@yuneta/gobj-ui 7.25.18 → 7.25.20

package/README.md

@@ -227,6 +227,16 @@ seen*:
   for its treedb nodes, so the graph tab returns to `<graph>/<topic>` with its
   focus and the data tab to the topic that was open.
 
+- **A node hears every event of the navs it projects** (they are its pure
+  children). Since `7.25.19` it declares all three, in every state -- a ZONE
+  projection outlives the path, so a root that is off it still hears its
+  rail: `EV_NAV_CLICKED` (a navigation), `EV_NAV_ITEM_CLOSE` (handed to the
+  shell, which re-publishes it to the app, as for the shell's own closable
+  items) and `EV_DRAWER_CLOSE_REQUESTED` (the node closes its own drawer
+  projection: the shell's `yui_shell_close_drawer()` knows only the shell's
+  navs). Before, the last two answered *"Event NOT DEFINED in state"*, and so
+  did a click on a zone nav of a root in `ST_OFF`.
+
 - **`nav_mode` — the three shapes as one runtime knob.** The two bullets above
   describe what a tree *declares*; `nav_mode` is how a user *chooses* between
   the shapes without the app rewriting anything:
@@ -576,13 +586,33 @@ Two layout facts the browser taught this component, both worth keeping:
   `"graph"`), `$container` (mounted by the parent).
 - Input events: `EV_SET_JSON {json}` (replace the whole document; `ST_EMPTY` →
   `ST_READY`), `EV_SUBTREE_LOADED {path, json}` (splice a fetched subtree),
-  `EV_SUBTREE_ERROR {path, error}`, `EV_SET_VIEW_MODE {mode}` (`"tree"` /
+  `EV_SUBTREE_ERROR {path, error | i18n, by_design}`, `EV_SET_VIEW_MODE {mode}` (`"tree"` /
   `"text"` / `"graph"`; **no mode advances** to the next view, which is what
   the two-view toggle did when the list was two long), plus `EV_REFRESH` /
   `EV_SHOW` / `EV_HIDE` / `EV_LANGUAGE_CHANGED`.
 - Output event: `EV_EXPAND_PATH {path, size}` (`EVF_OUTPUT_EVENT`) — the parent
-  must declare it in its own FSM (CHILD subscription model). That is the ONLY
-  one, on purpose: this viewer is a child of its host and subscribes it to
+  must declare it in its own FSM (CHILD subscription model), and ANSWER it: the
+  stub shows "loading" until `EV_SUBTREE_LOADED` or `EV_SUBTREE_ERROR` arrives.
+  A host that cannot read a subtree says so with the error, as
+  `C_YUI_TREEDB_TOPIC_WITH_FORM` does for its schema, cell and table viewers
+  since `7.25.19` (before, it declared nothing: *"Event NOT DEFINED"*):
+  ```js
+  gobj_send_event(src, "EV_SUBTREE_ERROR", {
+      path:      kw.path,
+      i18n:      "this part cannot be loaded here",  // a KEY: t(key) + data-i18n
+      by_design: true     // not a failure: logged as a warning
+  }, gobj);
+  ```
+  `error` is text, drawn as it came (a backend's comment); `i18n` is a key,
+  so the stub changes language with the app -- a text translated by the host
+  stays in the language it was sent in. The name `i18n` is the one a
+  `validate-locales` scan looks for, so the key stays visible to it. `by_design`
+  says the host CANNOT, which is not a failure: the viewer logs it as a
+  WARNING, and anything else as an ERROR. Before `7.25.20` the form sent
+  `t(...)`, and every click on a stub logged an ERROR, which broke a
+  clean-console check and reached the backend through the remote log.
+
+  `EV_EXPAND_PATH` is the ONLY output event, on purpose: this viewer is a child of its host and subscribes it to
   everything it publishes, so every output event is a mandatory declaration in
   every host's FSM. The graph child's `EV_JSON_ITEM_CLICKED` stops here
   (`7.21.0` forwarded it and broke exactly that way); a host that wants node
@@ -1003,7 +1033,7 @@ graph does not report `__graphs__`, its own bookkeeping.
     treedb_name: "treedb_test",           // the view's own treedb
     topic_name:  "yunos",                 // the topic written (the CHILD's, for a link)
     record:      {id: "1", x: 1, y: 2},   // the node the store answered, as written
-    created:     false,                   // true for create-node
+    created:     false,                   // true for a NEW record (see below)
     command:     "update-node",           // create/update/delete-node, link/unlink-nodes
     parent_ref:  "realms^r1^yunos",       // graph, link/unlink only
     child_ref:   "yunos^1"                // graph, link/unlink only
@@ -1018,6 +1048,21 @@ do -- so `treedb_name` and `record` arrived empty, and a link named no topic.
 delete answers the node deleted, a link the child), and a link echoes its topic
 and refs in `__md_command__`.
