New native Version of QR-Codes seems harder to read.

[RolfBremer]

Thanks for this phantastic package!

I noticed that the new version produces QR-Codes that are harder to read. Here is a sample I created to test this:

The code to generate the codes is here:

#[
  #import "@preview/codetastic:0.1.0"

  #h(1fr)
  #box[#codetastic.qrcode(width: 42pt, "Dies ist ein dichter QR-Code mit vielen Informationen")]
  #h(1fr)
]
#[
  #import "@preview/codetastic:0.2.0"

  #h(1fr)
  #box[#codetastic.qrcode(width: 42pt, "Dies ist ein dichter QR-Code mit vielen Informationen")]
  #h(1fr)
]

Any idea what´s wrong?

[jneug]

QR-Codes are masked with a masking pattern, to create a code that is easy to read for scanners. Version 0.2.0 of codetastic uses a simplified algorithm to calculate the best pattern, to speed up the creation process. But this approach will not always produce the most readable code (though they are correct qr-codes). Using the original algorithm did take considerable time for medium to large codes.

I'm thinking of ways to get better result while still keeping rendering times feasible. Maybe using the original algorithm up to a certain size and switching to the approximation for larger versions.

For now, you could set the masking pattern manually with the mask argument. You can get the best mask from online generators like this one.

In your example, mask: 2 is used by the linked generator (while mask: 6 is used in the first code generated by codetastic).

#[
  #import "@preview/codetastic:0.1.0"

  #h(1fr)
  #box[#codetastic.qrcode(width: 42pt, "Dies ist ein dichter QR-Code mit vielen Informationen")]
  #h(1fr)
]
#[
  #import "@preview/codetastic:0.2.0"

  #h(1fr)
  #box[#codetastic.qrcode(width: 42pt, "Dies ist ein dichter QR-Code mit vielen Informationen", mask:6)]
  #h(1fr)
]
#[
  #import "@preview/codetastic:0.2.0"

  #h(1fr)
  #box[#codetastic.qrcode(width: 42pt, "Dies ist ein dichter QR-Code mit vielen Informationen", mask:2)]
  #h(1fr)
]

Some other things I noticed in your screenshot:

• The individual modules have a minimal border to mask the thin white lines the old codes had. But this might be a problem for small module sizes. I will look into this.
• The alignment pattern (square with a black dot in the bottom right quadrant) seems to be off in the new version. I can still scan the code, though. That might be a result of the error correction working, but the pattern is not in the correct place and thus makes scanning even harder.

[jneug]

I just submitted version 0.2.2 that fixes the last two issues.

The masking pattern thing will take some more time to optimize the code.

[RolfBremer]

Cool, I will look into it! Thanks!

[RolfBremer]

Great, Version 0.2.2 works perfectly. Reading occurs instantly! The new code appears a lot smaller than in version 0.1.0, also both have a width of 42pt.

I do not think, that is a problem. I see this Issue solved. Thanks again!

[jneug]

The smaller code size is intended. The sizing before was wrong, because the quiet zone (white space around the code) was ignored while calculating the size. The qr-code spec suggests to have a quiet-zone of at least four modules around the code.

If you want the actual codes to be 42pt, either set quiet-zone to 0 or set the width to 53.58pt, to take the quiet-zone into account.