Is `<-` allowed?

[ryotayamanaka]

a -> b :knows

is the syntax.

b <- a :knows

allows?

[nichtich]

No, but I'd suppport extending the Direction rule to also include <-.

[ryotayamanaka]

Need to consider whether <- improves usability.

FYI,

Formatter https://g2glab.github.io/pg-formatter/

Visualization https://sandbox.blitzboard.io/

[ryotayamanaka]

Thank you for the suggestion!

We also need to consider if <- should exist or not.

This implementation is based on the definition of the ISWC paper below (Definition 5): https://arxiv.org/abs/2203.06393

So, we should update this definition if it is necessary.

[nichtich]

I stick to the property graph model (Definition 4 in https://arxiv.org/abs/2203.06393). The model is too abstract to directly be serialized so there are two minor differences in my current implementation (can be discussed):

• the abstract model has no excplicit node identifiers (NODE_ID). You seem to allow numbers and strings, I changed to only allow strings so 1 is read as "1" (otherwise there could be two confusing nodes with distinct ids 1 and "1")
• the abstract model does not define a set of data types, PG syntax (Definition 5) allows string and number. I extended to also allow Boolean and Null as described in #34.

So, we should update this definition if it is necessary.

For a final specification it needs to be updated anyway because NODE_ID, STRING and NUMBER are not included and whitespace should better be included as well. I'll try to write a proposal to be discussed.

[hchiba1]

To be honest, I don't see the necessity for <-.

Pros(?)

• The benefits of introducing it seem limited. It does not add new meaning compared to using ->, nor is it a shorthand expression (i.e., it does not reduce typing or size). Might be useful for people who frequently change directions(?)

Cons

• Once this notation is recognized as valid, all applications and visualizers developed so far should also adapt to it. Things might get complicated.

• If <- is allowed, users may come up with even more complicated notations, e.g., a <- b -> c But it is difficult to support such notations.

• < is typically observed at the beginning of strings for expressing URIs. Therefore, I do not want it to be a special character.

If the merit is limited, then we should keep the specification simple.

[nichtich]

Thanks for summary of your arguments! For me it seemed just a minimal backwards-compatible addition to syntax which is being extended anyway. I can roll back the implementation in pgraphs and move the idea to the list of possible extensions not to be included in version 1.0.0 of PG format, if you prefer.

[nichtich]

I've update pgraphs so with version 0.4.0 direction <- is not allowed in PG format. This issue can be closed.