 
+Since `7.25.19` the topics view publishes a successful **delete** too (it
+published nothing: the answer went to an empty `break`), and `created` is
+TRUE for its **+New**: that goes out as an `update-node` with `create_only`,
+so comparing the command with `"create-node"` answered `false` for every
+record the table created. The request now echoes `created: true` in its
+`__md_command__`. The graph creates with `create-node`, and says so the same
+way.
+
+```js
+// the table's +New, answered
+{treedb_name: "treedb_test", topic_name: "users", record: {id: "n1"}, created: true,  command: "update-node"}
+// a row deleted, answered with the node deleted
+{treedb_name: "treedb_test", topic_name: "roles", record: {id: "x"},  created: false, command: "delete-node"}
+```
+
 ### Read-only treedbs: `readonly`
 
 `C_YUI_TREEDB_TOPICS` and `C_YUI_TREEDB_GRAPH` take a **`readonly`** attr; the
@@ -1844,8 +1889,12 @@ the history buttons -- they decide the Save from `history.canUndo()`, and they
 run on every history change, every change of mode and every redraw of the
 theme -- and a refusal heard in reading was never shown. The engine now keeps
 the topic as owed: the Save stays lit while one is, in edition and on entering
-it; the Save that writes the topic again pays it, and a reload of the data
-forgets it (the view holds what the backend holds).
+it; the Save that writes the topic again pays it, a reload of the data
+forgets it (the view holds what the backend holds), and so does an ECHO of the
+topic's `__graphs__` record (since `7.25.19`): the echo is what the backend
+holds, taken as shown and as saved, so nothing is owed any more. Before, a
+refused Save #1 answered after Save #2's echo -- or another browser saving the
+topic -- left the Save lit with nothing to write until a reload.
 
 ```js
 gobj_send_event(engine, "EV_GRAPHS_WRITE_REFUSED", {topic: "users"}, host);  // Save lit
@@ -1863,10 +1912,34 @@ worked only in a test whose fake transport echoed the whole request. The host
 now puts the topic in the echo it asks for:
 
 ```js
-kw.__md_command__ = {topic_name: "__graphs__", graph_topic: "users"};
+kw.__md_command__ = {topic_name: "__graphs__", graph_topic: "users", graphs_load: 3};
 // the answer: __md_iev__.command_stack[0].kw === that echo, and nothing else
 ```
 
+**A refusal belongs to the LOAD its write was sent from** (since `7.25.20`).
+"A reload forgets it" held only for a refusal that arrived BEFORE the reload:
+one answered after it was applied to the fresh load -- Save, Refresh, then the
+refusal lands -- and lit Save over nothing changed, and pressing it wrote a
+record identical to what the backend holds. The engine now counts its loads
+(each `EV_CLEAR_DATA` starts one) and puts the number in every `EV_UPDATE_NODE`
+of `__graphs__` as `graphs_load`; the host echoes it in `__md_command__` and
+hands it back with `EV_GRAPHS_WRITE_REFUSED {topic, graphs_load}`. A refusal of
+an earlier load is logged as a warning and ignored. A host that sends no
+`graphs_load` is taken at its word, as before:
+
+```js
+// in the host: the engine's load, from the EV_UPDATE_NODE it published
+gobj_send_event(engine, "EV_GRAPHS_WRITE_REFUSED",
+    {topic: record.topic, graphs_load: kw.graphs_load}, gobj);
+```
+
+**A `__graphs__` create echo moves only its own topic** (since `7.25.20`), as
+an update echo has since `7.25.19` (`apply_graphs_echo()`). The first record of
+a topic's arrangement -- from this browser or another -- rebuilt the saved
+copies of EVERY topic from the live objects the view arranges in place, so an
+unsaved change of another topic (a look reset in edition, say) counted as saved
+and the next Save did not write it.
