Replace "code" with "index"

[gdalle]

The whole MetaGraphsNext package uses the word "code" for what is usually called the index of a vertex. Maybe we could change that to get closer to LightGraphs conventions?

[gdalle]

Not sure who's still watching this repo, but @bramtayl I might move forward with this

[bramtayl]

I'm still watching. Still agree with #16 (comment) but I'm ok changing it too.

[gdalle]

If I understand correctly, you worry that there may be a confusion between

• the method that takes a label and returns the vertex index (currently named code_for)
• the method getindex that takes a label and returns the vertex metadata

I hadn't understood this before, and I think you're right. What would you think about an approach à la MetaGraphs, where getindex actually returns the vertex index instead of the metadata? It would make some of the code a bit heavier, but I think we may gain in clarity?

[gdalle]

Of course that would mean going from a Dict to an Array for vertex- and edge-level metadata, but maybe we would even gain some performance in the process?

[bramtayl]

In base, getindex means roughly get_at_index: you pass an index to a dict, and Julia returns what exists at that index. It has the convenient dictionary[index] syntax.

setindex! roughly means set_at_index!: you pass an index and a value to a dict, and Julia puts the value at that index. It has the convenient dictionary[index] = value syntax.

Because these have a convenient syntax, I wanted to reserve them for very common operations (using labels instead of codes, which I encourage). I think it might be confusing if we call codes indices, but pass labels (not codes/indices) to getindex and setindex!.

getindex actually returns the vertex index instead of the metadata

That might be ok, but then how would `setindex! work?

[gdalle]

I see how this might be confusing, however due to the convenient dictionary[index] syntax, users will never call the function getindex themselves. As for setindex!, I think in my setting we would throw an error.
In the end it comes down to the most common operation you want to perform:

• if you love tampering with metadata, g[label] = data is fine
• if you love applying graph algorithms, g[label] = index / code is better because it gives you access to index-based LightGraphs routines

[gdalle]

I guess I'm more used to building the metadata on vertices and edges once and for all, and applying lots of graph algorithms afterwards, which is why the second option made more sense to me

[gdalle]

Closing this for now