Attach input location to tokens (add spans feature)

[not-my-profile]

Hey, thanks for this library ... it looks really promising :) I am working on an HTML linter for which I require the spans of parser errors, tag names, attribute names and attribute values. These spans would ideally be reported as core::ops::Range<usize>, so that I can pass them directly to the codespan_reporting library (codespan_reporting::diagnostic::Label::range in particular). Since span tracking is of course overhead it would be behind an off-by-default feature flag.

I recently implemented this in my fork of the html5ever tokenizer ... which I frankly would love to abandon for a more sound library :) If you are interested in this I can probably implement it.

[untitaker]

i think the easiest way to hack this in without adding overhead to the common case is to wrap the Reader in one that keeps track of the currently read character, then you can get the current file position and attach it to all the tokens you retrieve from the tokenizer. File position is not really a span but maybe close enough?

i'm interested in this feature but I have no idea how it could be done without adding overhead. probably using a generic that resolves to () for the common case?

[not-my-profile]

File position is not really a span but maybe close enough?

I want to output the nicest error messages possible ... for which I need spans ... not just the current position.

I have no idea how it could be done without adding overhead

I was thinking of a bunch of #[cfg(feature = "spans")]{ ... } in the giant (unreadable) match block.

[untitaker]

yes compile-time feature probably works, ideally it would be configurable per tokenizer instance for testability reasons though. perhaps it can be done via generics

[not-my-profile]

I am afraid I don't see how this could be nicely done with generics. What are these testability reasons you speak of?

[untitaker]

if I want to for example test with codespans, and once without, or benchmark with/without codespans I would have to recompile everything many times

generics idea:

enum Token<S> {
    StartTag(StartTag<S>),
    ...
}

struct StartTag<S> {
    name: String,
    attributes: BTreeMap<String, String>,
    span: S,
}

type Span = Range<char>;
type TokenWithoutSpans = Token<()>;
type TokenWithSpans = Token<Span>;

and make DefaultEmitter also generic over S.

Your entire tokenizer already has the emitter as type param, so now it can be Tokenizer<DefaultEmitter<()>, _> for running without spans, and Tokenizer<DefaultEmitter<Span>, _> for with spans

so during monomorphization you get two separately optimized codepaths

[jbirnick]

I would love to have Spans as well!

I want to find math formulas in HTML and replace them with a SVG-rendered image. And when the math syntax is wrong, I want to produce nice errors.

[untitaker]

fixed by #120, take a look at examples/spans.rs. Thanks to @Akida31 !