+
 **Every saved look has a way back, in the context menu** (since `7.23.134`,
 replacing `reset sizes` / `reset topic sizes`). In edition, the node, the port
 and the edge menus each offer three resets — this one, its kind (`reset topic
@@ -2344,7 +2417,13 @@ else.
   consumer's locales carry). A write that ends with an error -- a drop
   included -- still applies a position that waited, as a load does. Out of
   session a load is not asked at all (no *"not in session"* errors); the
-  reconnect loads.
+  reconnect loads. **Nor is a write** (7.25.20): each one of the queue asks
+  first, and out of session the writes end with `cannot reach the treedb`, a
+  warning, and a reload owed to the next session. A direct `C_IEVENT_CLI`
+  answers a command out of session with `null`, the answer of a request that
+  left, so before this the editor sat in `ST_SAVING`, inert, until another
+  drop. (gui_agent's `C_AGENT_TREEDB_LINK` answers an error string there, so
+  it was not hit.)
 - **A dialog is built on ONE model, and a load that replaces it closes it**
   (7.25.9). Every load that lands bumps `model_gen`; a dialog records the one
   it was opened on, and its Save, its import Preview/Run, an orphan delete and

package/dist/gobj-ui.cjs.js

@@ -10901,7 +10901,7 @@ function register_c_yui_map() {
 *
 *          Graphics, Charts with uPlot
 *
-*          Copyright (c) 2025, ArtGins.
+*          Copyright (c) 2025-2026, ArtGins.
 *          All Rights Reserved.
 ***********************************************************************/
 /***************************************************************
@@ -11151,8 +11151,8 @@ function ac_add_serie(gobj, event, kw, src) {
 	let colors = (0, _yuneta_gobj_js.gobj_read_attr)(gobj, "colors");
 	if (!Array.isArray(colors) || !colors.length) colors = CHART_PALETTE;
 	let color = colors[Math.max(0, idx - 1) % colors.length];
-	(0, _yuneta_gobj_js.kw_get_str)(gobj, kw, "stroke", color.stroke, _yuneta_gobj_js.kw_flag_t.KW_CREATE);
-	(0, _yuneta_gobj_js.kw_get_str)(gobj, kw, "fill", color.fill, _yuneta_gobj_js.kw_flag_t.KW_CREATE);
+	if (kw.stroke === void 0) kw.stroke = color.stroke;
+	if (kw.fill === void 0) kw.fill = color.fill;
 	uplot$4.addSeries(kw);
 	return 0;
 }
@@ -14743,10 +14743,18 @@ function push_collapsed_row(ctx, value, segments, depth, key, rows) {
 		},
 		is_pending ? "loading" : "click to load"
 	]);
-	if (err) children.push([
+	if (err && err.is_key) children.push([
+		"span",
+		{
+			class: "JSON_STUB_ERR has-text-danger is-size-7 ml-2",
+			"data-i18n": err.error
+		},
+		(0, i18next.t)(err.error)
+	]);
+	else if (err) children.push([
 		"span",
 		{ class: "JSON_STUB_ERR has-text-danger is-size-7 ml-2" },
-		String(err)
+		String(err.error)
 	]);
 	let attrs = {
 		class: "JSON_ROW JSON_COLLAPSED" + (is_pending ? " is-pending" : ""),
@@ -14963,14 +14971,32 @@ function ac_subtree_loaded(gobj, event, kw, src) {
 	return 0;
 }
 /************************************************************
-*   EV_SUBTREE_ERROR { path, error } — mark the failed branch
+*   EV_SUBTREE_ERROR { path, error | i18n, by_design }
+*   — mark the branch that was not loaded.
+*
+*   `error` is text, drawn as it came (a backend's comment).
+*   `i18n` is a KEY, drawn as t(key) with data-i18n so the stub
+*   changes language with the app (a text translated by the host
+*   stays in the language it was sent in). The name is the one a
+*   locale check scans for (`i18n: "<key>"`), so the key of every
+*   host stays visible to it.
+*   `by_design`: the host answers that it CANNOT read a subtree
+*   (a form holds records, a pasted pad has no backend) -- that is
+*   not a failure, and it is logged as a warning. Anything else is
+*   a load that failed, logged as an error.
