Goal: by the end of this lesson you can narrow the Explorer to the namespaces you work in, to everything built on a given fn, to fns with a given effect, or to what nothing uses — combine those, save the set as a view, and share a view with your team as an fn in the graph. Browser filters affect your Explorer; saving or editing a graph view changes its shared definition.
Concepts introduced: filter (one chip, one predicate),
view (a named set of filters), the view chip, + filter,
the personal hide (⊘), and a view saved in the graph
(the explorer-view fn).
The Explorer shows the whole graph — every namespace, every fn. On a real shared graph that is overwhelming, and most of it isn’t yours. Everything that narrows the tree is the same thing at three depths:
in core,
uses core.logic.const, fx io, unused;2 filters, or a view’s name;Changing a browser filter leaves the graph intact. A filter is about what is listed, never about what is reachable: fns outside the set still exist, still run, and references to them still resolve. Search always looks across the whole graph, so a filtered-out fn is still findable by name.
Under the kind chips sits + filter. It opens the menu of filters that need a value:
ADD A FILTER
Namespaces — only these
☑ core ☐ app ☐ web ← the graph's root namespaces
Hidden by you — restore
↺ core.tests
Uses a function
+ pick a fn — only what is built on it
Effect
io db network state time random env process raw-sql
In a graph view
☐ api-surface ← the views saved in the graph
Name contains
[ part of a name ]
☐ Unused — nothing references or extends it
view api-surface.name …
chip that stays on and can be part of a view.Each becomes a chip with an ×; the × is the way back. ◍ all clears every filter at once. Two honest edges: a server-evaluated set stops at 500 fns and the tree says Showing 500 of N — add a filter to narrow; a chip that names a fn or view deleted since you saved it turns ⚠ and matches nothing — its × is the fix.
Sometimes the noise is one namespace inside your scope — a scratch
area, an archive. Hover any namespace row and click its ⊘ (next
to rename / add / publish): the namespace vanishes from your tree
at any depth, and a not core.tests chip appears. Remove the chip
(or ↺ it in the + filter menu) to restore. It is a filter like
the others: the shared graph is untouched, teammates still see the
namespace, no permission is needed.
Top-left in the Build context bar — alongside the branch and packages chips (Lessons 23 and 32) — the view chip names what you are looking at. Click it:
VIEWS
◍ All functions — no filters
Device only
● on-const ×
web-services ×
In the graph
api-surface Edit filters Edit graph
Save the current filters
[ View name ] Save in graph Device only
Save in graph is the primary action (also Enter in the name field). It creates a versioned view for this branch. Device only saves an independent named filter set in this browser and does not update a same-named graph view. Device views and the active filters survive reload; another device starts with its own preferences.
Pick a saved view to restore its conditions. Changing a chip detaches the active set from its saved label. To update an existing graph view, use its Edit filters button: the name field is prefilled and Save changes updates that exact view identity. Ordinary filter edits keep this explicit edit context; ◍ all, applying another view or saving Device only leaves it.
Save in graph turns the active set into an ordinary fn:
{:name :api-surface
:parent :explorer-view
:args {:namespaces ["web" "app"]
:effects ["network"]
:uses :http-server}}
explorer-view runs a query over the current branch. Its conditions include
name, uses, effects, kinds, problems, namespaces, exclude,
unused, and membership in other views (also). The save operation retains
every selected uses / also reference, not just the first. References
carry fn identities, so a rename does not change what they mean. Multiple
uses and views intersect; namespace and kind choices OR within their axis.
Problem predicates evaluate the current failed / type-error / lint state.
The Apps kind uses the same installed addon data as the Apps panel.
The graph view is an ordinary fn, versioned and reviewable like any other change. Edit graph opens it by identity; ▶ Run returns matching functions, the total count, and whether the displayed list was capped. Saving or editing sends all conditions in one transaction. A failed save leaves the draft in place. A view with computed clauses cannot be flattened into editable chips; Edit graph is the explicit way to change it. Inherited fixed clauses can also refuse replacement, preserving the original view and explaining the graph fallback.
You can write one by hand too — extend explorer-view like any
base-fn, from the editor or an fns.edn.
Prefer to be shown? This lesson exists as a guided in-editor tour: open the demo with the tour running (no sign-up), or pick “Interactive tutorial” in the editor’s account menu.
core — the Explorer collapses to
core.*, an in core chip appears, the view chip reads
1 filter, and the “(primitives)” bucket disappears.2 filters.core.strings in the Explorer, click ⊘ — a
not core.strings chip appears and the namespace is gone from
your tree. Click the chip’s × to bring it back.const, pick const
(core.logic). Inspect the computed members; then ◍ all clears
this exercise.tutorial- and press Enter. Open the view chip, name the set
tutorial-function-view and choose Save in graph. Both predicates
must be saved. Device only is the separate browser option.tutorial-function-view; then reopen the view
chip and Save changes. The Functions condition stays, and the same
view is updated.Lesson 23 — Branches: fork, edit, diff, merge: a change that lives on its own branch until you fold it back.