isorite 0.9.1 → 0.11.0

package/dist/core.js

@@ -106,7 +106,7 @@ function normpath(p) {
 	}
 	return "/".repeat(initial) + out.join("/") || ".";
 }
-/** SHA-256 as hex, synchronous and dependency-free, so the core runs anywhere. */
+/** SHA-256 as hex, synchronous and dependency-free, so the core runs anywhere: of text (as UTF-8), or of bytes. */
 function sha256(message) {
 	const K = new Uint32Array([
 		1116352408,
@@ -174,7 +174,7 @@ function sha256(message) {
 		3204031479,
 		3329325298
 	]);
-	const data = new TextEncoder().encode(message);
+	const data = typeof message === "string" ? new TextEncoder().encode(message) : message;
 	const len = data.length;
 	const padded = new Uint8Array(len + 9 + 63 >> 6 << 6);
 	padded.set(data);
@@ -366,7 +366,7 @@ function linkTexts(body) {
 	return out;
 }
 /** Every link and image in a body, in order: where it points, its title, and whether it is an image
-*  (`![…](…)`, which for an artifact or a picture means "shown here", an embed). */
+*  (`![…](…)`, which for an artifact or a picture means "shown here", an embed), with an image's alt text. */
 function citeRefs(body) {
 	const { tokens } = parse$1(body);
 	const out = [];
@@ -380,7 +380,8 @@ function citeRefs(body) {
 		else if (c.type === "image") out.push({
 			href: c.attrGet("src") ?? "",
 			title: c.attrGet("title"),
-			embed: true
+			embed: true,
+			alt: c.content
 		});
 	}
 	return out;
@@ -424,6 +425,94 @@ function section(body, heading) {
 	return null;
 }
 //#endregion
+//#region packages/core/src/picture.ts
+const TYPES = {
+	png: "image/png",
+	jpg: "image/jpeg",
+	jpeg: "image/jpeg",
+	gif: "image/gif",
+	svg: "image/svg+xml",
+	webp: "image/webp",
+	avif: "image/avif"
+};
+/** A picture's media type, by its extension, or null when the path isn't a picture's (M22). */
+const pictureType = (path) => {
+	const dot = path.lastIndexOf(".");
+	return dot < 0 || path.lastIndexOf("/") > dot ? null : TYPES[path.slice(dot + 1).toLowerCase()] ?? null;
+};
+const OTHER = {
+	html: "text/html",
+	htm: "text/html",
+	csv: "text/csv",
+	tsv: "text/tab-separated-values",
+	json: "application/json",
+	jsonl: "application/jsonl",
+	ndjson: "application/x-ndjson",
+	txt: "text/plain",
+	tex: "text/plain",
+	pdf: "application/pdf",
+	glb: "model/gltf-binary",
+	gltf: "model/gltf+json",
+	stl: "model/stl",
+	obj: "model/obj",
+	zip: "application/zip",
+	parquet: "application/vnd.apache.parquet",
+	arrow: "application/vnd.apache.arrow.file",
+	npy: "application/octet-stream",
+	xlsx: "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
+	docx: "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
+	epub: "application/epub+zip",
+	odt: "application/vnd.oasis.opendocument.text"
+};
+/** Any file's media type in the notes (M23), by its extension: a picture's, one listed, or octet-stream. Null for a
+*  note (Markdown is read as a note, never as a file) and for a path with no extension. */
+const fileType = (path) => {
+	const dot = path.lastIndexOf(".");
+	if (dot < 0 || path.lastIndexOf("/") > dot) return null;
+	const ext = path.slice(dot + 1).toLowerCase();
+	return ext === "md" ? null : pictureType(path) ?? OTHER[ext] ?? "application/octet-stream";
+};
+/** Width and height in pixels, or null. */
+function pictureSize(bytes, type) {
