Generator: text boxes (shape, fill, stroke), inline and display
formulas, comments, bookmarks with cross-references, ruby, rowspan
merges. Display rides a formula owning its line; a stroke width needs
a stroke color; a vertical merge keeps a second column so no row is
all covered cells. Normalize canonicalizes comment ids and dates
(naive local vs UTC are one instant) and trims a formula's latex
(the OMML side keeps the serializer's trailing space on purpose).
New in the pool: paragraph shading/rule lines (also in cells and lists), rtl
paragraphs, keepNext/keepLines/widowControl, tab stops, custom bullet chars,
heading levels to 8, image alt/rotation, row heights and custom note labels.
noHyphenation stays out: without the document-wide hyphenate flag the
importers suppress it as meaningless. 1000 sweep seeds clean on both legs.
colspan merges (grid-consistent), orderedList listStyleType (upper-* only on
nested lists: lower/decimal forms are depth defaults the importers suppress),
and inline PNG images. 300 sweep seeds clean on both legs.
New in the pool: underline/strike line styles, super/subscript, caps,
textPosition, links, page breaks, footnotes/endnotes, cell shading and
vertical alignment. normalize() canonicalizes importer-generated note ids and
drops the link mark's default plain:false; FUZZ_SEEDS widens the sweep beyond
the default 50. This is the suite that surfaced the last four export/import
fixes; all 300 sweep seeds now round-trip clean on both legs.
Only a pure delete was turned into a deletion mark; typing or pasting over
a selection ran as it was, so the replaced text left the document unmarked
and rejecting the change lost the original for good. A one-step replacement
now marks its range deleted and puts the new content after it, as both word
processors do.
The document's content ended in an optional noteSection, so the content
match after the first child offered noteSection as its defaultType — and
that is what TipTap's clearNodes converts a block into. Every style command
on a top-level block past the first threw "Invalid content for node type
noteSection" and applied nothing. Spelling the same language as an
alternation puts a paragraph first again.
Replace the hardcoded node if/else chain in describeTransaction with a
NODE_LABELS map covering image, text box, formula, page break, TOC,
fields, cross-reference, index/bibliography entries, ruby and columns;
foot-/endnotes labelled via their kind attr. MARK_LABELS gains link,
comment, bookmark, character style and tracked change.
exportPdf gains a test seam: a window.__edentextPdfSink takes the bytes
instead of doc.save(), whose download hangs in headless Chromium. The
smoke test clicks the real export menu over the imported corpus document
and asserts %PDF magic and a plausible size. The test browser is pinned
to en-US so the labels it clicks are deterministic across machines.
The two framework-free 0%-coverage modules: the recent-files list
(dedupe by name, eight-entry cap, corrupt-localStorage tolerance) and
the labelled undo/redo mirror against a live jsdom editor (labels per
edit kind, undo/redo moves, redo branch discarded by a fresh edit).
The same seeded documents go through buildOdt->importOdt and
buildDocx->importDocx; the two imports must agree, so one exporter
silently losing what the other keeps now fails with the seed named.
Column widths are ignored here (ODT leaves defaults implicit, DOCX
always writes w:tcW); each format's own leg still covers them.
Found by the cross-format fuzz leg. importDocx baked the resolved cell
chain in as direct formatting even where it equals what an unset attr
already renders (0/0 spacing; single line height over a single default) —
every re-imported own file grew spacing attrs on every cell paragraph.
And a numbering level kept the first list kind it saw, so a nested bullet
list beside a nested ordered one came back ordered; a kind conflict now
forks its own numbering reference, as the ODT side mints a list style.
tests/fuzzDoc.ts generates documents from a seeded PRNG (marks, lists,
tables, indents, XML-hostile text); fuzz-roundtrip.test.ts runs 50 seeds
through buildOdt -> importOdt and diffs the normalized JSON, naming the
failing seed. normalize/firstDiff moved out of roundtrip.test.ts into
tests/normalize.ts for reuse. The generator documents what is lossy by
design: default-equal marks in headings, indents and start in nested lists.
tests/smoke/run.mjs boots the production build in Playwright Chromium and
checks what jsdom cannot: the bundle boots, typing and the bold shortcut
reach ProseMirror, autosave survives a reload, a corpus ODT imports and
paginates; any uncaught page error fails the run. Wired as npm run
test:smoke and a CI job (build + chromium install), playwright-core is
now a devDependency.
