scrape_gs.js

/**
 * usage: node scrape_gs.js USERKEY
 *
 * Determine h-index for papers published AFTER each year found in a Google
 * scholar profile. The USERKEY is found in your Google scholar citations
 * page url.
 */

var request = require('request');
var cheerio = require('cheerio');
var _ = require('underscore');

var arguments = process.argv.slice(2);
var userKey = arguments[0];

var url = "http://scholar.google.com/citations?user=" + userKey + "&pagesize=1000"

request(url, function(err, resp, body) {

    $ = cheerio.load(body);

    var values = $(".gsc_a_tr").map( function() {
        var title = $(this).find( "a.gsc_a_at" ).text();
        var citations = parseInt( $(this).find( "a.gsc_a_ac" ).text() ) || 0;
        var year = parseInt( $(this).find( "span.gsc_a_h" ).text() ) || 0;
        // console.log( title, year, citations );
        return { year: year, citations: citations };
    });

    console.log( "Year TotalCitations h-index" );

    var byYear = _.groupBy( values, "year" );

    var cumCitations = []

    _.each( _.keys( byYear ).sort().reverse(), function( year ) {

        cumCitations = cumCitations.concat( 
            _.map( byYear[ year ], function ( v ) { return v.citations } ) );

        cumCitations.sort( function( a, b ){ return a - b } ).reverse()

        var h = 0;
        for ( var i = 0; i < cumCitations.length; i++ ) {
            if ( i >= cumCitations[i] ) {
                h = i;
                break; 
            }
        }

        var totalCitations = _.reduce( cumCitations, function( a, b ) { return a + b }, 0 );

        console.log( year, totalCitations, h );

    });

});