+	const v = new DataView(bytes.buffer, bytes.byteOffset, bytes.byteLength), n = bytes.length;
+	const at = (i, s) => [...s].every((c, k) => bytes[i + k] === c.charCodeAt(0));
+	const ok = (w, h) => w > 0 && h > 0 ? {
+		width: w,
+		height: h
+	} : null;
+	try {
+		if (type === "image/png" && n >= 24 && at(12, "IHDR")) return ok(v.getUint32(16), v.getUint32(20));
+		if (type === "image/gif" && n >= 10 && at(0, "GIF")) return ok(v.getUint16(6, true), v.getUint16(8, true));
+		if (type === "image/jpeg") {
+			for (let i = 2; i + 9 < n;) {
+				if (bytes[i] !== 255) return null;
+				const m = bytes[i + 1];
+				if (m >= 192 && m <= 207 && m !== 196 && m !== 200 && m !== 204) return ok(v.getUint16(i + 7), v.getUint16(i + 5));
+				i += 2 + v.getUint16(i + 2);
+			}
+			return null;
+		}
+		if (type === "image/webp" && n >= 30 && at(0, "RIFF") && at(8, "WEBP")) {
+			if (at(12, "VP8 ")) return ok(v.getUint16(26, true) & 16383, v.getUint16(28, true) & 16383);
+			if (at(12, "VP8L")) {
+				const b = v.getUint32(21, true);
+				return ok((b & 16383) + 1, (b >> 14 & 16383) + 1);
+			}
+			if (at(12, "VP8X")) return ok(1 + (bytes[24] | bytes[25] << 8 | bytes[26] << 16), 1 + (bytes[27] | bytes[28] << 8 | bytes[29] << 16));
+			return null;
+		}
+		if (type === "image/svg+xml") {
+			const head = new TextDecoder().decode(bytes.subarray(0, 4096)), tag = /<svg\b[^>]*>/i.exec(head)?.[0] ?? "";
+			const attr = (name) => new RegExp(`\\s${name}\\s*=\\s*["']\\s*([\\d.]+)\\s*(px)?\\s*["']`, "i").exec(tag);
+			const w = attr("width"), h = attr("height");
+			if (w && h) return ok(Math.round(Number(w[1])), Math.round(Number(h[1])));
+			const box = /\sviewBox\s*=\s*["']\s*[-\d.]+[\s,]+[-\d.]+[\s,]+([\d.]+)[\s,]+([\d.]+)/i.exec(tag);
+			return box ? ok(Math.round(Number(box[1])), Math.round(Number(box[2]))) : null;
+		}
+	} catch {}
+	return null;
+}
+//#endregion
 //#region packages/core/src/kinds.ts
 function kindOf(type) {
 	const t = type.trim().toLowerCase();
@@ -507,6 +596,19 @@ function outwardLinks(body) {
 		}
 		seen.add(href);
 	}
+	for (const { href, embed, alt } of citeRefs(body)) {
+		if (!embed || seen.has(href)) continue;
+		const c = codeTarget(href);
+		if (!c || !pictureType(c.path)) continue;
+		out.push({
+			kind: "code",
+			href,
+			text: alt || c.path.split("/").at(-1),
+			path: c.path,
+			picture: true
+		});
+		seen.add(href);
+	}
 	return out;
 }
 //#endregion
@@ -514,10 +616,19 @@ function outwardLinks(body) {
 /** The base's own instructions, for Iso and any other agent, at the top of the notes (M15). Not a note. */
 const AGENTS_FILE = "AGENTS.md";
 const OKF_VERSION = "0.2";
-const ARTIFACT = /\.html?$/i;
-const IMAGE = /\.(png|jpe?g|gif|svg|webp|avif)$/i;
-/** What kind of citable file a path is, or null. */
-const citeKind = (path) => ARTIFACT.test(path) ? "artifact" : IMAGE.test(path) ? "image" : null;
+const KINDS$2 = [
+	["artifact", /\.html?$/i],
+	["image", /\.(png|jpe?g|gif|svg|webp|avif)$/i],
+	["data", /\.(csv|tsv|json|jsonl|ndjson|parquet|arrow|feather|npy|npz|xlsx|xls|h5|hdf5|nc)$/i],
+	["document", /\.(pdf|docx?|odt|rtf|epub|pptx?|odp|txt|tex)$/i]
+];
+/** What kind of file a path is: null for a note, or a path with no extension (a note or a folder named without one);
+*  anything not listed is misc rather than dropped. */
+const citeKind = (path) => {
