为何char数组无法使用parallel的ref遍历？D语言技术疑问

Great question! Let's break down exactly why you're seeing this error with char arrays but not int arrays when using std.parallelism.parallel().

1. The core reason: Character arrays have special encoding safety constraints #

D's char type is an alias for ubyte, but it carries special semantic meaning: it represents a UTF-8 code unit. Unlike int (which is a standalone, independent value), char elements are part of a UTF-8 encoded text sequence—where multi-byte characters span multiple char elements.

The std.parallelism library is designed to avoid accidental corruption of valid UTF-8 sequences. When you use parallel() on a char[], the library restricts you from getting modifiable ref access to individual elements. This is because parallel modification could easily break multi-byte characters (e.g., if one thread modifies a byte that's part of a 2-byte UTF-8 character while another thread processes adjacent bytes). Even if your array only contains single-byte ASCII characters, the library enforces this restriction universally to prevent unsafe behavior in all cases.

In contrast, int[] elements are fully independent—modifying one int doesn't affect the validity of any other element. So parallel() safely allows ref access for numeric arrays.

2. How to work around this (if you know your data is safe) #

If you're certain your char[] only contains single-byte characters (like ASCII) and you want to modify elements in parallel, you can cast the array to a ubyte[] (the underlying raw byte type) to bypass the UTF-8 safety restrictions:

auto result = [ 'w','e','l','c','o','m','e','t','o','d'];
// Cast to ubyte[] to enable ref parallel iteration
foreach(ref c; parallel(cast(ubyte[])result)) {
    c = cast(ubyte)'a';
}
writeln(result); // Will output ["a", "a", ..., "a"]

This works because ubyte[] is treated as a raw byte array with no UTF-8 semantics, so parallel() allows modifiable ref access.

3. Confirming the type behavior #

Your code already prints the types of arr and result—they're int[] and char[] respectively. The difference in how parallel() handles these types comes down to the library's safety checks for character encodings, not the raw storage size or basic type properties.

内容的提问来源于stack exchange，提问作者user2346536