Missing node identifier produces invalid dot file about dotnetgraph HOT 7 CLOSED

I'm ok with your first point, I'll add a piece of code to use a GUID if no identifier is set.
About the uniqueness of the identifiers, I think this is not the responsability of my current compiler implementation. I'd rather develop a parser that would validate graphs and return those type of errors.
I'm open to discussion :)

from dotnetgraph.

With GUIDs the unique identifier thing, isn't really an issue anymore, as I can simply never set one and the GUID will guarantee uniqueness. As such I can live without it.

Personally, I'd have such a constraint, either on the graph itself or latest in the compiler. And depending on whether you count a graph with duplicated node identifiers as valid or not, it can well be argued that it's the compiler's responsibility to generate a valid graph/dot file. 🤔

from dotnetgraph.

With Guid.NewGuid() this becomes a bit harder to test, as you now introduce a random element.
Do you think, it would make sense to introduce a identifier generator or similar that could be mocked?
Or would you rather adjust the test, to no do exact string matching on the output?

from dotnetgraph.

I think yes it could make sense to introduce a identifier generator since it's not a lot of work and it allows more freedom.

I've tested to create a graph with two nodes having the same identifier and it compiles, you can have a look here: https://dreampuf.github.io/GraphvizOnline/#digraph%20G%20%7B%0A%0A%20%20a%20%5Bcolor%3Dred%5D%0A%20%20a%20%5Bcolor%3Dgreen%5D%0A%7D

Maybe letting the developer to choose wether or not to check uniqueness (with simple bool parameter, disabled by default) could be a good compromise.

What do you think?

from dotnetgraph.

Yeah, dot does generate some graph, as such it might technically not be wrong (don't really fancy reading the spec), it's just that you'll usually end up with unexpected out comes.

Sounds like a good compromise to me.

from dotnetgraph.

I've started to implement it, you can check PR #11
For my current implementation if a node has no identifier it's generating one when the node is added to the graph. I am not sure this is the best design decision, and maybe it should be generated when the graph is compiled, so the IdentifierGenerator is linked to the compiler instead of the graph object.
What would be most practical for your use case?

from dotnetgraph.

Fixed by #28
Default identifier generator might be implemented in v3

from dotnetgraph.