There was an error while loading. Please reload this page.
📝 docs: the Photon frame is clickable — the hit map and the shared origin
📝 docs: a Photon frame from the engine itself The cube button renders the buffer through the native engine in a disposable child process — the Reference rasterizer's pixels, which the GPU parity gates are pinned against. Documented with the two promises that make it safe to point at user code: a hung Build() dies in ten seconds as a sentence, and a throwing page renders the framework's own diagnostic surface.
📝 docs: native projects preview, and state survives the recompile Two additions the page was silent on: a native (Photon) project previews through the web realizer of the same abstract tree — with the toolbar saying so — and a page's own fields survive a recompile, carried through the SSR handoff's own door.
📝 docs: the screen as a list The layers tree, in both languages: what it reaches that the canvas cannot — a container filled edge to edge by its own children has no pixel of its own — and that picking a row is the same selection a click makes, not a second one.
📝 docs: it installs now "Built from source and points at the design host in the same checkout" was the honest instruction until this morning. It packages into a .vsix with the host inside it, so an install needs no checkout — and working ON the extension still uses the repository's own build, which is the part worth saying out loud, because a packaged host that answered for edited source would do it silently.
📝 docs: duplicate, and the keys Two additions the page was silent on: a node can be duplicated in place, and every gesture on it is also a key — arrows walk the tree, Alt+arrows move the node, Delete removes, Cmd/Ctrl+D duplicates, Esc lets go. Documented as a table, because a keyboard map is the one thing a reader scans rather than reads.
📝 docs: a pile is not a row The page described the drag as a caret between children, which is true everywhere except the one container where children overlap. A Stack's list order is paint order, so there the mark covers the child the node would land in front of and the panel's arrows are named for depth.
📝 docs: the fence is around the factory call, not around every construction The page said the inspector never rewrites the form the file is written in, and gave the init-only members as the example. Half of that stopped being true: a construction already written with `new` has CHOSEN that form, so its braces are added when something needs to go in them. Only a factory call still refuses, and now the page says which is which. Added with it: a closed set is a list, and the type is not the only thing that decides what is closed. Space is a static class of const float, so a `float gap` reads as a number until you notice what is written in it.
📝 docs: the inspector edits values as rows, and every list as a list The page said the panel descends one level into a value, which was true and made it sound like a detail. It is the shape of the thing: a BoxStyle is a sheet of key/value rows under the property that carries it, with a select of what is not set yet on the last line and an × on each row. Documented with the layout, because "descends one level" tells a reader nothing about what they will see. And the fence one level up moved too: a Grid's columns, a menu's items and a dialog's actions are the same shape as a children, so they get the same gestures. What differs is what goes in them, and that is read off the signature rather than written down as a rule about grids.
📝 docs: where a component could go, and whose components those are Two additions the page was silent on. The canvas now marks the insertion point itself — a + at each end of a hovered child, and for a container's own space, including an empty one, the nearest gap with a dashed line and a single +. Documented with the reason the dash is a dash: the caret marks a commitment, the dash marks a possibility. And the palette sentence was quietly wrong by omission. It said the palette is derived from the factory surface, which was true and left a reader thinking the framework's is the only one. The app's own components are in it too, first and under their own heading.
📝 docs: the three gestures are on the pointer too The page described insert, move and remove as things the panel does, which was true this morning. They are on the canvas now: drag a child to reorder it, hover one for a + at each end. Documented with the two refusals that are drawn BEFORE the gesture — the screen and the C# list must line up child for child, and a drop outside its container does nothing — because those are the part a reader cannot infer from using it, and the part that keeps a visual editor from corrupting a file. "No dragging on the canvas" became "no dragging between containers", which is what is actually still missing.
📝 docs: the visual editor page catches up with the editor The page was written when there was a preview and an identity mechanism and nothing else, and it has been saying "no editing from the canvas" through four slices that added exactly that. Nothing compiles a wiki, so a stale page stays green forever and the site publishes it. Now documented, EN and pt-BR: the three tiers a selected node is classified into and why they exist (a row built five hundred times in a foreach has one source site and no separate existence, so "delete this row" is wrong for it); the inspector, its three honesty rules — compiled before offered, WorkspaceEdit rather than a file write, never rewriting the form the file is written in — and the one level it descends so that padding and background are reachable at all; and the three structural gestures with the fence they live inside. The distinction the whole design turns on is now stated rather than implied: UNDERSTANDING is not fenced, only WRITING is. The editor opens code that already exists, and real code splits itself across helper methods and files. Both Roadmap pages carried a premise the work disproved on day one — that click-to-select would ride the V3 source maps eqc already emits. It cannot: the whole Build body is emitted through one call, so the finest position a map can name is the start of the method. The repo's ROADMAP said so; the wiki went on telling readers the opposite.
docs: the Origin mark names a search string only this change introduced `symbol="Origin"` is a substring of half the framework, so the guard that resolves a mark to the release it shipped in found a commit from preview.2 and called the mark wrong. It was right to: the search is `git log -S`, and a generic symbol resolves to whoever used the word first. `string? Origin` resolves to the commit that added the property and nothing else.
docs: the editor that renders what you are typing, in both languages Track E opens with a design host and an identity mechanism, and both are documented before either grows a user interface — because the part worth explaining is not the panel, it is why the preview compiles the BUFFER instead of the file, and why a node has to remember the C# that built it. The page says plainly what does not exist yet: the extension is built from source, there is no editing from the canvas, no native preview, no ServerAction calls, and the preview renders under the baseline theme. A wiki page that implies more than ships is the failure mode this one is written against. It also records the correction, because the roadmap said otherwise for weeks: click-to-select cannot ride the V3 source maps. The whole Build body is emitted through a single call, so the finest position a map can name is the start of the method — one shipped module has a 942-character line. Both pages in the same commit, as always: the translation staleness guard reads git history, and a canonical page committed ahead of its twin breaks the site.