 ************************************************************/
 function ac_subtree_error(gobj, event, kw, src) {
 	let priv = gobj.priv;
 	let path = kw.path || "";
+	let is_key = !!kw.i18n;
+	let error = kw.i18n || kw.error || "error";
 	priv.pending.delete(path);
-	priv.errors.set(path, kw.error || "error");
-	(0, _yuneta_gobj_js.log_error)(`${GCLASS_NAME$13}: subtree load failed at '${path}': ${kw.error || ""}`);
+	priv.errors.set(path, {
+		error,
+		is_key
+	});
+	if (kw.by_design) (0, _yuneta_gobj_js.log_warning)(`${GCLASS_NAME$13}: subtree not loaded at '${path}': ${error}`);
+	else (0, _yuneta_gobj_js.log_error)(`${GCLASS_NAME$13}: subtree load failed at '${path}': ${error}`);
 	render_view$1(gobj);
 	return 0;
 }
@@ -16020,10 +16046,11 @@ function ac_hide_diff(gobj, event, kw, src) {
 *  document has no backend to ask for it: the SOURCE truncated
 *  it, and that is the answer, to the viewer that asked.
 ***************************************************************/
-function ac_expand_path(gobj, event, kw, src) {
+function ac_expand_path$1(gobj, event, kw, src) {
 	(0, _yuneta_gobj_js.gobj_send_event)(src, "EV_SUBTREE_ERROR", {
 		path: kw && kw.path || "",
-		error: (0, i18next.t)("collapsed in the source")
+		i18n: "collapsed in the source",
+		by_design: true
 	}, gobj);
 	return 0;
 }
@@ -16081,7 +16108,7 @@ function create_gclass$12(gclass_name) {
 			],
 			[
 				"EV_EXPAND_PATH",
-				ac_expand_path,
+				ac_expand_path$1,
 				null
 			]
 		]],
@@ -16118,7 +16145,7 @@ function create_gclass$12(gclass_name) {
 			],
 			[
 				"EV_EXPAND_PATH",
-				ac_expand_path,
+				ac_expand_path$1,
 				null
 			]
 		]],
@@ -24937,7 +24964,7 @@ function answered_record$1(data) {
 /************************************************************
 *  Command to remote service
 ************************************************************/
-function treedb_update_node$1(gobj, treedb_name, topic_name, record, options, form_write) {
+function treedb_update_node$1(gobj, treedb_name, topic_name, record, options, form_write, created) {
 	let command = "update-node";
 	let gobj_remote_yuno = (0, _yuneta_gobj_js.gobj_read_pointer_attr)(gobj, "gobj_remote_yuno");
 	let kw = {
@@ -24949,6 +24976,7 @@ function treedb_update_node$1(gobj, treedb_name, topic_name, record, options, fo
 	};
 	kw.__md_command__ = { topic_name };
 	if (form_write) kw.__md_command__.form_write = form_write;
+	if (created) kw.__md_command__.created = true;
 	let ret = (0, _yuneta_gobj_js.gobj_command)(gobj_remote_yuno, command, kw, gobj);
 	if (ret) {
 		(0, _yuneta_gobj_js.log_error)(ret);
@@ -24972,7 +25000,11 @@ function treedb_delete_node$1(gobj, treedb_name, topic_name, record, options) {
 	};
 	kw.__md_command__ = { topic_name };
 	let ret = (0, _yuneta_gobj_js.gobj_command)(gobj_remote_yuno, command, kw, gobj);
-	if (ret) (0, _yuneta_gobj_js.log_error)(ret);
+	if (ret) {
+		(0, _yuneta_gobj_js.log_error)(ret);
+		return -1;
+	}
+	return 0;
 }
 /************************************************************
 *  True on a phone-width viewport (Bulma's mobile breakpoint).
