isorite 0.1.0 → 0.2.0

package/dist/core.js

@@ -1467,6 +1467,111 @@ function authors(entries) {
 	}
 	return owners;
 }
+/** A line's words, spaces and marks, in order: what is compared to tell who wrote what within a line. */
+const tokens = (line) => line.match(/\s+|[\p{L}\p{N}_]+|[^\s\p{L}\p{N}_]/gu) ?? [];
+/** Who wrote each word of the note as the journal last has it, line by line, as runs. Where a change replaced
+*  lines, the old and the new are compared word by word, so a word kept keeps its writer and only the words put in
+*  are the changer's. Text kept from a proposal is its proposer's, as in `authors`. Null when the chain is broken. */
+function wordAuthors(entries) {
+	const proposers = /* @__PURE__ */ new Map();
+	let lines = null, text = null;
+	const line = (l, who) => {
+		const t = tokens(l);
+		return {
+			t,
+			w: t.map(() => who)
+		};
+	};
+	for (const e of entries) {
+		if (e.kind === "proposal") proposers.set(e.id, e.who);
+		if (!isChange(e)) continue;
+		const who = e.why === "keep" && e.proposal ? proposers.get(e.proposal) ?? e.who : e.who;
+		if (lines === null || text === null) {
+			if (e.text === void 0) return null;
+			text = e.text.split("\n");
+			lines = text.map((l) => line(l, who));
+			continue;
+		}
+		if (hashText(text.join("\n")) !== e.before) return null;
+		const next = [];
+		let at = 0;
+		for (const h of e.hunks) {
+			next.push(...lines.slice(at, h.at));
+			const old = lines.slice(h.at, h.at + h.del.length);
+			const a = [], aw = [];
+			old.forEach((l, k) => {
+				if (k) {
+					a.push("\n");
+					aw.push("");
+				}
+				a.push(...l.t);
+				aw.push(...l.w);
+			});
+			const b = [];
+			h.ins.forEach((l, k) => {
+				if (k) b.push("\n");
+				b.push(...tokens(l));
+			});
+			const bw = b.map(() => who);
+			const kept = [];
+			let i = 0, j = 0;
+			for (const d of diffLines(a, b)) {
+				while (i < d.at) {
+					kept.push([j, i]);
+					i++;
+					j++;
+				}
+				i += d.del.length;
+				j += d.ins.length;
+			}
+			while (i < a.length && j < b.length) {
+				kept.push([j, i]);
+				i++;
+				j++;
+			}
+			const words = (ts) => ts.filter((t) => /[\p{L}\p{N}]/u.test(t)).length;
+			if (kept.filter(([jj]) => /[\p{L}\p{N}]/u.test(b[jj])).length * 2 >= words(b)) for (const [jj, ii] of kept) bw[jj] = aw[ii];
+			let cur = {
+				t: [],
+				w: []
+			};
+			const made = [];
+			b.forEach((tok, k) => {
+				if (tok === "\n") {
+					made.push(cur);
+					cur = {
+						t: [],
+						w: []
+					};
+				} else {
+					cur.t.push(tok);
+					cur.w.push(bw[k]);
+				}
+			});
+			if (h.ins.length) made.push(cur);
+			next.push(...made);
+			at = h.at + h.del.length;
+		}
+		next.push(...lines.slice(at));
+		lines = next;
+		text = applyHunks(text, e.hunks);
+	}
+	return (lines ?? []).map((l) => {
+		const runs = [];
+		let pos = 0;
+		l.t.forEach((tok, k) => {
+			const who = l.w[k], last = runs.at(-1);
+			if (last && last.who === who && last.to === pos) last.to += tok.length;
+			else runs.push({
+				from: pos,
+				to: pos + tok.length,
+				who
+			});
+			pos += tok.length;
+		});
+		return runs;
+	});
+}
 /** The note as it was after an entry. */
 const textAt = (entries, id) => replay(entries, id)?.text ?? null;
 /** The entry for a change of the note's file to `after`, or null when nothing changed. The difference is taken
@@ -1678,7 +1783,8 @@ function moments(entries) {
 			const what = e.why === "start" ? e.who === "external" ? "found" : "wrote" : e.why === "edit" ? "edited" : e.why === "restore" ? "restored" : e.why === "keep" ? "kept" : e.why === "move" ? "moved" : "changed elsewhere";
 			out.push({
 				...base,
-				what
+				what,
+				...e.asked ? { asked: e.asked } : {}
 			});
 		}
 	}
@@ -1867,6 +1973,18 @@ function mapModel(base, study = null) {
 		edges
 	};
 }
+/** The map without the study base: its folders and notes, and the edges to them, left out. The Atlas draws study
+*  folders only in its Understand lens (M10). */
+function withoutStudy(model) {
+	const strip = (d) => ({
+		...d,
+		children: d.children.filter((c) => !c.study).map((c) => c.kind === "dir" ? strip(c) : c)
+	});
+	return {
+		root: strip(model.root),
+		edges: model.edges.filter(([a, b]) => !a.startsWith("@") && !b.startsWith("@"))
+	};
+}
 /** What lights when a note is chosen on the Atlas (refs, as the map model names notes). Where the base rates its
 *  links, only "requires" is followed back; where it rates none, every drawn link is. A study note requires what it tests. */
 function reach(model, note) {
@@ -2047,6 +2165,11 @@ function understanding(events, now = Date.now()) {
 				s.last = when;
 				s.evidence.push(e.id);
 			}
+		} else if (e.event === "understood" && typeof e.note === "string") {
+			const s = at(e.note), by = e.by;
+			s.level = clamp(s.level + (typeof by === "number" && Number.isFinite(by) ? Math.max(-1, Math.min(1, by)) : 0));
+			s.last = when;
+			s.evidence.push(e.id);
 		} else if (e.event === "moved" && typeof e.from === "string" && typeof e.to === "string") {
 			const s = state.get(e.from);
 			if (s) {
@@ -2310,4 +2433,4 @@ function ago(at, now = Date.now()) {
 /** A moment is named by its last entry: the one the note is shown at. */
 const momentKey = (m) => m.ids.at(-1);
 //#endregion
-export { Bundle, CONFIG_FILE, Concept, DEVICE_RE, DiffError, Directory, ExerciseError, FrontmatterError, GITATTRIBUTES, ID_RE, INTERVALS, KINDS, LEVELS, LearnerError, MAX_EVENT_BYTES, MAX_WHY, OKF_VERSION, RATINGS, SESSION_GAP_MS, SNAPSHOT_EVERY, STRENGTH, STUDY_DIR, STUDY_FORMAT, STUDY_PREFIX, SearchIndex, StudyError, WORKING, agentName, ago, anchorOf, applyHunks, authors, blocks, change, citeKind, cleanEvent, cmp, cmpTuple, contentHash, describe, diffLines, dumpFrontmatter, dumpStudyConfig, editFrontmatter, exercise, exerciseText, findAnchor, findPassage, floatRepr, frontmatterText, hashText, headings, impliedLinks, insertAt, isStudyId, iso, journalLine, journalPath, legacyId, linkRefs, locate, lowest, mapModel, markingReply, mergeRecords, momentKey, moments, newId, noteIds, pagerank, parseStudyConfig, parseYaml, passageOf, pending, placeMarks, pyRepr, rating, reach, readJournal, readRecord, reconcile, recordPath, replay, round3, section, sessionMoments, sha256, slugify, snippet, spliceText, splitFrontmatter, splitSource, sprout, stamp, stepOf, strengthEdges, strip, stronglyConnected, studyId, studyList, testsOf, text, textAt, toTime, tokenize, touched, understanding, withNext, written };
+export { Bundle, CONFIG_FILE, Concept, DEVICE_RE, DiffError, Directory, ExerciseError, FrontmatterError, GITATTRIBUTES, ID_RE, INTERVALS, KINDS, LEVELS, LearnerError, MAX_EVENT_BYTES, MAX_WHY, OKF_VERSION, RATINGS, SESSION_GAP_MS, SNAPSHOT_EVERY, STRENGTH, STUDY_DIR, STUDY_FORMAT, STUDY_PREFIX, SearchIndex, StudyError, WORKING, agentName, ago, anchorOf, applyHunks, authors, blocks, change, citeKind, cleanEvent, cmp, cmpTuple, contentHash, describe, diffLines, dumpFrontmatter, dumpStudyConfig, editFrontmatter, exercise, exerciseText, findAnchor, findPassage, floatRepr, frontmatterText, hashText, headings, impliedLinks, insertAt, isStudyId, iso, journalLine, journalPath, legacyId, linkRefs, locate, lowest, mapModel, markingReply, mergeRecords, momentKey, moments, newId, noteIds, pagerank, parseStudyConfig, parseYaml, passageOf, pending, placeMarks, pyRepr, rating, reach, readJournal, readRecord, reconcile, recordPath, replay, round3, section, sessionMoments, sha256, slugify, snippet, spliceText, splitFrontmatter, splitSource, sprout, stamp, stepOf, strengthEdges, strip, stronglyConnected, studyId, studyList, testsOf, text, textAt, toTime, tokenize, touched, understanding, withNext, withoutStudy, wordAuthors, written };