+	const name = basename(path);
+	if (/\.md$/i.test(name) || !/\.[^.]+$/.test(name.replace(/^\./, ""))) return null;
+	return KINDS$2.find(([, re]) => re.test(name))?.[0] ?? "misc";
+};
 const isEntry = (v) => typeof v === "object" && v !== null && !Array.isArray(v);
 var Concept = class {
 	id;
@@ -697,7 +808,7 @@ var Bundle = class Bundle {
 			concept.links = links;
 			concept.outward = outwardLinks(concept.body);
 			for (const link of links) if (link.broken) this.issue(concept.path, "warning", "broken-link", `broken link to ${link.target}`);
-			for (const { href, title, embed } of citeRefs(concept.body)) {
+			for (const { href, title, embed, alt } of citeRefs(concept.body)) {
 				const path = this.filePath(concept, href), kind = path === null ? null : citeKind(path);
 				if (path === null || kind === null) continue;
 				const t = strip((title ?? "").toLowerCase());
@@ -707,10 +818,11 @@ var Bundle = class Bundle {
 					embed,
 					broken: !this.files.has(path),
 					rel: rating(title),
-					align: t === "left" ? "left" : t === "center" || t === "centre" ? "center" : null
+					align: t === "left" ? "left" : t === "center" || t === "centre" ? "center" : null,
+					...embed && alt ? { alt } : {}
 				};
 				concept.cites.push(cite);
-				if (cite.broken) this.issue(concept.path, "warning", "broken-link", `broken link to ${path}`);
+				if (cite.broken) this.issue(concept.path, "warning", "broken-link", kind === "image" ? `picture not in the notes: ${path}` : `broken link to ${path}`);
 			}
 		}
 		const byName = /* @__PURE__ */ new Map();
@@ -2087,13 +2199,60 @@ function mapModel(base) {
 			s
 		]);
 	}
+	const holdsFiles = /* @__PURE__ */ new Set();
+	const parentOf = (p) => p.includes("/") ? p.slice(0, p.lastIndexOf("/")) : "";
+	for (const f of base.files) holdsFiles.add(parentOf(f));
+	const pictureFolder = (d) => {
+		const dd = base.directories.get(d);
+		const here = [...base.files].filter((f) => parentOf(f) === d);
+		return !!d && (!dd || !dd.concepts.length && !dd.children.length) && here.length > 0 && here.every((f) => citeKind(f) === "image");
+	};
+	const refOf = (f) => citeKind(f) === "image" && pictureFolder(parentOf(f)) ? parentOf(f) + "/" : f;
+	const fileEdges = /* @__PURE__ */ new Map();
+	for (const c of base.concepts.values()) for (const x of c.cites) {
+		if (x.broken || !base.files.has(x.target)) continue;
+		const ref = refOf(x.target);
+		if (!fileEdges.has(c.id)) fileEdges.set(c.id, /* @__PURE__ */ new Set());
+		fileEdges.get(c.id).add(ref);
+	}
+	const rankOf = /* @__PURE__ */ new Map();
+	for (const [from, refs] of fileEdges) for (const ref of refs) {
+		edges.push([
+			from,
+			ref,
+			STRENGTH.uses
+		]);
+		rankOf.set(ref, Math.max(rankOf.get(ref) ?? 0, order.get(from)?.depth ?? 0));
+	}
+	const filesIn = /* @__PURE__ */ new Map();
+	for (const f of [...base.files].sort()) {
+		const of = citeKind(f), ref = of ? refOf(f) : null;
+		if (!of || !ref) continue;
+		const at = ref.endsWith("/") ? parentOf(ref.slice(0, -1)) : parentOf(f);
+		const list = filesIn.get(at) ?? [];
+		if (!list.some((n) => n.ref === ref)) list.push({
+			kind: "concept",
+			id: "f:" + ref,
+			ref,
+			weight: 1,
+			rank: rankOf.get(ref) ?? 0,
+			of
+		});
+		filesIn.set(at, list);
+	}
 	const dir = (id) => {