Ctrl+F7, Review ▸ Thesaurus and the context menu open one dialog: the
word at the caret, the sense groups it is in, click one to replace it,
and a box to look up any other word. The data is MyThes re-packed to one
group per line (de 3.0MB/37k groups, en 6.7MB/143k) — the format repeats
every group once per member, ten times the bytes. Loaded on first use for
the document's language and scanned, not indexed: ~4ms a lookup, where a
word→group index measured ~100MB of memory against the text's ~8MB.
The CI grep for "render-parity/fixtures" outside tests/render-parity
flags any occurrence of that path, including comments. corpus.test.ts
only mentioned the path in prose; reworded it so the guard passes.
View ▸ Multiple Pages (1–4): rows of pages filled left to right, scrolling
as one canvas, every page editable. Neither reference product stops at two —
LibreOffice takes a column count, Word fits what the window allows — and both
re-zoom for it, so a whole row is fitted when the count changes. A pane is a
grid cell clipped to one page, a view showing only a contiguous slice; two
rows of cells take turns as the upper one. Cost: columns x 2 live views, each
holding the whole document (measured: +94k nodes and +16ms per keystroke per
view at 177 pages).
A new document starts with it off and an import adopts the file's own
flag, so an editor left recording no longer marks the next document it
opens. Both formats carry it, probed against soffice: ODF on the change
registry (text:track-changes, written for the flag alone where nothing
is recorded yet), Word as w:trackRevisions in settings.xml — LibreOffice
writes and reads exactly those two, and reading w:trackChanges instead
would have found nothing.
With revision recording on, setContent inserted the whole file in one
transaction, so the recorder marked every run as the reader's own
insertion — the document then rendered underlined in the author colour
(#c69200) from the first character. Measured: 19 843 of 19 843 characters
on an 11-page import, 436 revision spans.
Both loads (new document, import) now carry the RECORDING meta; what is
typed after the open is still recorded, as in both word processors.
View ▸ Split (Ctrl+Alt+S): a second live view of the editor's state, stacked
below the first and editable the same way. The panes share a width, so the
second renders the pagination the first measures — pageBreaks, columnsFlow
and tabStops skip the pass on a split pane rather than run it twice.
Every widget decoration now builds its DOM in a function: two views render
one decoration set, and a shared node is taken back and forth between the
two documents until the tab locks up (measured on a two-page document).
A polygon, a polyline, a curve or a connector's elbow used to be dropped
with a warning; it now rides the box as an SVG path in the 0…100 box a
preset's points use, and the editor draws it. Read from all four dialects
the two products write (ODF points and svg:d, a non-primitive
enhanced-path, DrawingML custGeom, VML's reversed-case path), written
back as enhanced-path and custGeom — LibreOffice renders both exactly as
the editor draws them. An open outline is stroked, never filled.
Ten documents the docx lib authors directly, plus the ODT twin
LibreOffice converts each into, now live in tests/corpus/ instead of the
gitignored fixtures dir. corpus.test.ts asserts both formats of a
document read the same and that each survives our export and import
either way. The heading styles gained the outline level Word writes —
without it LibreOffice converts a heading to a paragraph that merely
carries the style, and the two formats disagreed.
Both come from the document's language, read off Intl: the symbol's side
and spacing, the short date's order and padding. Each minted data style
names that language — without it LibreOffice prints its own locale's
separators (probed: a German document's 1.234,00 € came out $1,234.00).
A date is a serial day from LibreOffice's day 0, 1899-12-30. Word's
formula dialog offers no date, so the DOCX leg uses the field's \@
picture. Also drops a document's name from three older comments.
The three attrs LibreOffice's Format… dialog is mostly about, on both
sides of the bar. They live together on one TextStyle mark and a mark of
one type replaces that type whole, so each run's own attrs are the base of
the new one — recolouring leaves the font alone.
