= 'hello';
    return data;
  }

  /**
   * A function that will handle and asynchronous request to the source.
   * Defaults to a GET request.
   * @param source
   */
  public request(source: Source<I>): Promise<AxiosResponse> {
    return this.get(this.url);
  }

  /**
   * A function that will parse the result of the request.
   * @param source
   * @param result The result of the request.
   */
  async parse(source: Source<Instructions>, result: AxiosResponse): Promise<Article[]> {
    // Reuse the rss parser for this example
    const rss = new RssParser();
    rss.source = source;
    source.instructions = {
      extra_fields: [],
      assign_fields: {}
    };

    return rss.parse(source, result);
  }
}

Which to choose

We recommend a specific order for using the available parsers.

• If the desired website is based an WordPress and the WordPress articles API is enabled, then choose the wordpress-v2 parser.
• If the desired website supports RSS feed. then choose the rss parser.
• If the desired website is loading data using API requests with structured responses (e.g. lazy loading), then choose the json or xml parser.
• If the desired website has a structured form, the use the html parser.
• If none of the above is possible (bad html or custom API) then the dynamic parser is our last choice.

Article

We have created a universal format for the parsed news, and we named it Article.

type Article = {
  title: string;
  content: string;
  url: string;
  publication_date: string;
  thumbnail_url: string;

  author_name: string;
  author_image_url: string;

  attachments: {
    attribute: string;
    value?: string;
    text?: string;
  }[];
  
  categories: {
    name: string;
    links: string[];
  }[];

  timestamp: number;
  extra: Record<string, any>;
  source: string;
};

Source files

What is a source file?

A source file is a json file that represents a website. These files are generated from the user and guide Saffron on how to parse a website.

Creating a source file

The source file type is as follows:

type Source<I> = {
  // The unqiue name of the source.
  name: string;
  // The url(s) of the source.
  url: string | string[] | {url: string; categories: string[]}[];

  // The parser to be used.
  parser: 'rss' | 'json' | 'html' | 'dynamic' | 'wordpress-v2' | 'xml';
  // The custom instructions for the parser.
  instructions: I;

  // An object set by the user for the user, will not be used by Saffron.
  extra?: any;

  // Override Saffron's default options.
  options?: {
    ignore_certificates?: boolean;
    delay_between_requests?: number;
    axios?: AxiosRequestConfig | ((source: Source<I>) => Promise<AxiosRequestConfig>);
    preprocessor?: (responses: any, source: Source<I>) => Promise<any>;
    articles?: {
      amount?: number;
      extract_attachments_from_content?: boolean;
    };
    on_request_fail?: 'fail' | 'skip';
    dynamic_sources?: Parser<any>[];

    encoding?: string;
  };
};

Wordpress Instructions

type Instructions = {
  paths?: {
    posts?: string;
    categories?: string;
  };
  articles?: {
    include?: string[];
    dates?: {
      gmt?: boolean;
      fallback?: boolean;
    };
    filter?: {
      search?: string;
      author?: string;
      author_exclude?: string;
      after?: string;
      before?: string;
      slug?: string;
      status?: string;
      categories?: string;
      categories_exclude?: string;
      tags?: string;
      tags_exclude?: string;
      sticky?: boolean;
    };
    pagination?: {
      page?: number; // page
      size?: number; // per_page
    };
    thumbnail?: string;
    disable_thumbnail?: boolean;
  };
};

RSS Instructions

type RssInstructions = {
  extra_fields: string[];
  assign_fields: {
    [assign: string]: string;
  };
}

JSON / XML Instructions

type JsonSkipOptions = {
  find?: (string | number)[];
  text?: string;
  type?: 'exact' | 'contains'; // Default is 'exact'
} | {
  position: number;
};

type Instructions = {
  container: (string | number)[];
  skip?: JsonSkipOptions[];
  article: {
    [field: string]: {
      parent?: string;
      find?: (string | number)[];
      static?: string;
    };
  };
  unpaired_tags?: string[];
}

HTML Instructions

type HtmlSkipOptions = {
  selector?: string;
  attribute?: string;
  text?: string;
  type?: 'exact' | 'contains'; // Default is 'exact'
} | {
  position: number;
};

type Instructions = {
  container: string;
  scripting_enabled?: boolean;
  skip?: HtmlSkipOptions[];
  article: {
    [field: string]: {
      parent?: string;

      class?: string;
      find?: string[];
      attributes?: string[];
      multiple?: boolean;
      html?: boolean;

      static?: string;
    };
  };
}

Dynamic Instructions

type Instructions = {
  implementation?: string;
}

Extensions

An extension is a function that gets executed before the articles are returned to the scheduler. Extension functions can be useful for logging, article formatting or sorting.

The order where the extensions are executed is the order where they were registered. Each extension event can be called more than once.

Register a middleware

app.use("name", (...args: any) => {
    //...
});

Format article

For changing the contents of the articles. It gets as parameter every article that was found from the parsers and must return the same object when it changed.

app.use("article.format", (article: Article, source: Source<any>) => {
    // If possible set publication_date with milliseconds.
    let ms = new Date(article.publication_date).getTime();
    if (!isNaN(ms)) article.publication_date = ms;

    // Append source name before title for every article
    article.title = `[${source.name}] ${article.title}`;

    // Return the changed article.
    return article;
});

Note that any changes made on the source object will also affect the saved source.

Articles

This extension can be used to edit the articles in bulk. You can sort or filter them as you want. The only requirement is to return an array (empty or not) of articles.

app.use("articles", (articles: Article[], source: Source<any>) => {
    sort(articles);
    return articles.filter((article) => article.title != null && article.title !== "");
});

Note that any changes made on the source object will also affect the saved source.

Standalone

Saffron supports immediate parsing using the static function scrape.

import {initializeApp} from "@unistudents/saffron";

const app = initializeApp();

try {
    const articles = app.scrape({
        name: "source-name",
        url: "https://example.com",
        parser: "html",
        // ...
        instructions: {
            // ...
        },
    });

    console.log("Result:", articles);
} catch (e) {
    console.log("Encountered an error during parsing:", e);
}