Re: Empty Alternatives (a.k.a epsilon).

On Wed, Aug 26, 2026, at 04:31, Bethan Tovey-Walsh wrote:
> I can't see how forbidding "" provides any particular benefit to grammar authors. It might not be your preferred syntax, but I'd like to understand why you think having it as an option is harmful, and what you feel the benefits are of having "" be an error. 

I'm coming to this from the perspective of a not-so-smart person who doesn't have ownership of the grammar; I might have responsibility, but I'll be part of a group of not-so-smart people who can all modify the grammar. (Even when this group is me and instances of past-me, I find that this basis of concern holds.)

In this sort of group context, ambiguity can mean formal ambiguity (the machine can get more than one place) or it can mean the kind of informal ambiguity where someone reading the grammar is unsure of the intent of the author of that portion of the grammar. It's this kind of informal ambiguity which I would prefer to avoid. It removes a whole category of "did they mean to do that?" questions. Not necessarily difficult questions, but still an effort cost and an opportunity for mistakes simply by existing.

It's the same reason why I would much rather see concat($value,' ') than concat($value,' ') in XPath; the later has no formal ambiguity but it does have some informal ambiguity, especially in a context where the value might be (or might have been intended to be) a non-breaking space rather than a plain space. Or, similarly, (part of) why [not(normalize-space($value))] is preferable to [$value = ''] -- the not-normalize-space version is less ambiguous in intent.

A grammar where empty quotes are an error is that little bit less informally ambiguous; when I see empty quotes, I can be sure that the author of that line didn't mean to do that. If they might have meant to do that, questions of intent become that little bit more difficult and my chances of making a mistake increase. I think that's a positive to remove the informal ambiguity.

It's not quite like the recognition that the target audience for technical documentation is tired and stressed out, by and large and in the main; it might legitimately be hoped that people modifying a grammar will not always be experiencing deadline pressure while not at their best. But I generally would prefer when making decisions about syntax to err on the side of supposing the human person reading the thing isn't going to be in an optimal state when they need to comprehend what it means. As little informal ambiguity as possible is a good thing.

-- Graydon

--  
Graydon Saunders  | graydonish@fastmail.com
Þæs oferéode, ðisses swá mæg.
-- Deor  ("That passed, so may this.")