@yuneta/gobj-ui 7.23.155 → 7.23.157

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md CHANGED
@@ -1387,8 +1387,8 @@ G6 5.1 registers these layouts, and this is how each meets that shape:
1387
1387
  | `circular`, `concentric`, `grid`, `mds`, `random` | rings, rings by degree, a grid, projection, noise | none knows a parent from a child |
1388
1388
  | `combo-combined`, `fishbone` | combos / cause-effect | not this data |
1389
1389
 
1390
- So three of our own, in `treedb_layout.js` (pure, tested), registered as G6
1391
- layouts by three thin adapters in `c_g6_nodes_tree.js`:
1390
+ So two of our own, in `treedb_layout.js` (pure, tested), registered as G6
1391
+ layouts by two thin adapters in `c_g6_nodes_tree.js`:
1392
1392
 
1393
1393
  - **`treedb-tree`** — the classic tidy tree, **top to bottom**. A row per
1394
1394
  depth as tall as its tallest card, a node centred over the block of its
@@ -1402,25 +1402,6 @@ layouts by three thin adapters in `c_g6_nodes_tree.js`:
1402
1402
  keeps the other reading for a host that wants it). The outline that shipped
1403
1403
  beside it (`treedb-outline`, one row per node) was removed in `7.23.75`: a
1404
1404
  list that indents is a JSON viewer, and this library already has one.
1405
- - **`compact-tree`** (`treedb-compact`, `7.23.148`) — the same tree packed by
1406
- **contour** instead of by block. The tidy tree gives each subtree its whole
1407
- block, so a hall opened with twenty-four devices pushes the siblings of its
1408
- ancestors apart at depths where nothing of theirs touches; here every
1409
- sibling slides against the outline of the ones before it, and a closed
1410
- branch sits beside an open one at the depth they share. Same rows and order
1411
- as `treedb-tree`, so switching moves cards sideways and never reorders them.
1412
- One region open with a 24-device hall is 70% of the tidy width on a treedb
1413
- shaped like a real one; a uniform tree is the same width. The contour
1414
- arithmetic is Moen's (1990) as mxGraph wrote it in `mxCompactTreeLayout`
1415
- (JGraph Ltd, Apache-2.0), read in maxGraph — a port, not a dependency,
1416
- because maxGraph's layouts drive its own graph model. Since `7.23.154` a run
1417
- of three or more **leaf** children is **stacked**: two columns under the
1418
- parent with a corridor between them, row after row — the contour alone
1419
- gained nothing on a whole level opened with the stepper, where every parent
1420
- still needed the row of its children beside it. Against `tree`: regions
1421
- only 10% of the width, one region open with a 24-device hall 21%, a whole
1422
- level open 41%. The tree grows taller instead; with elbow edges or curves
1423
- alike, since `7.23.155` a curve keeps off the cards too.
1424
1405
  - **`radial`** (`treedb-radial`) — the root in the middle, a ring per depth,
1425
1406
  every subtree an angular **sector** proportional to its leaves, and the
1426
1407
  radius of each ring the largest of three: one `ranksep` out from the ring
@@ -1467,36 +1448,12 @@ what gets routed is a detour with a third card in its way, an edge dropping
1467
1448
  several rows, an edge between two cards of one row. Since `7.23.152` a node is
1468
1449
  measured **whole** — card, ports and label — so a line keeps clear of the
1469
1450
  ports as well; the edge's own ports stick out of its own cards, so its first
1470
- and last segments are not tested against their own box.
1471
-
1472
- The edges sharing an **end** do not lie on top of each other either
1473
- (`7.23.153`): the forward edges leaving one card — all of them, not only one
1474
- port's, or two hooks side by side would stack their runs — and those reaching
1475
- one card (all its fkeys enter by one point) are **staggered**, each turning at
1476
- its own height in the channel, 10px apart at most and closer when the channel
1477
- is short. Leaving, the farthest turns highest; arriving, the farthest turns
1478
- lowest: the orders in which the lines of one end nest without crossing. A
1479
- routed line leaves its port and reaches the other one **sideways**, because
1480
- the column straight under or over a port is where the staggered edges of that
1481
- port run. Only the stub at the port is shared. Into a lower row of a stack of
1482
- the compact tree an edge runs a **comb** (`7.23.154`): along the channel to a
1483
- lane of its own in the stack's corridor, down it, and into the gap above its
1484
- card; lanes go outermost first by row, so none crosses another, and a comb is
1485
- drawn as laid out, never routed.
1486
-
1487
- **Curved edges keep off the cards as well** (`7.23.155`). On a layout with
1488
- rows a curve is `treedb-curve-v` / `-h`, a subclass of G6's cubic: the cubic
1489
- as before, unless that very curve — the path G6 would draw, sampled, not the
1490
- rectangle round it — would cross a card other than its own two, or the edge
1491
- does not run forward and so would cross its own. Then it takes the way its
1492
- elbow would (comb, detour, route, the same lanes and staggered turns) with
1493
- corners rounded to 24px. On a layout with no rows the edges stay G6's plain
1494
- cubic. G6's `shortest-path` router is not used: it is
1495
- not exported, it
1451
+ and last segments are not tested against their own box. G6's `shortest-path`
1452
+ router is not used: it is not exported, it
1496
1453
  falls back in silence to a route that crosses cards, and it rewrites its
1497
1454
  module defaults with every config it is given.
1498
1455
 
1499
- All three take the **same spanning tree**, chosen deterministically: roots are the
1456
+ Both take the **same spanning tree**, chosen deterministically: roots are the
1500
1457
  nodes with no incoming edge, in node order; a node belongs to the **first
1501
1458
  parent that reaches it** in a breadth-first walk (the place, not the
1502
1459
  controller, because the place's column came first) and its other links are