Four orthogonal dials — layout · join · shape · view-settings (is a pivot a left join?)

The question that prompted this

While building crosswalk coverage views (2026-06-03), a sharp question came up: “a pivot could also be a left join — is that even a thing in our system?” It is — but only because a join and a pivot are two different dials, plus two more that get conflated with them. This log pins down the four so we can account for each independently.

The four dials

When you look at a coverage matrix, four independent decisions are baked into what you see. Turning one does not turn the others.

Dial	Concern	The question it answers	Choices	Layer
Layout	where the note physically lives	”where on disk does this concept/edge sit?“	folder · file · heading · tag · wikilink	Tier 1 emission (hierarchy primitives)
Join / verbs	which rows are in the set	”matches only? everything? just the gaps?“	inner · left · right · full · anti (+ filter, traverse, aggregate, diff…)	Layer A (relational)
Shape	how the set is drawn	”matrix? list? tree? graph?“	pivot · table · list · hierarchy · graph · timeline	Layer B (presentation)
View-settings	knobs on the chosen shape	”which axis? heatmap? sort? empty=gap/zero?“	rowsBy · colsBy · cellOp · heatmap · sort · empty	Layer B config

The whole point: a crosswalk-edge note (layout) can be inner/left/anti-joined (verb) and drawn as a pivot/table/list (shape) with any settings. One dataset answers many questions precisely because these stay orthogonal. Bake join semantics into the view, or layout into the query, and you lose that.

”Is a pivot a left join?”

No — a pivot is a shape; a join is a verb that produces the row-set the pivot draws. But the join type is exactly what changes whether the pivot shows gaps:

Feed the pivot with…	…and the matrix shows
inner join (CSF ⋈ mappings)	only CSF subcategories that have a mapping — populated cells only
left join (CSF ⟕ mappings)	every CSF subcategory, even unmapped — unmapped appear as empty rows / gap cells
anti join (CSF ▷ mappings)	only the unmapped subcategories — the gap list itself

So “a pivot of a left join” is a real, useful thing: it’s the coverage matrix with the gaps left visible instead of silently dropped. “A pivot of an anti-join” is the gap report. Same shape, different verb feeding it.

What Crosswalker supports on each dial today

The honest state (2026-06-03)

The dials exist as designed, but they aren’t all wired together yet — notably, today’s pivot is implicitly an inner join.

Dial	Shipped	Gap
Layout	folder · file · heading (src/render/mechanisms/)	tag · wikilink reserved for v0.2
Join / verbs	inner · left · right · full · anti join + filter · bind · set-op · diff as Layer A primitives (src/views/*-primitive.ts); traverse/closure in Tier 2	not yet wired into the pivot feed
Shape	table · list · cards (Bases-native) · pivot (crosswalkerPivot)	hierarchy (v0.1.7–8) · graph · timeline (v0.2)
View-settings	rowsBy · colsBy · cellOp · heatmap · sort · empty	reading them from the .base YAML is currently flaky — Bases moved the config location, so the view falls back to sensible defaults

The load-bearing nuance: the pivot view currently reads only the materialized crosswalk-edge notes — which are an already-inner-joined set (a note exists only where a mapping exists). So today’s coverage matrix can’t show gaps, because the unmapped subjects were never in its input. To get the left/anti behaviors (gaps, orphans) the pivot must be fed the full concept universe (the framework’s concept notes) joined against the edges — i.e. wire the shipped join primitives (or a Tier 2 query) into the pivot feed, instead of pointing it at the edge folder alone.

That is why, of the three coverage views, “gap / orphan” is a different beast than “density” or “lookup”: density and lookup live entirely in the edge set (inner); gaps require the universe + a join.

Takeaway — account for all four, separately

1. Layout is an emission decision (where the markdown lives). It does not constrain how you later query or draw it.
2. Join/verb is what decides the rows — and is where “show me the gaps” lives. It is the dial people mean when they say “left join.”
3. Shape is purely how the resulting set is rendered. Pivot is one shape; it is never itself a join.
4. View-settings are the per-shape knobs.

Next concrete step toward this: feed the pivot a left/anti-joined set (concepts ⟕ edges) so a single coverage view can toggle between “matches only” and “with gaps” — turning the join dial without touching the shape.

Related

• Ingestion tooling dev log — where the crosswalk-edge notes (the layout dial) come from
• GRC corpus end-to-end — the pipeline these dials sit on
• Hierarchy primitives (layout) · ETL & import · Terminology