Clone

HTTPS: git clone https://vervis.peers.community/repos/yEzqv

SSH: git clone USERNAME@vervis.peers.community:yEzqv

Branches

master

1.mdwn

[[!template id=ticket class=task]] [[!tag /projects/smaoin-hs/decisions]]

[[!meta title=“Data structures”]]

Issue

At the time of writing, the top-level data layer in [[!rel4darcs smaoin]] is statements and statement blocks. There are no “models”, i.e. no triplestore or quadstore or anything like that.

I did plan to have them, using containers, but now I’m not sure where to put them. The questions I’d like to answer here are:

In which package should they be? smaoin or razom-basic-store or something else?
Which models should be provided, using which containers?

Process

Here are some ideas:

Set-based model
- List
- Ordered set (red-black tree etc.)
- Unordered set (hash table etc.)
Tree-based model (e.g. statement blocks)
Graph-based model
- See Data.Graph
- See what [[!hackage rdf4h]] uses

Thoughts on package:

If they are added to smaoin, it will have dependencies which most users won’t need. For example, users of set-based models don’t need any extra dependencies used by the graph model code. On the other hand, these would be commonly used packages installed anyway and having many dependants. And GHC hopefully optimizes away code the program doesn’t use.
If they are added to razom-basic-store, there is some level of mixing/coupling between the basic store API (the package’s purpose) and the data structures (which can be used with any other frontend API).
A simple small package, e.g. smaoin-model, can contain just the data definitions and a minimal set of utility functions. But it should be minimal enough compared to the basic store API, to justify the separate package.

Result

No final decision yet.

I’m starting in the smaoin package. Will possibly move to a separate smaoin-model package later.