Allow variable strings to be defined for a string offset

[Girgias]

This stops truncating strings with multiple bytes to one byte and using the first byte as the char to replace the string offset.

At the same time allow to pass an empty string to remove the byte from the string.

This partially invalidates Bug 71572.

[nikic]

This is an interesting idea... implicit truncation is definitely not good, but I'm not totally sure if I like these semantics better than making the assignment of multi-char strings an error.

[Girgias]

This is an interesting idea... implicit truncation is definitely not good, but I'm not totally sure if I like these semantics better than making the assignment of multi-char strings an error.

I'll bring it up on internals for discussion seems like the best way forward to see what people think.

[claudepache]

Problem: What is the expected result of

$str = "あello world"; // utf-8-encoded
$str[0] = "H";

? I guess, a non-surprising result is not really possible, because it would need the knowledge of the expected encoding of the string.

[Girgias]

Problem: What is the expected result of

$str = "あello world"; // utf-8-encoded
$str[0] = "H";

? I guess, a non-surprising result is not really possible, because it would need the knowledge of the expected encoding of the string.

You'll just get a byte stream which doesn't have a correct encoding. This is the case currently and has been since forever. This doesn't change any of the behaviour on write.

Now if one day we have unicode (or some sort of it) support then I would expect the codepoint to be altered and not the byte.

[cmb69]

You'll just get a byte stream which doesn't have a correct encoding.

That. And I'd rather not "improve" support for something that can't generally work; instead I'd let multibyte assignments fail.

[Girgias]

You'll just get a byte stream which doesn't have a correct encoding.

That. And I'd rather not "improve" support for something that can't generally work; instead I'd let multibyte assignments fail.

ACK.

Closing this in favour of an implementation which adds an explicit warning for multi bytes values.
Will try to code this when I've got time.