Tartarus Dataset Fetch and Cleanup Tools

Various scripts for downloading and building datasets.

TL;DR

# Download English Wikimedia dumps
npx -p @tartarus/data td fetch wikimedia --output /tmp/

Prerequisites

Requires node and wget.

(Tested on MacOS with node==10.14.2 and wget==1.20.3.)

Usage

# Download Project Gutenberg library and catalogue
td fetch gutenberg --output /my/download/path

# Download Wikimedia dumps for English and Spanish language Wikipedia and Wikiquote 
td fetch wikimedia --output /my/download/path \
  --language en \
  --language es \
  --site wiki \
  --site wikiquote
  
# Crawl an API endpoint
td spider --output /my/download/path -- site /my/site/config/file.ts  

Spiders & Crawling

td spider collects sequential data from APIs. Both iterative counters (e.g. page number) and extractable URLs ('next page') are supported.

SpiderSiteConfig

A spider requires a JS/TS configuration file to customize its use.

import { SpiderJsonData, SpiderHttpNavigator, SpiderStore, SpiderHandle } from '@tartarus/data' 

export default {
  name: 'myexamplesite',
  
  // Determine how the received data will be parsed
  // Supported: SpiderJsonData, SpiderYamlData, SpiderXmlData, SpiderCsvData, SpiderHtmlData, SpiderTextData
  data: new SpiderJsonData(),
  
  // Determine what to query (required)
  navigator: new SpiderHttpNavigator(
    {
      // Base URL
      baseTarget: 'https://api.domain.ext/v1/list',
      
      // Determine how URLs are formed (return null to stop spidering)
      target: (h: SpiderHandle): string | null => `${h.getBaseTarget()}&page=${h.getIteration()}`,
      
      // Optional callback to test whether spidering should be ceased
      isDone: (h: SpiderHandle): boolean => false
    }
  ),
  
  // Determine how responses are stored
  store: new SpiderStore(
    {
      // Number of sub-directories to be used 
      subDirectoryDepth: 3,
      
      // Filename to be used; return null to stop spidering
      filename: (h: SpiderHandle) => `${h.getIteration()}.json`,
    }
  ),
  
  // Request decorator
  request: {
    headers: {
      'User-Agent': 'Tartarus-Data-Spider/1.0 ([email protected])'
    },
    
    method: 'get',
    responseEncoding: 'utf8'
  },
  
  // Spider behavior
  behavior: {
    delay: 1500, // Delay between requests in milliseconds
    retryDelay: 15000, // Delay before retrying failed requests
    maxRetries: 15, // Number of times to retry a failed request 
  }
}

SpiderHandle

The spider interface exposes information of its current status and the latest downloaded page by passing an instance of SpiderHandle class to the callback functions.

SpiderHandle.getResponseData(): SpiderParsedData

Returns an object that describes the data received in response to a successful query. The data element contains a parsed (JSON) object of the response. The raw element contains a string representation of the response data.

interface SpiderParsedData {
  raw: string;
  data: any;
}

SpiderHandle.getResponse(): SpiderNavigatorFetchResponse | null

Returns an object that contains a descriptor of a successful query (rawResponse) and the raw data received (rawData). The contents of the rawResponse element are dependent on the type of SpiderNavigator in use -- for SpiderHttpNavigator it will be an AxiosResponse<any>; for SpiderFileNavigator it will be set to null.

interface SpiderNavigatorFetchResponse {
  rawData: string;
  rawResponse: any;
}

SpiderHandle.getBaseTarget(): string

Returns the value passed in baseTarget element to the SpiderNavigator instance. Typically an URL.

SpiderHandle.getIteration(): number

Returns the current iteration.

SpiderHandle.getPath(relativeFilename: string): string

Returns an absolute path to relativeFilename in the output directory.

SpiderHandle.getSiteConfig(): SpiderSiteConfig

Returns the contents of the site configuration as described above.

SpiderHandle.getSpider(): SpiderTask

Returns the Task instance of the spider.