371. An affix pins to one edge per axis

Date: 2026-09-17

Status

Accepted. Closes book/src/TODO.md’s “An affix pins on one axis”.

Context

An affix took one edge=. The TODO entry described the case for two: “a header that is both sticky at the top and held against the left of a horizontally scrolling table”, which “needs two shifts and a rule about which wins”.

There is nothing for either to win. A vertical edge’s shift is a translateY and a horizontal edge’s is a translateX, and each is computed from the same three rectangles on its own axis: the hole, the clip and the container (ADR-0360).

Decision

An affix has an edge and an optional cross edge on the other axis, and each axis’s shift is computed independently.

• Affix.alsoPinnedTo(Edge) and edge="top left" in markup. A second edge on the same axis is refused in Java; in markup it is ignored, as an unknown edge word already is.
• AffixState.shiftFor(edge, offset, self, clip, container) is the per-edge arithmetic, lifted out unchanged. The state keeps a shift per axis, and :affixed is on when either is non-zero.
• AffixSlot and AffixContent carry shiftX and shiftY instead of an edge and a shift, and the content translates by both.

Consequences

• One offset applies to both edges.
• The table’s own sticky header (ADR-0360) is unchanged; a table in a horizontally scrolling viewport that wants its header held on the left too says alsoPinnedTo(Edge.LEFT) on an affix of its own.