Chemical structure of trivial molecules appear without their nodes when using flagging

[adamjohnwright]

Migrated from reactome/PathwayBrowser#41, opened by @EliotRagueneau on 2025-11-13. That repository is archived; the Angular pathway browser now lives here.

---

e.g. https://reactome.org/beta/PathwayBrowser/R-HSA-975634?select=R-HSA-975634&flag=class:Complex!drug&path=R-HSA-1430728;R-HSA-196854

Should have the trivial node presents when initially loading a diagram with flagging, and not hidden.

[adamjohnwright]

Two defects found and fixed in a76379e — but I could not reproduce the original report on demand, so leaving this open for confirmation on beta.

1. The zoom handler was often not detached. Flagging detaches the zoom handler so trivial molecules stop fading, looking it up via cy.data('reactome').interactivity. But a single Style is bound to several graphs (diagram, legend, comparison) and that field only holds whichever was bound last. The legend binds inside an async subscribe, so it is a race. Confirmed at runtime — on the legend, style.cy points at the diagram rather than itself. cy.off() matches on function identity, so losing the race means the diagram’s own handler is never removed and keeps re-applying zoom-based opacity. Now looked up per graph via interactivityOf(cy).

2. Visibility rode on an inline style that nothing could restore. The base rule is .trivial { opacity: 0 }, and flagging overrode it with an inline opacity: 1. Inline styles beat stylesheet rules in cytoscape — but with the handler detached, nothing recomputes that value. If it is lost, the base rule takes over and no amount of zooming brings the molecules back, while their chemical structures keep being drawn by a separate handler that was never detached. That is exactly the reported symptom, and I hit it directly during this work. It is now a durable class, .trivial.always-visible, with the stale inline value cleared first.

Verified on the URL from #150 (R-HSA-975634, flag=class:Complex!drug): flagged → opacity 1 at zoom 0.25, 0.5 and 1.5; unflagged → 0, 0.5, 1, so the normal fade is untouched. Still correct after navigating to another pathway with the flag active.

Since both reports describe this as intermittent ("sometimes, but not always"), please re-check on beta.reactome.org and close if it looks right.

[adamjohnwright]

Curators have given us a reproduction, which this issue did not have before.

Their words: "Not all trivial molecules stay visible at any zoom, e.g., H2O and H+ disappear with zooming out."

Previously the reports said "sometimes, but not always" and we could not make it happen on demand. Zooming out with H₂O and H⁺ on screen is specific enough to write a test against, so this moves from cannot reproduce to testable.

Not started yet.

[adamjohnwright]

Fixed in #170, live on beta now.

What was wrong. Flagging marks trivial molecules .always-visible so they
stop fading with zoom. The zoom handler writes an inline opacity across every
trivial element, and in cytoscape an inline style beats any stylesheet rule — so
it painted straight over that mark. The earlier fix detached the handler from the
zoom event, which was not enough: a restyle calls it directly, and so does
opening an interactor. The class is now the authority and the handler leaves
anything wearing it alone.

That also covers the original report here. A structure drawn with no molecule
underneath it was the same fault seen from the other side: the node had been
faded to nothing while a different handler carried on drawing its structure.

How to check it. On beta, open a diagram with something flagged — for
example https://beta.reactome.org/PathwayBrowser/R-HSA-70171?FLG=PKM — and zoom
all the way out and back in. H₂O and H⁺ should stay visible throughout. With
nothing flagged they should still fade as you zoom away, which is the behaviour
that makes a busy diagram readable; please tell us if that part looks wrong.

There is now an automated check for this, so it cannot come back silently.