explore-services/services/sitemaps/extractUrlsFromSearch.ts

'use strict';

import {properties} from "../../explore/src/environments/environment";
import {SearchResearchResultsService} from "../../explore/src/app/openaireLibrary/services/searchResearchResults.service";
import {ResultPreview} from "../../explore/src/app/openaireLibrary/utils/result-preview/result-preview";

import {Identifier} from "../../explore/src/app/openaireLibrary/utils/string-utils.class";
import {SearchFields} from "../../explore/src/app/openaireLibrary/utils/properties/searchFields";
import {ContextsService} from "../../explore/src/app/openaireLibrary/claims/claim-utils/service/contexts.service";

const request = require('superagent');

function get(resultsPerUrl) {
  setTimeout(() => {
    let searchFields = new SearchFields();
    let fieldIdsMap = searchFields.RESULT_FIELDS;

    request.get(refineUrl, async function (err: any, refineResponse: any) {
      if (!refineResponse && err) {
        console.error("Error getting refine filters ",err);
      } else {
        let keys = refineResponse.body['refineResults'] ? Object.keys(refineResponse.body['refineResults']) : null;
        console.log("number of keys: " + keys.length);

        let allUrls = new Set();

        let promiseArray = [];

        for (let key of keys) {
          if(key == "community") {
            await communitiesPromise;
          }
          console.log("key: "+key+", number of values: " + refineResponse.body['refineResults'][key].length);

          for (let value of refineResponse.body['refineResults'][key]) {
            if(!value || !value.name || !value.id
              || value.name.toLowerCase().includes('unknown') || value.name.toLowerCase().includes('not available')
              || value.name == "unidentified" || value.name == "Undetermined") {
              console.log("filtered out: "+(value ? ("name: "+value.name + " - id: "+value.id) : value));
              continue;
            }

            if(key=="community") {
              let valueId = "";
              if(value.id) {
                let idArray = value.id.split("||");
                if(idArray) {
                  valueId = idArray[0];
                }
              }
              if(!valueId || !publicCommunities.includes(valueId)) {
                console.log("hidden community: "+valueId);
                continue;
              }
            }

            const url = resultsUrlPrefix + "&fq=" + key + " " + fieldIdsMap[key].equalityOperator + " \"" + encodeURIComponent(value.id) + "\"" + "&type=results&page=0&size=" + resultsPerUrl;

            promiseArray.push(new Promise((resolve, reject) => {
              request.get(url, function (err: any, response: any) {
                if (!response && err) {
                  reject(err);
                } else {
                  parseAllUrls(response, allUrls);
                  resolve();
                }
              })
            }).catch(error => {
              console.error("Error getting results ", error);
              fs.appendFileSync("./"+errorFileName, "no response  "+url);
              fs.appendFileSync("./"+errorFileName, err);
              fs.appendFileSync("./"+errorFileName, "\n");
            }));
          }
          console.log("");
        }

        await Promise.all(promiseArray);
        console.log("\nDuplicate urls: "+alreadyin + " vs unique urls: "+notin);

        fs.appendFile("./" + fileName, "\n</urlset>", function (err) {
          if (err) {
            return console.log("Error appending in file "+fileName+": ", err);
          }
          console.timeEnd("total_time");
        });
      }
    })
  })
}
// });

function parseAllUrls(response: any, allUrls: any) {
  // let allUrls: any = [];

  let responses: any = response.body['results'];
  let searchResearchResultsService: any = new SearchResearchResultsService();

  // if(responses) {
  //   let length = Array.isArray(responses) ? responses.length : 1;
  //   for (let i = 0; i < length; i++) {
  //     let resData = Array.isArray(responses) ? responses[i]['result']['metadata']['oaf:entity']['oaf:result'] : responses['result']['metadata']['oaf:entity']['oaf:result'];
  //
  //     if (resData['pid']) {
  //       if (!Array.isArray(resData['pid'])) {
  //         if (resData['pid'].classid && resData['pid'].classid == 'doi') {
  //           if (resData['pid'].content != '' && resData['pid'].content != null) {
  //             console.log("|"+resData['pid'].content+"|   "+(typeof resData['pid'].content));
  //             resData['pid'].content.replace("https://doi.org/", "");
  //           }
  //         }
  //       }
  //     }
  //   }
  // }


  let searchResults: any = searchResearchResultsService.parseResults("result", responses, properties);
  if(searchResults.length < 100 && searchResults.length > 0) {
    console.log("num of results: "+searchResults.length + "   " + response.request.url);
  }

  if(searchResults.length == 0) {
    fs.appendFileSync("./"+errorFileName, response.statusCode+"  "+response.request.url+"/n");
  }
  for(let j=0; j<searchResults.length; j++) {
    let resultPreview: any = ResultPreview.searchResultConvert(searchResults[j], searchResults[j].entityType);

    let pid: any = Identifier.getResultPIDFromIdentifiers(resultPreview.identifiers);
    let url;
    if(pid && pid.id) {
      url = getUrlByType(resultPreview.resultType, pid, pid.id);
    } else {
      url = getUrlByType(resultPreview.resultType, null, resultPreview.id);
    }
    if(allUrls.has(url)) {
      alreadyin++;
    } else {
      allUrls.add(url);
      let urlPre = "<url>\n" +
        "    <loc>";
      let urlSuf = "</loc>\n" +
        "    </url>";
      fs.appendFileSync("./"+fileName, urlPre + url + urlSuf);
      notin++;
    }
  }
  return allUrls;
}


