| 1 | // Generates the Yjs updates behind CARLOS Docs' sample documents |
| 2 | // (internal/docs/seed.go). Same reasoning as generate.mjs: there is no |
| 3 | // Go-side Yjs encoder, so seed content has to arrive as a real update |
| 4 | // produced by real yjs, exactly like a person's own typing would |
| 5 | // produce one — never a text column written directly. |
| 6 | // |
| 7 | // The .txt files under internal/docs/seeddata/ are the editable |
| 8 | // source; this script is what turns an edited .txt back into the |
| 9 | // matching .update.bin the app actually applies. Run it after editing |
| 10 | // any seed .txt file, and commit both together — the .bin is the |
| 11 | // artefact, the .txt is the reason it looks the way it does. |
| 12 | // |
| 13 | // As of the fragment task, the update built here is a Y.XmlFragment |
| 14 | // on root "prosemirror" (the same root and shape |
| 15 | // fixtures/generate-fragment.mjs uses for core/src/fragment.rs's own |
| 16 | // wire fixtures — see that file's own comment), not a bare Y.Text — |
| 17 | // the read view now renders these through the CRDT's fragment |
| 18 | // renderer, and a Y.Text-only update reads back empty through that |
| 19 | // path. |
| 20 | import * as Y from 'yjs' |
| 21 | import { readFileSync, writeFileSync, readdirSync } from 'node:fs' |
| 22 | import { join, dirname, basename } from 'node:path' |
| 23 | import { fileURLToPath } from 'node:url' |
| 24 | |
| 25 | const seedDir = join( |
| 26 | dirname(fileURLToPath(import.meta.url)), |
| 27 | '..', |
| 28 | 'internal', |
| 29 | 'docs', |
| 30 | 'seeddata', |
| 31 | ) |
| 32 | |
| 33 | const names = readdirSync(seedDir) |
| 34 | .filter((f) => f.endsWith('.txt')) |
| 35 | .map((f) => basename(f, '.txt')) |
| 36 | .sort() |
| 37 | |
| 38 | // titles mirrors internal/docs/seed.go's seedDocs Title field, by Base |
| 39 | // — kept in sync by hand, the same way this script's own Base names |
| 40 | // already mirror seed.go's. blocksFor (below) uses it to decide |
| 41 | // whether a document's opening block is a literal echo of its own |
| 42 | // title (drop it) or real content that merely looks similar (keep |
| 43 | // it). If the two ever drift, the failure is visible and harmless |
| 44 | // either way: a document whose title changed here keeps a stray |
| 45 | // opening paragraph until this map is updated, never the reverse |
| 46 | // (silently eating real content it doesn't recognise as a title). |
| 47 | const titles = { |
| 48 | 'packing-list': 'Packing List — Thursday Site Visit', |
| 49 | 'proposal': 'Proposal: Refitting the Elm Street Reading Room', |
| 50 | 'meeting-notes': 'Notes — Tuesday Planning Sync', |
| 51 | 'documentation': 'Quick Reference: Resetting the Site Printer', |
| 52 | 'letter': 'Letter to the Harbourview Allotment Committee', |
| 53 | 'history': 'A History of the Millbrook Waterworks', |
| 54 | } |
| 55 | |
| 56 | // blocksFor turns a .txt source into the paragraph/heading blocks its |
| 57 | // matching Y.XmlFragment gets built from: split on blank lines, then |
| 58 | // classify each block. A block reads as a heading only when it looks |
| 59 | // like one — short, a single line, no trailing sentence-ending |
| 60 | // punctuation — and never for the very first block (i === 0): that |
| 61 | // guard exists independently of the title check below, because a |
| 62 | // short, unpunctuated opening line is not always a title echo — |
| 63 | // letter.txt's own opening block, "Dear Members of the Harbourview |
| 64 | // Allotment Committee,", fits the heading shape (52 chars, one line, |
| 65 | // trailing comma) but is a real salutation, not a section break. This |
| 66 | // is what gives "Early years" and its neighbours in the Millbrook |
| 67 | // waterworks history their real shape, rather than leaving every |
| 68 | // block a paragraph. |
| 69 | // |
| 70 | // title, when given, is compared against the *first* paragraph only |
| 71 | // (trimmed, exact match) — fix round 1: an earlier version of this |
| 72 | // function dropped every document's opening block unconditionally, on |
| 73 | // the assumption that a seed .txt always opens by restating its own |
| 74 | // title. That is true for five of the six seed documents but not |
| 75 | // letter.txt, whose opening line is real content the unconditional |
| 76 | // version would have silently deleted. Comparing against the actual |
| 77 | // title, rather than guessing from shape or position, is what lets |
| 78 | // this stay correct for both cases at once. The comparison runs after |
| 79 | // classification (on the finished `blocks` array), not before it on |
| 80 | // `paras`: dropping the first *paragraph* before classifying would |
| 81 | // shift every remaining block's index down by one, and the i !== 0 |
| 82 | // guard above would then wrongly protect the new index-0 block (e.g. |
| 83 | // history.txt's "Early years") from ever becoming a heading. |
| 84 | function blocksFor(text, title) { |
| 85 | const paras = text |
| 86 | .split(/\n\s*\n/) |
| 87 | .map((s) => s.trim()) |
| 88 | .filter(Boolean) |
| 89 | const blocks = paras.map((p, i) => { |
| 90 | const looksLikeHeading = |
| 91 | p.length < 60 && !p.includes('\n') && !/[.!?:]$/.test(p) && i !== 0 |
| 92 | if (looksLikeHeading) { |
| 93 | const h = new Y.XmlElement('heading') |
| 94 | h.setAttribute('level', '2') |
| 95 | const t = new Y.XmlText() |
| 96 | t.insert(0, p) |
| 97 | h.insert(0, [t]) |
| 98 | return h |
| 99 | } |
| 100 | const el = new Y.XmlElement('paragraph') |
| 101 | const t = new Y.XmlText() |
| 102 | t.insert(0, p) |
| 103 | el.insert(0, [t]) |
| 104 | return el |
| 105 | }) |
| 106 | if (blocks.length > 0 && title !== undefined && paras[0] === title.trim()) { |
| 107 | return blocks.slice(1) |
| 108 | } |
| 109 | return blocks |
| 110 | } |
| 111 | |
| 112 | for (const name of names) { |
| 113 | const text = readFileSync(join(seedDir, `${name}.txt`), 'utf8') |
| 114 | |
| 115 | const doc = new Y.Doc() |
| 116 | // Fixed clientID, same reasoning as generate.mjs: running this twice |
| 117 | // must produce byte-identical output. It only needs to differ from |
| 118 | // the *other* clients that ever touch this specific document, and a |
| 119 | // freshly seeded document has none yet — the app's own random |
| 120 | // per-document ClientID (internal/docs/store.go's Store.Create) is |
| 121 | // what a real edit would be attributed to. |
| 122 | doc.clientID = 1 |
| 123 | const blocks = blocksFor(text, titles[name]) |
| 124 | doc.getXmlFragment('prosemirror').insert(0, blocks) |
| 125 | |
| 126 | const update = Y.encodeStateAsUpdate(doc) |
| 127 | writeFileSync(join(seedDir, `${name}.update.bin`), Buffer.from(update)) |
| 128 | console.log(`${name}: ${text.length} chars -> ${blocks.length} blocks -> ${update.length} bytes`) |
| 129 | } |
| 130 | |
amadan