Misc #354
Misc #354
Conversation
src/handlers/graphical.rs
Outdated
| Line { | ||
| line_number: context_data.line() + line_number + 1, | ||
| offset: context_data.span().offset() | ||
| + unsafe { line.as_ptr().offset_from(context.as_ptr()) as usize }, |
There was a problem hiding this comment.
I believe this unsafe block is sound, but it would be good to have a // SAFETY: comment explaining the reasoning. I had to do a bit of digging to convince myself that it's not possible to have context.len() > isize::MAX.
Alternatively, you could avoid the offset_from call entirely by keeping a mutable offset variable and doing offset += line.len() or similar inside the map closure. That's a little ugly, but I'd personally prefer it to unsafe.
There was a problem hiding this comment.
That's the exact opposite of what I wanted to do. The whole reason I went with offset_from is to avoid using an offset variable, because that variable is a possible source of errors if not computed right (and indeed, such problem in the previous code is why I started to look at this piece of code in the first place).
Also, from a semantic point of view, the "offset from" meaning is what we are looking for ("the offset of line from the beginning of the data buffer"). A + line.len() is only indirectly related to that info. It's only correct if, and only if, the various lines follow each other exactly, without gaps nor overlaps. In particular, it requires the use of split_inclusive instead of the more common split1, creating, IMO, an unnecessary dependency between the split and the offset computation.
And yes it's safe per design:
The theoretical upper bound on object and array size is the maximum isize value. This ensures that isize can be used to calculate differences between pointers into an object or array and can address every byte within an object along with one byte past the end.
Footnotes
-
I would have used
splitexcept that it causes some other issues in some corner cases (I can't remember if it was when the data terminates with a newline or when the data was empty) ↩
- Fix a panic when a 0 length span covers the end of a document - Fix incorrect `line_count` - Add new unit tests and complete existing ones - Improve readability
The first two (refactor) are simplification of the code.
On the last one (fix), regarding the default SourceCode implementations:
\rwas handled as a newline, but is not supported in other places (Graphical/Narrated). So this too, depending on how one squints, it could be a breaking change or not.