// function parseAllUrls1(response) {
//   let allUrls = [];
//
//   let responses = response.body['results'];
//   let length = Array.isArray(responses) ? responses.length : 1;
//
//   for (let i = 0; i < length; i++) {
//     let p = new parsingFunctions.ParsingFunctions();
//     let resData = Array.isArray(responses) ? responses[i]['result']['metadata']['oaf:entity']['oaf:result'] : responses['result']['metadata']['oaf:entity']['oaf:result'];
//
//     let type = "result";
//     if (resData['resulttype']) {
//       type = resData['resulttype']['classname'];
//     }
//
//     if (resData['pid']) {
//       let identifiers = p.parseIdentifiers(resData['pid']);
//       let pid = string_utils.Identifier.getResultPIDFromIdentifiers(identifiers);
//
//       if(pid && pid.id) {
//         allUrls[i] = getUrlByType(type, pid, pid.id);
//       } else {
//         let canId = parsingFunctions.ParsingFunctions.parseRelCanonicalId(Array.isArray(responses) ? responses[i] : responses, "result");
//         allUrls[i] = getUrlByType(type, null, canId);
//       }
//     } else {
//       let canId = parsingFunctions.ParsingFunctions.parseRelCanonicalId(Array.isArray(responses) ? responses[i] : responses, "result");
//       allUrls[i] = getUrlByType(type, null, canId);
//     }
//   }
//   return allUrls;
// }
//
function getUrlByType(type: any, pid: any, id: any) {
  let parameter: any = "";

  if (type === "publication") {
    parameter = "articleId";
  } else if (type === "dataset") {
    parameter = "datasetId";
  } else if (type === "software") {
    parameter = "softwareId";
  } else if (type === "other") {
    parameter = "orpId";
  } else {
    parameter = "id";
  }
  if(pid) {
    parameter = "pid";
  }
  return landingPrefix+type+"?"+parameter+"="+id;
}

function getCommunities() {
  communitiesPromise = new Promise((resolve, reject) => {
    request.get(contextUrl, async function (err: any, communitiesResponse: any) {
      if (!communitiesResponse && err) {
        reject(err);
      } else {
        const contextsService = new ContextsService();
        publicCommunities = contextsService.parseCommunities(communitiesResponse.body, false).map(value => value.id);
        resolve();
      }
    })
  }).catch(error => console.error("Error getting communities ", error));
}

function buildSiteMap(resultsPerUrl) {
  console.time("total_time");

  let date = new Date();
  fileName = "sitemap_"+date.getFullYear()+"_"+(date.getMonth()+1)+"_"+date.getDate()+".xml";//+"_"+date.getTime();
  errorFileName = "error_"+date.getFullYear()+"_"+(date.getMonth()+1)+"_"+date.getDate()+".txt";//+"_"+date.getTime();
  console.log("Buiding sitemap in file: "+fileName+"\n");

  let sitemap = "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n" +
    "<urlset xmlns=\"http://www.sitemaps.org/schemas/sitemap/0.9\" xmlns:xsi=\"http://www.w3.org/2001/XMLSchema-instance\" xsi:schemaLocation=\"http://www.sitemaps.org/schemas/sitemap/0.9 http://www.sitemaps.org/schemas/sitemap/0.9/sitemap.xsd\">";

  fs.writeFile("./"+fileName, sitemap,  function(err) {
    if(err) {
      return console.log("Error writing in file "+fileName+": ", err);
    }
  });

  getCommunities();
  get(resultsPerUrl);
}


let fileName;
let errorFileName;
const fs = require('fs');

let alreadyin = 0;  // duplicate urls
let notin= 0;

let communitiesPromise;
let publicCommunities = [];

const refineUrl = "https://beta.services.openaire.eu/search/v2/api/resources2/?format=json&refine=true&fields=resultbestaccessright&fields=relfunder&fields=instancetypename&fields=resultlanguagename&fields=community&fields=resulthostingdatasource&fields=country&type=results&page=0&size=0";
const resultsUrlPrefix = "https://beta.services.openaire.eu/search/v2/api/resources2/?format=json";
const landingPrefix = "https://beta.explore.openaire.eu/search/";
const contextUrl = "https://beta.services.openaire.eu/openaire/contexts/";

buildSiteMap(100);