Version Selection is wrong for certain string length

[LXMK]

Hi, I've noticed that the version is sometimes not selected correctly, depending on length of the string.
ABCDEFGHIJKLMNO1 -> Version 1 (21x21)
ABCDEFGHIJKLMNO12 -> Version1 (21x21)
ABCDEFGHIJKLMNO123 -> Version2 (25x25)
ABCDEFGHIJKLMNO1234 -> Version2 (25x25)
ABCDEFGHIJKLMNO12345 -> Version1 (21x21)
ABCDEFGHIJKLMNO123456 -> Version1 (21x21)

Forcing to Level 1 (fit=False) gives the error:
qrcode.exceptions.DataOverflowError: Code length overflow. Data size (156) > size available (152)

[LXMK]

[LXMK]

qr = qrcode.QRCode(version=1,
                           error_correction=qrcode.constants.ERROR_CORRECT_L,
                           box_size=20,
                           border=2)
        qr.add_data(self.text)
        qr.make(fit=False)
        img = qr.make_image()

[SmileyChris]

It isn't just the length of the string taken into account, but the contents. Digits can be encoded more succinctly, so when 4 in a row are found, we switch encoding for that block meaning those extra numbers can use less space than encoded as alphanumeric

[LXMK]

It has nothing to do with digits! I've just used them for illustrating. It is the same when only letters are used!

[LXMK]

ABCDEFGHIJKLMNOPQRSTU -> Version 1 (21x21)
ABCDEFGHIJKLMNOPQRST -> Version 1 (21x21)
ABCDEFGHIJKLMNOPQRS -> Version2 (25x25)
ABCDEFGHIJKLMNOPQR -> Version2 (25x25)
ABCDEFGHIJKLMNOPQ -> Version 1 (21x21)
ABCDEFGHIJKLMNOP -> Version 1 (21x21)
ABCDEFGHIJKLMNO1 -> Version 1 (21x21)
ABCDEFGHIJKLMNO12 -> Version1 (21x21)
ABCDEFGHIJKLMNO123 -> Version2 (25x25)
ABCDEFGHIJKLMNO1234 -> Version2 (25x25)
ABCDEFGHIJKLMNO12345 -> Version1 (21x21)
ABCDEFGHIJKLMNO123456 -> Version1 (21x21)

[LXMK]

[SmileyChris]

It was still an issue with mode encoding, the case was just a bit different.
add_data() optimizes the chunks of data into the correct mode type, but if the data was shorter than the minimum optimization length, it wasn't attempting to pick the correct mode type -- now it does.