No way of resizing elements in diagrams #16
No Assignees
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: sorokin.oleg.edu/ARCADIA-Studio#16
Reference in New Issue
Block a user
You can't resize entities, activities and other diagram elements. This feature needs to be present to allow users to draw more good looking diagrams and make connection routing easier
This is a part of #1
Codex-956atf /root claims a bounded resizing design investigation alongside owned #3/#15. No other claim appears in current issue comments. Scope: map stored width/height and renderer minima, nested/allocation auto-fit, state symbol restrictions, route attachment and direct-manipulation lifecycle; compare concrete handle/constraint candidates and derive acceptance before implementation. Read-only investigation while current deletion verification runs; no production files assigned yet. Root branch codex/behavior-authoring-20260910. Acceptance for this design checkpoint: identify exact persisted semantics and an implementable direct-resize contract that preserves model identities, undo/source and usable nested layouts; final feature remains open.
Codex-956atf /root design checkpoint at
f7d84fa: full eight-handle persisted-rectangle resizing retained. Fresh independent critique supplied actual-core falsifiers; standalone Windows probe compiled -D warnings against the exact core library from the passing818-test source42511f0 and passed. Visible-size normalization is not idempotent:224x232 becomes232x232 from automatic port capacity; preserving inactive authored height makes feasible214x232 resize succeed. Manual anchors do not remove this effect. Valid100x50→110x50 also disproves a universal preferred floor on inactive axes. Durable evidence: harness/diagram-resize-foundation-20260912.md and exact seven-payload archive. Root accepts preserving inactive authored fields and exact projected preview/reopen equality instead of blanket normalization. Bounded inverse-geometry experiment now compares all eight handles and feasible candidate strategies; production gesture implementation not yet claimed. Child authored coordinates/semantics must remain unchanged; process/scenario layout persistence remains separate. Existing issue ownership unchanged.Codex-956atf experiment result (source
f7d84fa, unchanged core dependencies from full425): finite inverse-resize probe executed on 12 fixtures × 8 handles × 6 requests. 326 exact/no-op, 118 static-minimum, 48 dynamic unresolved and 84 ineligible outcomes; all static cases resolved. A 52,622-candidate finite grid found no missed feasible witness in 14 targeted searches. The unresolved port cases expose a projection constraint: changing an active edge can rotate automatic ports onto another face and impose a minimum on an otherwise unchanged dimension. Three targets are structurally impossible under the current capacity rule. Successful plans used interval seeds (at most five projections); residual iterative correction demonstrated no added value. This rejects shipping an iterative solver as a presumed solution. Next bounded step is compare port-capacity policy against eight-handle, inactive-field preservation and exact preview/reopen criteria. No resize UI acceptance claimed; source/report/receipts are being archived with the checkpoint.Codex-956atf experimental checkpoint published at
419c78con codex/release1-956atf: harness/resize-port-policy-experiment-20260912.md and its 73-payload source/receipt/geometry archive. Actual Windows comparison tests (8), allocator tests (3), and port-anchor tests (13) pass for isolated candidate 7e25b3d, but two independently designed acceptance falsifiers both fail at their intended criteria: 20-port owner height 259.9→260 causes 12 needless side changes; a top fallback port has a blocked mandatory exit stub and actual NoClearPath beside an unrelated obstacle while both prior modes avoid it. Candidate is not adoptable and has not entered production. Exact failed test source a95db10, binary/provenance and all 36 geometry artifacts are retained. Stable-ranking-only follow-up source 6203b5c is isolated and unrun; separate review also identifies common-capacity, cross-gap placement, primary-preference and gap-tier churn risks. No interval policy is accepted. Root compiler is reserved free during current native canvas acceptance.