Starting with 3.2.1, x and y coordinates in pdfData.Pages[].Texts are incorrect

[dpmott]

For example, when parsing a PDF such as test/pdf/misc/i28_line_break_210.pdf in this repo,

      const pdfParser = new PDFParser();

      pdfParser.on('pdfParser_dataError', (errData: PDFParserError) => {
        // handle error
      });

      pdfParser.on('pdfParser_dataReady', (pdfData: PDFData) => {
        try {
          // write pdfData.Pages[0].Texts to file
        } catch (error) {
          // handle error
        }
      });

      pdfParser.loadPDF(`test/pdf/misc/i28_line_break_210.pdf`);

The resulting output is something like this:

[ { x: -0.25,
    y: 48.75,
    w: 3,
    clr: 0,
    sw: 0.32553125,
    A: 'left',
    R: [ { T: '%20', S: -1, TS: [ 0, 15, 0, 0 ] } ] },
  { x: -0.25,
    y: 48.75,
    w: 110.016,
    clr: 0,
    sw: 0.32553125,
    A: 'left',
    R: [ { T: 'BY%20ORDER%20OF%20THE%20', S: -1, TS: [ 0, 16, 1, 1 ] } ] },
  { x: -0.25,
    y: 48.75,
    w: 3,
    clr: 0,
    sw: 0.32553125,
    A: 'left',
    R: [ { T: '%20', S: -1, TS: [ 0, 16, 1, 1 ] } ] },
  { x: -0.25,
    y: 48.75,
    w: 140.376,
    clr: 0,
    sw: 0.32553125,
    A: 'left',
    R: [ { T: 'SECRETARY%20OF%20THE%20AIR', S: -1, TS: [ 0, 16, 1, 1 ] } ] },
...

Sometimes the first item has a unique x/y coordinate, but thereafter all elements have the same x/y coordinate.

This makes spatial-aware parsing and grouping algorithms which depend on these coordinates useless.

Output was fine with version 3.2.0, and is broken in version 3.2.1 and 3.2.2.

The issue happens for both parseBuffer() and loadPDF().

The major refactor of version 3.2.1 added "Type3 glyph font support", but after reviewing the diff I was unable to identify where the above undesirable behavior was introduced.

[dpmott]

Here's a script that demonstrates the issue. Just put it in the root of the pdf2json project and run it.

const PDFParser = require('./dist/pdfparser.cjs');

async function main() {
  const pdfParser = new PDFParser();
  const promise = new Promise((resolve, reject) => {
    const uniqueXYPairs = {};

    pdfParser.on('pdfParser_dataError', (errData) => {
      reject(errData.parserError);
    });

    pdfParser.on('pdfParser_dataReady', (pdfData) => {
      try {
        const { Texts } = pdfData.Pages[0];
        const numTexts = Texts.length;
        Texts.forEach(text => {
          const key = `${text.x},${text.y}`;
          uniqueXYPairs[key] = true;
        });
        resolve([numTexts, Object.keys(uniqueXYPairs)]);
      } catch (error) {
        reject(error);
      }
    });

    pdfParser.loadPDF(`test/pdf/misc/i28_line_break_210.pdf`);
  });

  const [numTexts, uniqueXYPairs] = await promise;
  console.log(`Found ${numTexts} text items on the first page.`);
  console.log('Unique x,y text coordinates:', uniqueXYPairs.length);
};

main();

Expected:

Found 182 text items on the first page.
Unique x,y text coordinates: 182

Actual:

Found 182 text items on the first page.
Unique x,y text coordinates: 1

[modesty]

thanks for calling it out, just pushed a fix

[ekaoecanaan]

Oh.. I was just seeing this exactly the same problem, and about to report it. Thanks for fixing it.