 		const d = base.directories.get(id);
-		return {
+		if (id && pictureFolder(id)) return null;
+		const children = [
+			...[...d.children].sort().map(dir).filter((c) => c !== null),
+			...[...d.concepts].sort().map(note),
+			...filesIn.get(id) ?? []
+		];
+		return id && !children.length && (holdsFiles.has(id) || d.children.length) ? null : {
 			kind: "dir",
 			id: "d:" + id,
 			ref: id,
-			children: [...[...d.children].sort().map(dir), ...[...d.concepts].sort().map(note)]
+			children
 		};
 	};
 	return {
@@ -2101,6 +2260,8 @@ function mapModel(base) {
 		edges
 	};
 }
+/** Whether a map item is a file (M23), not a note: its ref has an extension, or is a folder of pictures. */
+const isFileRef = (ref) => ref.endsWith("/") || citeKind(ref) !== null;
 /** 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. */
 function reach(model, note) {
@@ -2216,4 +2377,4 @@ 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 { ABOUT, AGENTS_FILE, Bundle, CONFIG_FILE, Concept, DEVICE_RE, DONE, DiffError, Directory, FrontmatterError, GITATTRIBUTES, ID_RE, KINDS, LearnerError, MAX_EVENT_BYTES, OKF_VERSION, OWN_DIR, OWN_FORMAT, RATINGS, SESSION_GAP_MS, SETTINGS_FILE, SNAPSHOT_EVERY, STRENGTH, STUDY_DIR, STUDY_FORMAT, SearchIndex, StudyError, agentName, ago, anchorOf, applyHunks, authors, blocks, change, citeKind, cleanEvent, cmp, cmpTuple, codeTarget, contentHash, describe, diffLines, doneOf, dumpFrontmatter, editFrontmatter, embedFor, embedRefs, findAnchor, findPassage, floatRepr, frontmatterText, hashText, headings, idsOf, impliedLinks, insertAt, iso, journalLine, journalPath, kindOf, legacyId, linkRefs, linkTexts, locate, mapModel, mergeRecords, momentKey, moments, newId, noteIds, outwardLinks, pagerank, parseSettings, parseStudyConfig, parseYaml, passageOf, pending, pyRepr, rating, reach, readJournal, readRecord, reconcile, recordPath, replay, round3, section, sessionMoments, settingsText, sha256, slugify, snippet, spliceText, splitFrontmatter, splitSource, stamp, strengthEdges, strip, stronglyConnected, text, textAt, toTime, tokenize, touched, wordAuthors, written };
+export { ABOUT, AGENTS_FILE, Bundle, CONFIG_FILE, Concept, DEVICE_RE, DONE, DiffError, Directory, FrontmatterError, GITATTRIBUTES, ID_RE, KINDS, LearnerError, MAX_EVENT_BYTES, OKF_VERSION, OWN_DIR, OWN_FORMAT, RATINGS, SESSION_GAP_MS, SETTINGS_FILE, SNAPSHOT_EVERY, STRENGTH, STUDY_DIR, STUDY_FORMAT, SearchIndex, StudyError, agentName, ago, anchorOf, applyHunks, authors, blocks, change, citeKind, cleanEvent, cmp, cmpTuple, codeTarget, contentHash, describe, diffLines, doneOf, dumpFrontmatter, editFrontmatter, embedFor, embedRefs, fileType, findAnchor, findPassage, floatRepr, frontmatterText, hashText, headings, idsOf, impliedLinks, insertAt, isFileRef, iso, journalLine, journalPath, kindOf, legacyId, linkRefs, linkTexts, locate, mapModel, mergeRecords, momentKey, moments, newId, noteIds, outwardLinks, pagerank, parseSettings, parseStudyConfig, parseYaml, passageOf, pending, pictureSize, pictureType, pyRepr, rating, reach, readJournal, readRecord, reconcile, recordPath, replay, round3, section, sessionMoments, settingsText, sha256, slugify, snippet, spliceText, splitFrontmatter, splitSource, stamp, strengthEdges, strip, stronglyConnected, text, textAt, toTime, tokenize, touched, wordAuthors, written };