@@ -25114,7 +25146,7 @@ function request_print_tranger$1(gobj, path) {
 		let jv = priv.json_gobj;
 		if (path && jv && (0, _yuneta_gobj_js.is_gobj)(jv) && !(0, _yuneta_gobj_js.gobj_is_destroying)(jv)) (0, _yuneta_gobj_js.gobj_send_event)(jv, "EV_SUBTREE_ERROR", {
 			path,
-			error: (0, i18next.t)("no session")
+			i18n: "no session"
 		}, gobj);
 		return;
 	}
@@ -25255,17 +25287,18 @@ function ac_mt_command_answer$2(gobj, event, kw, src) {
 			break;
 		case "create-node":
 		case "update-node":
+		case "delete-node":
 			(0, _yuneta_gobj_js.gobj_publish_event)(gobj, "EV_RECORD_WRITTEN", {
 				treedb_name: (0, _yuneta_gobj_js.gobj_read_str_attr)(gobj, "treedb_name"),
 				topic_name: (0, _yuneta_gobj_js.kw_get_str)(gobj, kw_command, "topic_name", "", 0),
 				record: answered_record$1(data),
-				created: command === "create-node",
+				created: command === "create-node" || kw_command.created === true,
 				command
 			});
+			if (command === "delete-node") break;
 			settle_form_write(gobj.priv.form_writes, (0, _yuneta_gobj_js.kw_get_int)(gobj, kw_command, "form_write", 0, 0), (0, _yuneta_gobj_js.kw_get_str)(gobj, kw_command, "topic_name", "", 0));
 			answer_form_write(gobj, get_gobj_formtable(gobj, (0, _yuneta_gobj_js.kw_get_str)(gobj, kw_command, "topic_name", "", 0)), (0, _yuneta_gobj_js.kw_get_int)(gobj, kw_command, "form_write", 0, 0), true);
 			break;
-		case "delete-node": break;
 		default: (0, _yuneta_gobj_js.log_error)(`${(0, _yuneta_gobj_js.gobj_short_name)(gobj)} Command unknown: ${command}`);
 	}
 	return 0;
@@ -25533,7 +25566,7 @@ function ac_create_record(gobj, event, kw, src) {
 		create: true,
 		create_only: true,
 		autolink: true
-	}, form_write) < 0) {
+	}, form_write, true) < 0) {
 		answer_form_write(gobj, src, form_write, false);
 		return -1;
 	}
@@ -29451,6 +29484,25 @@ function refuse_if_readonly$2(gobj, event) {
 *  From external, at the beginning, load all topic data
 ************************************************************/
 /***************************************************************
+*  One of MY json viewers (the schema, a cell, the table's records)
+*  met a `__collapsed__` sentinel and the reader opened it. The
+*  viewer never fetches: it asks its host, and waits on "loading"
+*  until it is answered. This view holds records, not a store it can
+*  read a subtree of -- the lazy drill is `print-tranger`, which the
+*  TOPICS view runs for its own viewer, over a whole tranger -- so the
+*  answer is that it cannot, and the viewer shows that on the stub
+*  instead of loading for ever.
+***************************************************************/
+function ac_expand_path(gobj, event, kw, src) {
+	let path = kw && kw.path || "";
+	(0, _yuneta_gobj_js.gobj_send_event)(src, "EV_SUBTREE_ERROR", {
+		path,
+		i18n: "this part cannot be loaded here",
+		by_design: true
+	}, gobj);
+	return 0;
+}
+/***************************************************************
 *  The language changed (the shell publishes it after the app switched).
 *
 *  The column headers carry their key (col_label's title formatter emits a
@@ -30744,6 +30796,11 @@ function create_gclass$5(gclass_name) {
 			ac_show_schema,
 			null
 		],
+		[
+			"EV_EXPAND_PATH",
+			ac_expand_path,
+			null
+		],
 		[
 			"EV_CELL_EDITED",
 			ac_cell_edited,
@@ -30859,6 +30916,7 @@ function create_gclass$5(gclass_name) {
 		["EV_CHANGE_LOCALE", 0],
 		["EV_REFRESH", 0],
 		["EV_SHOW_SCHEMA", 0],
+		["EV_EXPAND_PATH", 0],
 		["EV_CELL_EDITED", 0],
 		["EV_REPULL_PAGE", 0],
 		["EV_PAGE_LOADED", 0],
@@ -59264,6 +59322,7 @@ var PRIVATE_DATA$4 = {
 	_graph_properties: {},
 	_saved_graph_properties: {},
 	_graphs_writes_owed: {},
+	_graphs_load: 0,
 	yet_showed: false,
 	edit_mode: false,
 	operation_mode: null,
@@ -60973,7 +61032,8 @@ function save_topic_graph_properties(gobj, topic_name) {
 			list_dict: true,
 			autolink: false,
 			create: true
-		}
+		},
+		graphs_load: priv._graphs_load
 	};
 	(0, _yuneta_gobj_js.gobj_publish_event)(gobj, "EV_UPDATE_NODE", kw_update);
 }
@@ -65754,6 +65814,7 @@ function ac_clear_data(gobj, event, kw, src) {
 	hide_create_popover(gobj);
 	(0, _yuneta_gobj_js.gobj_write_attr)(gobj, "records", {});
 	priv._graphs_writes_owed = {};
+	priv._graphs_load++;
 	priv._nodes_total = 0;
 	priv._nodes_placed = 0;
 	priv._fold_model = null;
@@ -65855,12 +65916,19 @@ function ac_save_graph(gobj, event, kw, src) {
 /************************************************************
 *  The host says a write of `__graphs__` did not land: the
 *  backend refused it, the transport refused it on the way out,
-*  or there was no session to send it on. kw: {topic}
+*  or there was no session to send it on.
