Typing Content

[cuihtlauac]

There are currently 21 “sorts” of content processed by ood.

Each sort entails two ocaml types. One corresponds directly to the Yaml stored content (internally named metadata), and the other is the type exposed to the rest of the application (called data). The former is encapsulated inside `Ood, and the latter is exposed to the rest of the application.

Currently, each data type occurs three times:

• Twice in tool/ood-gen/lib, one as code, one reified as a string
• Once in src/ocamlorg_data

Branch https://github.com/ocaml/ocaml.org/tree/dedupe-data shows it is possible to make the triplication disappear at the cost of creating an additional internal dependency, giving the dependency diagram a diamond shape. Types at the apex, frontend and packages independent in the middle and web at the bottom. In the current state of the code, this is a no brainer refactoring (the kind which will hopefully be in AI reach soon).

In a second time, to complete phasing out of ood, the metadata types need to be handled in an automatized way.

Data types are mostly “extensions” of metadata (in the oo-sense). They have additional fields.

type metadata = {
  author : string;
  date : string;
  title : string;
  text : string;
}

type data = {
  author : string;
  date : string;
  title : string;
  text : string;
  text_html : string;
  ocaml : string list;
  slug : string;
}

Most importantly, additional data fields always functionally depend on metadata field contents.

Identity correspondence:

data.author = metadata.author
data.date = metadata.date
data.title = metadata.title
data.text = metadata.text

Functional correspondence:

data.text_html = metadata.text |> render_as_html
data.ocaml = metadata.text |> extract_samples
data.slug = markdown_filename |> slugify 

This suggests usage of a means to express

1. by-field correspondence with a serialized format
2. field initialization from field-dependent expressions
   both expressed as ppx-like attributes. The idea would be to enable expressing something like this:

type data = {
  author : string; [@key "author"]
  date : string; [@key "date"]
  title : string; [@key "title"]
  text : string = [@meta "markdown_body"];
  text_html : string = text |> render_as_html;
  ocaml : string list = text |> extract_samples;
  slug : string = [@meta "filename"] |> slugify 
}

It may be possible to use https://github.com/andersfugmann/ppx_protocol_conv as a starting point w.r.t. 1.