Generating several `Identifier`s from a single `match` rule?

[loutr]

Hello, thanks a lot for your work on this framework. The following is more of a question, not really an issue.

From my understanding of Hakyll, I believe one file (matched by a match rule) can only generate one Identifier (along some compilation rules that are triggered when necessary). However I was considering the following:

• have a single JSON file that describes a collection of items (like a list of photos for an album ¹);
• generate several pages which correspond to a paginated representation of the items.

Generating a single Compiler (Item [MyObject]) while matching a valid JSON file is easy to do. However pagination only works with a given list of Identifiers (obtained from a Pattern):

buildPaginateWith
    :: MonadMetadata m
    => ([Identifier] -> m [[Identifier]])  -- ^ Group items into pages
    -> Pattern                             -- ^ Select items to paginate
    -> (PageNumber -> Identifier)          -- ^ Identifiers for the pages
    -> m Paginate

It is always possible to perform some IO with the preprocess :: IO a -> Rules a function and iterate on the create rule from there, but this completely escapes the dependency mechanism; also, it means doing it for several JSON files cannot naturally be done with a e.g. match ["dir/*.json"] rule.

So I was wondering if there is any construction in Hakyll that would allow this more "idiomatically"?

Footnotes

1. I understand that Hakyll seems very "file-based" and that a natural alternative to what is presented here would be to have instead one file per list item and have Hakyll manage that properly, like what is done in the repo of jaspervdj. My questioning arises mostly because I find using a single JSON file more convenient and less verbose from the perspective of git. ↩

[jaspervdj]

The idea that every Identifier will have only one output file is sort of baked in to Hakyll, as you already mentioned. "versions" are one solution to this, but I've only used them to generate a fixed number of documents (e.g. an HTML and PDF version of a page). But maybe they provide a solution?