Encoding binary data in QR Codes

The landscape of QR codes is complex, with many different standards and encoding schemes, there is a variety of guidance as to how to appropriately use these for encoding different types of data. This repository is home to a set of tests aimed at exploring the most appropriate way to encode raw binary into a QR code.

The main character encoding modes supported by QR codes are the following (table copied from here)

Mode	Characters	Compression	Max Characters
Numeric	0, 1, 2, 3, 4, 5, 6, 7, 8, 9	3 characters are represented by 10 bits	7,089
Alphanumeric	0–9, A–Z (upper-case only), space, $, %, *, +, -, ., /, :	2 characters are represented by 11 bits	4,296
Kanji	Characters from the Shift JIS system based on JIS X 0208	2 kanji are represented by 13 bits	2,953
Byte	Characters from the ISO/IEC 8859-1 character set	Each characters are represented by 8 bits	1,817

The problem

From the chart above the solution seems simple? Just use byte encoding. However the problem with this approach is that most QR code readers reading a QR code encoded in the 'byte' mode will actually process the bytes in the context of the ISO 8859-1 character set, which actually only supports a subset of values that a single byte can represent. This means a QR code that encodes raw binary not conforming to ISO 8859-1 wont reliably be read by QR code readers as they will often stop reading when they encounter a unsupported character, or alternatively not expose the ability to simply get the "raw bytes" encoded into a QR code.

Possible solutions

1. Encode in 'byte' mode with raw binary, only support readers that do not enforce ISO/IEC 8859-1 character encoding
   • Pro's - The most concise encoding of raw binary data
   • Con's - Difficulties with QR code readers that assume ISO/IEC 8859-1, there is also an inability to easily encode data in a way that can be formatted as a URI to support deep linking from outside of an application context.
2. Encode in 'byte' mode with base64 encoded content
   • Pro's - Supported by all QR code readers as it avoids collision with ISO/IEC 8859-1
   • Con's - Not very efficient, base64 is a 6 bit encoding scheme and the QR code 'byte' mode encoding is 8 bit, this means 2 bits of every byte in the encoded data wont be used. Leading to about 25% bloat in payload size
3. Encode in 'alphanumeric' mode with upper-cased base32 content.
   • Pro's - Second most efficient encoding of raw binary data as base32 is 5 bits and the QR code alphanumeric mode is a 5.5 bit encoding format (i.e 11 bits to every two bytes), information is also in text form suitable for URI formatting for deep linking
   • Con's slightly less efficient that raw binary ~ .5 bit per byte (6.25%).

In summary dependent on requirements, option 1 is the most concise, however option 3 offers the best tradeoffs in terms of broad QR code reader support and URI encode-ability whilst paying a 6.25% cost in size overhead.