+*  kw: {topic, graphs_load}
 *
 *  save_topic_graph_properties() recorded it as saved before it
 *  left. Forgotten now, and the Save is lit again: the next Save
 *  writes the topic again. Before, the view believed the backend
 *  held the arrangement, and the next Save had nothing to write.
+*
+*  `graphs_load` is the load the write was sent from, echoed by the
+*  host. A refusal of an EARLIER load is about an arrangement the
+*  reload replaced with what the backend holds: applied, it lit the
+*  Save over nothing changed (before gobj-ui 7.25.20). A host that
+*  echoes no load is taken at its word, as before.
 ************************************************************/
 function ac_graphs_write_refused(gobj, event, kw, src) {
 	let priv = gobj.priv;
@@ -65869,6 +65937,10 @@ function ac_graphs_write_refused(gobj, event, kw, src) {
 		(0, _yuneta_gobj_js.log_error)(`${(0, _yuneta_gobj_js.gobj_short_name)(gobj)}: ${event} names no topic`);
 		return -1;
 	}
+	if (kw.graphs_load !== void 0 && kw.graphs_load !== priv._graphs_load) {
+		(0, _yuneta_gobj_js.log_warning)(`${(0, _yuneta_gobj_js.gobj_short_name)(gobj)}: a refused write of the arrangement of '${topic_name}' belongs to an earlier load: ignored`);
+		return 0;
+	}
 	(0, _yuneta_gobj_js.log_warning)(`${(0, _yuneta_gobj_js.gobj_short_name)(gobj)}: the arrangement of '${topic_name}' was not saved: the next Save writes it again`);
 	forget_refused_graphs_write(priv._saved_graph_properties, topic_name);
 	priv._graphs_writes_owed[topic_name] = true;
@@ -65876,6 +65948,22 @@ function ac_graphs_write_refused(gobj, event, kw, src) {
 	return 0;
 }
 /************************************************************
+*  An echo of a `__graphs__` record is what the backend holds for
+*  its topic now, and the view has just taken it as shown and as
+*  saved: a write of that topic owed from an earlier refusal is
+*  owed no more. Left owed, the Save stayed lit with nothing to
+*  write -- the plan compares the two, and they are equal -- until
+*  a reload: a refused Save #1 answered after Save #2 landed, or
+*  another browser saving the topic, did exactly that.
+************************************************************/
+function settle_owed_graphs_write(gobj, rec) {
+	let priv = gobj.priv;
+	let topic_name = rec && rec.topic;
+	if (!topic_name || !priv._graphs_writes_owed[topic_name]) return;
+	delete priv._graphs_writes_owed[topic_name];
+	update_history_buttons(gobj);
+}
+/************************************************************
 *  Node created, from subscription
 ************************************************************/
 function ac_node_created(gobj, event, kw, src) {
@@ -65886,7 +65974,8 @@ function ac_node_created(gobj, event, kw, src) {
 	if (!priv.descs || !priv.graph) return 0;
 	if (topic_name === "__graphs__") {
 		priv.__graphs__.push(node);
-		build_graph_properties(gobj);
+		apply_graphs_echo(priv._graph_properties, priv._saved_graph_properties, priv.__graphs__, node);
+		settle_owed_graphs_write(gobj, node);
 		return 0;
 	}
 	if (!priv.descs[topic_name]) {
@@ -65922,6 +66011,7 @@ function ac_node_updated(gobj, event, kw, src) {
 		}
 		if (!found) priv.__graphs__.push(node);
 		apply_graphs_echo(priv._graph_properties, priv._saved_graph_properties, priv.__graphs__, node);
+		settle_owed_graphs_write(gobj, node);
 		return 0;
 	}
 	let desc = priv.descs[topic_name];
@@ -68265,7 +68355,7 @@ function request_print_tranger(gobj, path) {
 		let jv = priv.json_gobj;
 		if (path && jv && (0, _yuneta_gobj_js.is_gobj)(jv) && !(0, _yuneta_gobj_js.gobj_is_destroying)(jv)) (0, _yuneta_gobj_js.gobj_send_event)(jv, "EV_SUBTREE_ERROR", {
