如何使用PDF.js阅读页脚文本？

Question

I'm trying to extract DOIs from scientific papers, and as these are almost always located in the page footer, I'd like to try this strategy before going through the main text.

Here is my current approach, using Mozilla's pdf.js to search the first page of an arbitrary PDF.

var Promise = require('bluebird');
const doiRegex = new RegExp('\b(10[.][0-9]{4,}(?:[.][0-9]+)*/(?:(?!["&\'<>])[[:graph:]])+)\b', 'i');

function pdfgrep(fileObj) {
    return Promise.spawn(function* () {
        var pdf = yield pdfjs.getDocument(fileObj.path);
        console.log(pdf);
        var page = yield pdf.getPage(1);
        var text = yield page.getTextContent();

        for (var s of text.items) {
            var match = s.str.match(regex);
            if (match !== null) {
                return match;
            }
        }

        return null;
    });
}

Here is a PDF on which this method can be tested. Note that the DOI is located in the footer, and can be located using the search tool in any run-of-the-mill PDF viewer. However, pdf.getPage doesn't seem to include any text from the footer.

1. How can I access the footer text with PDF.js?
2. Failing that, are there any other tools I could use to do this?

Answer 1

The RegExp was not properly written:

• \\b are not escaped in the string, shall be \\\\b
• [:graph:] might not work

The following was meant:

var doiRegex = /\b(10[.][0-9]{4,}(?:[.][0-9]+)*\/(?:(?!["&\'<>])[\x21-\x7E])+)\b/i;

Answer 2

getTextContent() result returns text items with their positions on the page. Often PDF.js cannot combine individual characters into text runs because some PDF generators print individual glyph into separate positions, but that's improved in new versions of PDF.js (BTW, which version of PDF.js are you using?). Try to glue the text runs yourself before matching:

...
var text = yield page.getTextContent();
var str = text.items.map(function (s) {
    return s.str;
}).join('');
var match = str.match(regex);
return match;
...