F3 was bound twice from here on: it now expands an AutoText shortcut where
the caret has one, as it does in both word processors, and finds the next
match otherwise.
Word's Text Direction and LibreOffice's, on the frame's own toolbar: the
browser lays the vertical flow out, so the box needs nothing but the mode.
ODF keeps it in the frame style's *paragraph* properties — probed: in the
graphic properties LibreOffice drops it, there it keeps it and lays the box
out — and Word in the shape body's `vert`, the VML fallback's
`layout-flow:vertical` read too. Both survive a LibreOffice re-save.
The Formula dialog gains the number format Word's has: whole numbers, two
decimals, grouped thousands and the two percentages beside the general
one.
ODF keeps it on the cell *style* as a reference to a data style — odf-kit
shares one cell style across every cell, so a formatted cell gets a copy
of it — and LibreOffice writes both back unchanged. Word keeps it on the
field as its \# switch; LibreOffice's DOCX filter neither reads nor writes
that switch, so that leg holds through our own round trip.
Ruby, as both products' phonetic guide writes it: one element holding a
base and its reading, so the editor keeps it as one inline atom and the
browser sets it as a real <ruby>.
ODF gets <text:ruby> plus a minted ruby-family style (centred, above —
probed: LibreOffice writes that style back unchanged); Word a <w:ruby> run
from a post-pack pass, the package having none. Both legs verified through
a LibreOffice re-save, our .docx to .odt included.
LibreOffice's Tools ▸ AutoText, in the ribbon where Word keeps its Quick
Parts: store the selection under a name and a shortcut, insert it from the
library or by typing the shortcut and pressing F3 — LibreOffice's key for
it.
An entry is HTML parsed back through the schema, so it keeps its marks and
blocks; the library is app-level, as it is in both products, so no file
carries it.
LibreOffice's Word Completion, its defaults included: words from eight
letters up are collected as they are finished, and the first three
letters offer the rest. The offer is a widget decoration after the caret,
so it is outside the document until Enter takes it — nothing to undo when
it is not taken.
Only while typing: a bare caret move drops the offer, or an Enter meant to
split the paragraph would complete a word instead. The settings share the
AutoCorrect dialog, as they share LibreOffice's.
The ¶ toggle adds a format to each side: bold / italic / underline and a
paragraph style, as LibreOffice's Format… and its Paragraph Styles box.
A format narrows a text search to the runs carrying it, and with no term
the formatting is the search — every such run, or every paragraph in the
style, empty ones included.
A replacement with no text only reformats, which is how a document gets
restyled without retyping it; a heading style switches the node type over
the way setParagraphStyle does, so the extension takes the same sheet.
Both dialogs offer three sort keys and let you pick the type; only the
first key was offered, and the type was always the detected one. A later
key decides only where the ones before tie, and the sort stays stable.
Alphanumeric is the plain collation ("10" before "2") — that is what
choosing it over the detected type is for, so the two collators differ in
`numeric`. Sorted numerically, a cell that is no number sorts last either
way, as an empty cell does in a spreadsheet.
Cite by short name (LibreOffice's own), numbered, APA, MLA or Chicago:
the style sets the citation's label and the row's shape, and rides the
bibliography index node — ODF as the entry template LibreOffice
regenerates its rows from (verified through a re-save), Word as the
sources part's StyleName. Labels are derived and kept true by an
appendTransaction, so adding a citation renumbers the rest.
Word's Start at, per section: ODF writes style:page-number on the
paragraph that switches master page, DOCX a w:pgNumType w:start. That
section is no longer exported as a continuous break — probed, LibreOffice
drops the start off a continuous one, and neither word processor honours
it there. The footer counts from the nearest restart at or above the page.
Every recorded change in document order beside the comments pane: author,
date, the text, accept or reject in place. A change split by a paragraph
boundary is one row, so acceptRevision/rejectRevision take an id. The
author colour is a decoration, since a mark cannot know the document's
other authors; it nests inside the mark's span and carries the under/
strikethrough with it. shortcutHint no longer throws on an id nothing
binds, which heading 7-10 had just made reachable.
