Builds navigation, search, and content indexes from extracted entries.

This is the default DocShell.Presentation.Source — the one DocShell.Build uses unless a host substitutes its own. It takes the flat list of entries the extractors produced and derives the three indexes a documentation site needs:

• navigation — one DocShell.Presentation.NavigationItem per entry, sorted by kind then title, so modules, guides, and notebooks group together and each group reads alphabetically.
• search — one DocShell.Presentation.SearchEntry per entry, with the document flattened to plain text and pre-tokenized.
• content — a map from entry id to its AST nodes, so a renderer can load one page without parsing the whole set.

Deliberately flat

Every navigation item comes back with no children. That is not an omission: DocShell has no way to know whether your guides should nest under a section, whether modules should group by namespace, or whether the tree should follow the file layout at all. Those are product decisions, and a package that guessed at them would be wrong for most hosts and hard to override for the rest.

A host that wants structure has two options: reshape the flat list after DocShell.Build.run/1 returns it, or implement DocShell.Presentation.GraphProjector and own categorization outright.

Paths

Entry paths default to /docs/{kind}/{id}, which is a placeholder more than a recommendation. Hosts routing documentation anywhere else pass a :path_builder function:

StaticGenerator.project(
  entries: entries,
  path_builder: fn entry -> "/handbook/" <> entry["id"] end
)

The same function builds both navigation and search paths, so a search result and a nav link can never disagree about where a document lives.

Empty entries

Entries with no content are dropped. DocShell.Generate.ExDoc returns every module it can read, including ones marked @moduledoc false, because coverage reporting needs them — but a navigation tree listing every internal module as a blank page helps nobody. skip_empty: false keeps them.

Search text

Search content is the AST flattened depth-first with a space between nodes, which means code blocks, table cells, and link text are all searchable and no markup leaks into the index.

tokens is off by default and comes back []. It is the same text downcased and split on non-alphanumeric runs, and it costs roughly three quarters of the size of the text it duplicates — for a field the shipped renderer does not read, because it indexes title and content itself. Hosts wiring a search backend that wants a pre-split form set search_tokens: true.

Options

• :entries — the extracted entries to project; defaults to []
• :path_builder — a function from entry to path; defaults to default_path/1
• :skip_empty — drop entries with no content; defaults to true
• :search_tokens — populate SearchEntry.tokens; defaults to false