package/dist/embed-worker.js

@@ -1,5 +1,5 @@
 import { createRequire } from "node:module";
-import { existsSync, readFileSync } from "node:fs";
+import { existsSync, readFileSync, readdirSync } from "node:fs";
 import { dirname, join } from "node:path";
 import { fileURLToPath } from "node:url";
 import { parentPort, workerData } from "node:worker_threads";
@@ -3220,7 +3220,14 @@ const WASM = "ort-wasm-simd-threaded.wasm";
 const beside = join(dirname(fileURLToPath(import.meta.url)), WASM);
 ort.env.wasm.wasmBinary = readFileSync(existsSync(beside) ? beside : join(dirname(createRequire(import.meta.url).resolve("onnxruntime-web")), WASM));
 const tok = new Tokenizer(JSON.parse(readFileSync(join(dir, "tokenizer.json"), "utf8")), JSON.parse(readFileSync(join(dir, "tokenizer_config.json"), "utf8")));
-const session = await ort.InferenceSession.create(readFileSync(join(dir, "model.onnx")), { executionProviders: ["wasm"] });
+const externalData = readdirSync(dir).filter((f) => f.endsWith(".onnx_data")).map((path) => ({
+	path,
+	data: readFileSync(join(dir, path))
+}));
+const session = await ort.InferenceSession.create(readFileSync(join(dir, "model.onnx")), {
+	executionProviders: ["wasm"],
+	externalData
+});
 const typed = session.inputNames.includes("token_type_ids");
 async function embed(text) {
 	let ids = tok.encode(text).ids;