 			path,
-			error: (0, i18next.t)("no session")
+			i18n: "no session"
 		}, gobj);
 		return;
 	}
@@ -68374,7 +68464,7 @@ function treedb_create_node(gobj, treedb_name, topic_name, record, options) {
 /************************************************************
 *  Command to remote service
 ************************************************************/
-function treedb_update_node(gobj, treedb_name, topic_name, record, options) {
+function treedb_update_node(gobj, treedb_name, topic_name, record, options, graphs_load) {
 	let priv = gobj.priv;
 	if (!priv.gobj_remote_yuno) {
 		(0, _yuneta_gobj_js.log_error)(`${(0, _yuneta_gobj_js.gobj_short_name)(gobj)}: No gobj_remote_yuno defined`);
@@ -68389,7 +68479,10 @@ function treedb_update_node(gobj, treedb_name, topic_name, record, options) {
 		options: options || {}
 	};
 	kw.__md_command__ = { topic_name };
-	if (topic_name === "__graphs__") kw.__md_command__.graph_topic = graph_topic_of(record);
+	if (topic_name === "__graphs__") {
+		kw.__md_command__.graph_topic = graph_topic_of(record);
+		if (graphs_load !== void 0) kw.__md_command__.graphs_load = graphs_load;
+	}
 	let ret = (0, _yuneta_gobj_js.gobj_command)(priv.gobj_remote_yuno, command, kw, gobj);
 	if (ret) {
 		(0, _yuneta_gobj_js.log_error)(ret);
@@ -68693,7 +68786,7 @@ function ac_mt_command_answer$1(gobj, event, kw, src) {
 		case "delete-node":
 		case "link-nodes":
 		case "unlink-nodes":
-			if (result < 0 && command === "update-node") graphs_write_refused(gobj, (0, _yuneta_gobj_js.kw_get_str)(gobj, kw_command, "topic_name", "", 0), (0, _yuneta_gobj_js.kw_get_str)(gobj, kw_command, "graph_topic", "", 0));
+			if (result < 0 && command === "update-node") graphs_write_refused(gobj, (0, _yuneta_gobj_js.kw_get_str)(gobj, kw_command, "topic_name", "", 0), (0, _yuneta_gobj_js.kw_get_str)(gobj, kw_command, "graph_topic", "", 0), kw_command ? kw_command.graphs_load : void 0);
 			if (result >= 0 && (0, _yuneta_gobj_js.kw_get_str)(gobj, kw_command, "topic_name", "", 0) !== "__graphs__") {
 				let written = {
 					treedb_name: priv.treedb_name,
@@ -68948,17 +69041,18 @@ function ac_update_node(gobj, event, kw, src) {
 	let topic_name = kw.topic_name;
 	let record = kw.record;
 	let options = kw.options || {};
+	let graphs_load = kw.graphs_load;
 	if (refuse_if_readonly$1(gobj, event)) {
-		graphs_write_refused(gobj, topic_name, graph_topic_of(record));
+		graphs_write_refused(gobj, topic_name, graph_topic_of(record), graphs_load);
 		return -1;
 	}
 	if (topic_name === "__graphs__" && !is_connected(gobj)) {
 		(0, _yuneta_gobj_js.log_warning)(`${(0, _yuneta_gobj_js.gobj_short_name)(gobj)}: no session, the arrangement was not sent`);
-		graphs_write_refused(gobj, topic_name, graph_topic_of(record));
+		graphs_write_refused(gobj, topic_name, graph_topic_of(record), graphs_load);
 		return -1;
 	}
-	let ret = treedb_update_node(gobj, treedb_name, topic_name, record, options);
-	if (ret < 0) graphs_write_refused(gobj, topic_name, graph_topic_of(record));
+	let ret = treedb_update_node(gobj, treedb_name, topic_name, record, options, graphs_load);
+	if (ret < 0) graphs_write_refused(gobj, topic_name, graph_topic_of(record), graphs_load);
 	return ret;
 }
 /************************************************************
@@ -68973,9 +69067,11 @@ function graph_topic_of(record) {
 *  (C_G6_NODES_TREE) took it for granted when it left -- nothing
 *  answers it when it lands -- so it is told, and its next Save
 *  writes the topic again. Other topics are the operator's data:
-*  the answer's error says what happened to them.
