Just a couple of thoughts:

• I had a look at chooseMode. It currently iterates over the string on a byte basis. While this works in practice, a proper (although presumably slower) multibyte-aware iteration might be considered—though it’s not straightforward to implement using iconv alone (i.e., without relying on mbstring).
  Edit: A multibyte-aware loop would bring no benefit in this case, because the loop exits as soon as it encounters a byte that is neither numeric nor alphanumeric—both of which operate on single-byte characters.

• In the Encoder::encode() at hand, strlen() is sufficient when the mode is Mode::ALPHANUMERIC() or Mode::NUMERIC(). So iconv_strlen() is only strictly necessary for Mode::KANJI(), at the moment.
  Therefore, there might be a small optimization opportunity here: keeping strlen() for the Mode::ALPHANUMERIC() and Mode::NUMERIC() modes could avoid the overhead of iconv_strlen() when it's not needed.
  Edit: I've added a commit for this. strlen() is O(1), which is great.

Please rebase against main to get the tests working.

Rebased. Please let me know what you would like to see improved (including commit messages) before considering this for merging.

Looks like there's something wrong with your match syntax?

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 70.81%. Comparing base (bd2370d) to head (394a4fb).
⚠️ Report is 5 commits behind head on main.

@@             Coverage Diff              @@
##               main     #200      +/-   ##
============================================
+ Coverage     70.58%   70.81%   +0.22%     
+ Complexity      995      994       -1     
============================================
  Files            49       49              
  Lines          3182     3169      -13     
============================================
- Hits           2246     2244       -2     
+ Misses          936      925      -11     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

LGTM, thanks! :)

Looks like there's something wrong with your match syntax?

It's a shame, multiple expressions cannot be on separate lines…

Furthermore, default cannot be grouped with other expressions, i.e. the following is not possible:

Mode::KANJI(), default => iconv_strlen($content, 'utf-8'),

Actually, I just looked through the other modes – are we certain that they all require iconv_strlen?

I haven’t really dug into the other modes, so I can’t speak with confidence there. My changes were scoped specifically to Kanji mode.

Alternatively, it could look like this:

$numLetters = match ($mode) {
    Mode::BYTE()  => $dataBits->getSizeInBytes(),
    Mode::KANJI() => iconv_strlen($content, 'utf-8'),
    default       => strlen($content),
};

That said, I can’t say for sure whether this covers all cases correctly.

Well, as chooseMode in this context will only return those known modes, how about this:

Match on all expected modes coming from chooseMode, and in the default case throw an exception about an unexpected mode.

I had thought about this approach too (explicitly matching Mode::KANJI() and omitting the default branch).

I had discarded it because any hypothetical future mode addition could suddenly cause breakage, hence the idea of a default/fallback

But overall, that could still be an acceptable approach.

Note that match already throws an UnhandledMatchError on unmatched cases. But I suppose you would rather want to explicitly throw an exception.

oh, UnhandledMatchError is totally fine, I didn't know about that. But if somebody modifies chooseMode and adds new modes, they would then be warned by the match if it isn't handled explicitly, which is exactly what we want :)

Oh, right — encode() calls chooseMode() right at the beginning, which ensures the possible values for $mode.

And since chooseMode() is a private method only invoked there, its sole purpose is to simplify the logic within encode(), so the two are tightly coupled.

I’ve updated the PR accordingly, and also took the opportunity to rewrite appendBytes() using a match for consistency.