Hacker News Reader

Hacker News

Comment by barrkel | Hacker News Reader

Parent Full thread

barrkel

Strings aren't lists of characters. In that misconception, many common coding errors lie.

For example, you might think that reversing a string is equivalent to reversing its characters. It's not.

Similarly, you might think that indexing a string is the same as indexing a list. It's not.

Etc etc.

View on HN

That very much depends on your definition of a String.

I'm not sure what you mean by reversing a string not being equal to reversing its characters. It's certainly not always equal to reversing its bytes (depending on the encoding). Same goes for Strings that are implemented with UTF-16 characters (e.g. Java, win32 wchar_t etc.) all suffer from bad abstractions driven by implementation/efficiency concerns.

His point is that the reverse of the two character unicode string "LATIN-SMALL-LETTER-O COMBINIG-DIAERESIS" is not "COMBINING-DIAERESIS LATIN-SMALL-LETTER-O", but the original "LATIN-SMALL-LETTER-O COMBINIG-DIAERESIS". Blame it on the unicode consortium, but that is the way it is.

I wouldn't necessarily blame it on the unicode consortium but rather on the assumption that a character in a programming language should conform to the notion of a character in UCS-4 or some other unicode encoding.

From a programming language perspective, I think it would be ideal if strings could always be viewed as lists of encoding independent characters, such that reversing this list is equivalent to reversing the string.

If you want to be able to maintain a one-to-one mapping to unicode, you will need to use unique characters for an "ä" and an "a with combining-diaeresis" but ideally that should be hidden from the user of the language.

Thus, in my ideal world, both "LATIN-SMALL-LETTER-O-WITH-DIAERESIS" and "LATIN-SMALL-LETTER-O COMBINING-DIAERESIS" would each be one single element of a list of characters which is a string.

This problem exists even if we ignore unicode. What is the reverse of the following string?

    foo\r\n\bar

I imagine the most useful answer if you are writing a notepad-level application would be:

    rab\r\n\oof

At the level of abstraction I would expect from a String object, the reverse of \r\n should be \n\r. Basically, you example is another encoding issue. I think strings should be free from any encoding concerns. Those should be handled when parsing/serializing.

Even in UCS-4, strings are not reversible by reversing the code points. Are you suggesting that Unicode combining characters are an efficiency concern, rather than a correctness one?

A string is more like a bytecode program that can be executed to evaluate to a piece of human-readable text.

Reply on news.ycombinator.com