Add syntax for character code constants. about language

[lrhn]

I have never given up on this as a language feature. It's just not a particularly high priority since the charcode package handles most of the use-cases adequately.

Using a const constructor is ingenious, but ugly. (There is precedence in the fromEnvironment constructors, but they are also ugly).

That said, focusing on Unicode code points is not necessarily the correct level of abstraction. It's better than code units, but it still isn't complete grapheme clusters. That means that in many cases, you should not be looking at individual code points at all, and the places where it's correct, it's also very likely to be ASCII only. There is a reason that package:charcode has generally been adequate.

[jamesderlin]

This is an ugly idea, but for completeness' sake at the very least:

Since "9".charCodeAt(0) can't be constant, have we considered inverting it by adding a const constructor to int? e.g.:

const int.charCodeOf(String character)

It would be annoyingly verbose, but it wouldn't require any syntax changes. (It also might be on par with the existing int.fromEnvironment constructor.)

[lrhn]

This proposal is for code point (integer) constants written symbolically, so c"𠮷" wouldn't work for that. That's specified as evaluating to the code point of the single-code-point string.

Grapheme clusters are sequences of code points, which means that there is no representation distinction between that and a String, which is also a sequence of code points represented as a sequence of UTF-16 code units.
That makes g"𠮷" just a shorthand for "𠮷".characters, not a numeric constant.

(Having a shorthand for .characters is an interesting idea. I think it's aiming too low. I'd be interested in allowing arbitrary user-defined string prefixes, so you can define your own xml'<foo bar="baz">qux</foo>' and have that call the special (or not) xml function. It gets even more interesting if the prefix understands interpolations, so re'foo${bar}+' would call re with both the strings "foo" and "+" and the value of bar, and then it can do its own interpolation, like JavaScript's template literals.)

[gbracha]

Given that these are defined once and for all in the library and you can in fact write $F and get a readable, simple, concise constant, I'm not sure why we want a construct (consuming a valuable token that might be used for some more important syntactic construct).

---

Added this to the Later milestone.
Added Accepted label.

[lrhn]

I have a library now that handles a certain subset of characters. If I need other characters, or if other users need character codes, they will need to create their own library. It doesn't handle non-alphanumeric characters as well (q.v. $UNDERSCORE vs. #"_").

I think it makes sense for the language to support this feature instead of leaving it to libraries that are, by nature, either incomplete or very large.

[DartBot]

This comment was originally written by @simonpai

---

Just adding my 2 cents. Perhaps Dart can supply a library for char constants? In that case:

1. User can optionally include them
2. User can prefix the constants by importing with prefix, which avoids name collision

IMHO, this feature is like the String joining utility that Java misses. It doesn't really block anything if absent, but it will end up with everyone rebuilding the same wheel in every Dart project.

[efortuna]

We have this available in the dart:html library. http://api.dartlang.org/docs/bleeding_edge/dart_html/KeyCode.html

Can we make it available for both platforms instead?

[lrhn]

The key code library is a good example of creating just the thing you need for one purpose. It only has one A code. I can't see from the api whether that is a lower or upper case A, but a character code library in would need both - and not have a "windows key" entry.

In other words, I don't see the general applicability of the library.

[efortuna]

right. I guess I missed that when I read this bug the first time. Yes, for the KeyCode library, by design, we are providing constants for the numbers associated with a particular key on a keyboard, (so there is only one code for "a", lower and upper case because it is the same key on the keyboard) not the ascii (or other) char code for the letters.

[DartBot]

This comment was originally written by [email protected]

---

@Emily - I mixed up the key handlers and ascii in my email. Sorry for the confusion.

One advantage of having literals, is working with unicode characters. The $F constant example above obviously only works for characters which are also valid Dart identifiers.

For example - without literals:

const int _A_MACRON = 256; //'Ā'.codeUnits.first;

switch(c) {
   case _A_MACRON: print(c); break;
}

With literals - you can just write:

switch(c) {
   case c'Ā': print(c); break;
}

[DartBot]

This comment was originally written by [email protected]

---

Oops - I see that's a dup of LRN's comment above.

[lrhn]

Issue dart-lang/sdk#2093 has been merged into this issue.

[lrhn]

Issue dart-lang/sdk#18322 has been merged into this issue.

[kasperl]

Removed this from the Later milestone.
Added Oldschool-Milestone-Later label.

[kasperl]

Removed Oldschool-Milestone-Later label.

[Cat-sushi]

I'm from non-english speaking country Japan, and I would like to have g'𠮷' as a grapheme cluster constant.
It would include that package:characters should be a part of dart:core.

To be more precise, '𠮷' is represented by single code point U+20BB7, but by two code units of UTF-16 (a surrogate pair) 0xD842 0xDFB7.
I mean, each character constant should be a code point literal at minimum, but a grapheme cluster is better.

[Cat-sushi]

This proposal is for code point (integer) constants written symbolically, so c"𠮷" would work for that.

Good enough.

[eseidel]

Ran into this while writing an HTML tokenizer in Dart. https://github.com/RubberDuckEng/html6 The Tokenizer really wants to work on codepoints (int / Runes), but also wants to compare those against ascii literals (e.g. '<'). I could use package:codepoints but that seems like a bunch of namespace pollution for a few characters. 🤷 Not a big deal, just sharing the anecdote (feel free to hide the comment if not useful).

[lrhn]

If you use package:charcode v ^1.3.0, it can generate declarations for the constants you need.
It only needs to be a dev-dependency, and you can even remove it again after generating the file.

(I still want character constants, but until I get it, this covers most of my needs.)