an input range of characters (including strings) or a type that implicitly converts to a string type.
If r is not an auto-decodable string (i.e. a narrow string or a user-defined type that implicitly converts to a string type), then r is returned.
Otherwise, r is converted to its corresponding string type (if it's not already a string) and wrapped in a random-access range where the element encoding type of the string (its code unit) is the element type of the range, and that range returned. The range has slicing.
If r is quirky enough to be a struct or class which is an input range of characters on its own (i.e. it has the input range API as member functions), and it's implicitly convertible to a string type, then r is returned, and no implicit conversion takes place.
If r is wrapped in a new range, then that range has a source property for returning the string that's currently contained within that range.
import std.range.primitives; import std.traits : isAutodecodableString; auto r = "Hello, World!".byCodeUnit(); static assert(hasLength!(typeof(r))); static assert(hasSlicing!(typeof(r))); static assert(isRandomAccessRange!(typeof(r))); static assert(is(ElementType!(typeof(r)) == immutable char)); // contrast with the range capabilities of standard strings (with or // without autodecoding enabled). auto s = "Hello, World!"; static assert(isBidirectionalRange!(typeof(r))); static if (isAutodecodableString!(typeof(s))) { // with autodecoding enabled, strings are non-random-access ranges of // dchar. static assert(is(ElementType!(typeof(s)) == dchar)); static assert(!isRandomAccessRange!(typeof(s))); static assert(!hasSlicing!(typeof(s))); static assert(!hasLength!(typeof(s))); } else { // without autodecoding, strings are normal arrays. static assert(is(ElementType!(typeof(s)) == immutable char)); static assert(isRandomAccessRange!(typeof(s))); static assert(hasSlicing!(typeof(s))); static assert(hasLength!(typeof(s))); }
byCodeUnit does no Unicode decoding
string noel1 = "noe\u0308l"; // noël using e + combining diaeresis assert(noel1.byCodeUnit[2] != 'ë'); assert(noel1.byCodeUnit[2] == 'e'); string noel2 = "no\u00EBl"; // noël using a precomposed ë character // Because string is UTF-8, the code unit at index 2 is just // the first of a sequence that encodes 'ë' assert(noel2.byCodeUnit[2] != 'ë');
byCodeUnit exposes a source property when wrapping narrow strings.
import std.algorithm.comparison : equal; import std.range : popFrontN; import std.traits : isAutodecodableString; { auto range = byCodeUnit("hello world"); range.popFrontN(3); assert(equal(range.save, "lo world")); static if (isAutodecodableString!string) // only enabled with autodecoding { string str = range.source; assert(str == "lo world"); } } // source only exists if the range was wrapped { auto range = byCodeUnit("hello world"d); static assert(!__traits(compiles, range.source)); }
Refer to the std.uni docs for a reference on Unicode terminology.
For a range that iterates by grapheme cluster (written character) see std.uni.byGrapheme.
Iterate a range of char, wchar, or dchars by code unit.
The purpose is to bypass the special case decoding that std.range.primitives.front does to character arrays. As a result, using ranges with byCodeUnit can be nothrow while std.range.primitives.front throws when it encounters invalid Unicode sequences.
A code unit is a building block of the UTF encodings. Generally, an individual code unit does not represent what's perceived as a full character (a.k.a. a grapheme cluster in Unicode terminology). Many characters are encoded with multiple code units. For example, the UTF-8 code units for ø are 0xC3 0xB8. That means, an individual element of byCodeUnit often does not form a character on its own. Attempting to treat it as one while iterating over the resulting range will give nonsensical results.