theseus/bundled-addons/docx-editor/README.md
Hermes Dev 8a08879c49 test(docx-editor): the end-to-end suite moves into the repo
It lived in scratchpad/, which is untracked, and a tidy-up on 2026-10-22
deleted it — along with the pdf-editor's copy of the same harness. Forty
checks of coverage gone, and nothing in the repo so much as mentioned they
had existed. Anything load-bearing belongs next to the code it tests.

Rebuilt as test/e2e.mjs plus test/e2e-harness.mjs, runnable with
`npm run test:e2e`. It boots a throwaway Theseus, opens documents through
the sidebar panel, clicks the real ribbon, types with real key events, saves
and exports, then reads the result back with the add-on's own reader. The
editor carries no test-only hook, so a pass is a pass for what users get.

This is the only level that catches breakage which exists solely in the
browser — a bundle that will not load as a classic script from file://,
webfonts that register and then never fetch, a ribbon button wired to
nothing, a PDF on the wrong paper. All four have happened.

One trap is recorded in the harness: the throwaway profile must not live
under addon-build/, whose package.json declares "type": "module". Node picks
a file's module kind from the nearest package.json above it, so an add-on
seeded into a profile there is parsed as ESM and dies on its first require()
with "require is not defined in ES module scope" — which reads exactly like
the add-on being broken, and is not.

Run against today's build: 38/38, including the new start-on-first-use
activation.
2026-10-28 22:31:57 +02:00

123 lines
6.2 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

# Word editor — a Theseus extension
Opens, edits and saves Word documents (`.docx`) in a full Theseus tab, and
exports them as PDF.
Saving comes in three shapes. **Save** drops a copy in your downloads until the
document has a home; **Save as…** opens a real file dialog where the extension
you type — `.docx` or `.pdf` — decides the format, and from then on Save writes
straight to that file; **PDF** is the one-click export. PDFs are rendered by
Chromium's own print pipeline from `lib/doc-css.js`, the same stylesheet the
editor displays, on the paper size the document itself specifies — so the
export matches the preview rather than being a second opinion about it.
It is a first-party extension: written here, signed with the Silent Mode
operator key, and updated over the channel at
`theseus.x/extensions/docx-editor/` like every other add-on Theseus ships.
It does not go through the community catalogue, which is for extensions
published by whoever owns a BNS name.
It ships inside Theseus, so a fresh install already has it and Settings
lists it under **Built into Theseus**; the channel above then updates it
between releases without waiting for one. That is the same arrangement as
screenshot, aegis and pdf-editor, and it is what gives a user their first
copy at all — Settings can only *install* from the community catalogue, so
an extension that is neither bundled nor catalogued has a working update
channel and no way in.
The cost is about 400 KB compressed in the installer, most of it the
vendored editor libraries.
## The page, and the fonts
The sheet is drawn at the size the document says it is — A4 stays A4, margins
come from its own `sectPr` — and CSS `zoom` scales the whole thing to fit the
window. Scaling rather than widening is the point: a page stretched to the
window would break every line somewhere different from where the printed page
breaks it. The footer carries the control, and **Fit width** is the default
because a page marooned in the middle of a wide window wastes the screen. The
choice is remembered per editor, not per document.
Ubuntu and Fraunces ship with the add-on, in `fonts/`, because Windows has
neither and a font offered in the ribbon that the machine lacks is a font the
user picks and then cannot see. Regenerate with `npm run fonts`; licences are
in `fonts/LICENSES.txt`. They cost about 700 KB, most of it Ubuntu's Cyrillic
and Greek coverage — dropping those subsets would roughly halve it.
Two wrinkles worth knowing. The editor is on `file://`, where Chromium
registers `@font-face` rules but refuses to fetch the font files, so the
add-on reads them and hands the page a stylesheet with the woff2 inlined as
data URLs; the same inlined copy goes into the PDF export, whose print window
runs from a temp folder where a relative `url()` would resolve to nothing. If
either path fails the ribbon labels those two families "(not available)"
rather than pretending.
## The icon
`icon.svg` is the one drawing. `npm run icons` rasterises it to
`icon-16/32/48/128/256.png` and writes a minified copy into `addon.json`'s
`icon` field, which is what the dock button and the catalogue card read — so
there is never a second version to keep in step.
It is a blue document badge, paired with pdf-editor's red one so the two
editors read as a set in the dock — a solid fill stays legible at the 16–18px
the dock actually renders, which an outlined page does not.
Deliberately not Microsoft's Word icon: their mark is a blue sheet carrying a
white W, and dressing our extension up as theirs would be both a legal
problem and a dishonest one. A folded-corner page is the universal document
glyph, and "DOC" is the file format rather than anyone's trademark.
## Building
The vendored libraries (mammoth, ProseMirror, docx, JSZip) are bundled by a
build step that lives outside this folder, because mammoth needs local patches
before it can carry everything the editor edits:
cd ../../addon-build/docx-editor
npm install
npm run build # writes vendor/docx-vendor.js here
npm run icons # re-rasterises icon.svg after editing it
`vendor/docx-vendor.js` is committed, so the extension is installable straight
from a checkout; re-run `npm run build` after touching anything under
`addon-build/`.
## Publishing
The full recipe is in [docs/ADDON-UPDATES.md](../../docs/ADDON-UPDATES.md);
the short version, signed with the operator key and always from a clean copy
of the commit rather than the working tree:
git archive HEAD TheseusNavigator/extensions/docx-editor | tar -x -C <clean>
node scripts/sign-addon-update.mjs <clean>/TheseusNavigator/extensions/docx-editor https://navigate.st/bns/theseus.x/extensions/docx-editor "$USERPROFILE/.silentmode/ops/addon-update-key.pem" --out out/
node ../Argus/src/lib/sia-upload.js out/docx-editor-<version>.tar.gz bns/theseus/extensions/docx-editor
node ../Argus/src/lib/sia-upload.js out/updates.json bns/theseus/extensions/docx-editor
The channel lives on Sia only — nothing goes to the VPS filesystem — and is
served through the gateway at `https://navigate.st/bns/theseus.x/extensions/`.
A first release has no `updates.json` to prepend to, so it ships one holding
just its own entry. Versions must increase.
## Testing
Three levels, all re-runnable, all from `../../addon-build/docx-editor`:
npm test # the built-in fixture, round-tripped
node test/corpus.mjs <folder of .docx> # real documents
npm run test:e2e # a real Theseus, driven over CDP
The last one boots a throwaway Theseus, opens documents through the sidebar,
clicks the real ribbon, saves and exports, and reads the result back — about
forty checks. It is the only level that catches breakage which exists solely
in the browser: the vendored bundle failing to load as a classic script from
`file://`, webfonts that register but never fetch, a ribbon button wired to
nothing, a PDF on the wrong paper. Each of those has happened.
It used to live in `scratchpad/`, which is untracked, and was deleted by a
tidy-up on 2026-10-22 along with the pdf-editor's copy. It lives here now.
What survives a round trip, and what doesn't, is written up in
[ROUND-TRIP.md](ROUND-TRIP.md). Read that before promising anyone a Word
feature.