+*  the answer's error says what happened to them. `graphs_load` is
+*  the engine's load the write was sent from, handed back so it can
+*  tell a refusal of an earlier load from one of the current.
 ************************************************************/
-function graphs_write_refused(gobj, topic_name, topic) {
+function graphs_write_refused(gobj, topic_name, topic, graphs_load) {
 	let priv = gobj.priv;
 	if (topic_name !== "__graphs__") return;
 	if (!topic) {
@@ -68983,7 +69079,9 @@ function graphs_write_refused(gobj, topic_name, topic) {
 		return;
 	}
 	if (!priv.gobj_nodes_tree) return;
-	(0, _yuneta_gobj_js.gobj_send_event)(priv.gobj_nodes_tree, "EV_GRAPHS_WRITE_REFUSED", { topic }, gobj);
+	let kw = { topic };
+	if (graphs_load !== void 0) kw.graphs_load = graphs_load;
+	(0, _yuneta_gobj_js.gobj_send_event)(priv.gobj_nodes_tree, "EV_GRAPHS_WRITE_REFUSED", kw, gobj);
 }
 /********************************************
 *  Message from G6_nodes_tree
@@ -72956,8 +73054,12 @@ function current_topic(gobj) {
 *  One line for what the whole view has to say: loading, empty,
 *  or the reason nothing loaded. Carries its i18n key, so it
 *  follows a language change like any other text.
+*
+*  `detail` is a NAME (a treedb, a topic) shown as it is, unless
+*  `detail_is_key` says it is a key: a name is not text to
+*  translate, and a topic called `nodes` is not "Nodos".
 ***************************************************************/
-function show_notice(gobj, key, detail) {
+function show_notice(gobj, key, detail, detail_is_key) {
 	let $notice = $of(gobj, ".SCHEMA_NOTICE");
 	if (!$notice) return;
 	clear($notice);
@@ -72974,14 +73076,15 @@ function show_notice(gobj, key, detail) {
 		},
 		(0, i18next.t)(key)
 	]));
-	if (detail && detail !== key) $notice.appendChild((0, _yuneta_gobj_js.createElement2)([
-		"span",
-		{
-			class: "SCHEMA_NOTICE_DETAIL yui-text-quiet ml-2",
-			i18n: `${detail}`
-		},
-		(0, i18next.t)(`${detail}`)
-	]));
+	if (detail && detail !== key) {
+		let attrs = { class: "SCHEMA_NOTICE_DETAIL yui-text-quiet ml-2" };
+		if (detail_is_key) attrs.i18n = `${detail}`;
+		$notice.appendChild((0, _yuneta_gobj_js.createElement2)([
+			"span",
+			attrs,
+			detail_is_key ? (0, i18next.t)(`${detail}`) : `${detail}`
+		]));
+	}
 }
 /***************************************************************
 *  A button of the toolbar. Every one carries an icon and a
@@ -73168,7 +73271,7 @@ function render(gobj) {
 		return;
 	}
 	if (state === "ST_IDLE") {
-		show_notice(gobj, priv.load_error ? "cannot load the schemas" : "not connected", priv.load_error);
+		show_notice(gobj, priv.load_error ? "cannot load the schemas" : "not connected", priv.load_error, true);
 		destroy_diagram(gobj);
 		clear($body);
 		return;
@@ -75169,13 +75272,18 @@ function run_next_write(gobj) {
 	let write = priv.save_queue[0];
 	let command = write_command(write.op);
 	let options = write_options(write.op);
+	if (!transport_in_session(gobj)) {
+		(0, _yuneta_gobj_js.log_warning)(`${(0, _yuneta_gobj_js.gobj_short_name)(gobj)}: no session, '${command}' of '${write.topic_name}' was not sent`);
+		priv.reload_on_open = true;
+		return end_writes(gobj, "cannot reach the treedb");
+	}
 	priv.write_tag++;
 	if (remote_command(gobj, command, {
 		topic_name: write.topic_name,
 		record: write.record,
 		options,
 		record_id: write.record.id || ""
-	}, { write: priv.write_tag }) < 0) return end_writes(gobj, (0, i18next.t)("cannot reach the treedb"));
+	}, { write: priv.write_tag }) < 0) return end_writes(gobj, "cannot reach the treedb");
 	return 0;
 }
 /***************************************************************