HTML stops at h6, so 7-10 render as unknown elements editor.css makes
blocks; their built-in styles continue level 6 at 12pt, since LibreOffice
writes these styles with no properties and resolves them from its own
pool. Both importers read a two-digit level again (the style-name regexes
stopped at one digit). Verified through a LibreOffice re-save of both
formats and in the live editor.
LibreOffice's Format > Paragraph > Text Flow checkbox. Only "off"
travels (Word has no per-paragraph "on") and only under a document that
hyphenates, or it would say what is already true. ODF writes
fo:hyphenate="false" into the paragraph's text properties, DOCX gains
w:suppressAutoHyphens through a post-pack pass — the docx package
exposes no option for it. Both survive a LibreOffice re-save.
Both were selectable in the dialog and changed nothing — they only rode the
file formats, so an imported document numbered and placed its notes wrong.
noteLabels now counts within a bucket (class, chapter, or the page pageBreaks
publishes through setNoteAnchorPages); a collected footnote leaves position:
absolute and opens its own page, as LibreOffice does. Counting and position
move to the footnote tab, the only place LibreOffice offers them.
The separator's own alignment moves the line alone — probed, notes stay left.
Its picture nests a second w:drawing, and the importer's subtree
search found *that* one's wp:inline first: the box was read as the
bare picture, so every word of the caption in it was dropped —
which is what Word's own Insert Caption writes for a floating
picture. The root is now the drawing's direct child. The export
side gains the caption's SEQ field inside a box, so the running
number stays live there too. Leg 30 covers both.
Two faults on one picture. A block after a float keeps its box over
the float — only its lines move — so the caption on top swallowed
every click meant for the picture; the frame is lifted above the
in-flow boxes. And a caption can only sit under a side-wrapped frame
if it is boxed to it, so its span in the column is measured and
becomes the caption's indent, which also drops it clear of the float.
Measured at 100% and 150%: caption under the picture, its width, in
all four wrap modes.
A picture with "above and below" wrap is a frame centred in the
column while its paragraph carries no alignment at all, so the
caption still went to the left margin. framePlacement reads the
place off the frame: centred, the side it is set against, or its
offset as the caption's indent. The debug dump gains a frames
section — every other one measures the flow downwards, and this
is a sideways question.
It was inserted as a plain paragraph, so a centred or right-aligned
picture got its caption at the left edge of the text column.
LibreOffice frames picture and caption together; a loose paragraph
carries that over as the alignment of the block it captions.
Measured in the live editor: a centred 240px picture at 520…760
now has its caption text at 580…700.
Two endpoints rather than a box: no text, no fill, and the diagonal of
the frame with `flipV` picking which one. The three kinds share their
preset names, since the heads a file declares are what tell them apart.
ODF gives a line its own element and its heads a named marker, which
needs the draw: prefix declared in styles.xml — undeclared, the file
does not parse at all. Word keeps the frame and flips it.
The formula rides the cell, as ODF writes it; Word's field syntax is what
is stored, since it is the wider language. `ooow:` in the ODF attribute
value is a namespace prefix — undeclared, LibreOffice reads it as part of
the formula and prints "Expression is faulty".
Sort takes one key and detects the type per cell where both dialogs make
you pick. Number recognition is off by default, as it is in LibreOffice,
and a formula parses a cell's text either way — exactly as there.
The page grid was one repeating cycle, so every page had to be the same
size. It is now a table of height runs the placement pass grows as each
section's first page becomes known, and the sheets are one box per page
instead of a repeating gradient — .paper reserves the widest section's
paper, a narrower one takes the difference as a right inset.
Layout ▸ Orientation offers "this section"; the size stays document-wide
in the ribbon but round-trips per section from a file.
The `docx` package swaps the pair itself for a landscape section, so it
wants the portrait box; handed the swapped one it wrote a portrait page
merely labelled landscape — LibreOffice renders it portrait (measured:
21×29.7cm out of a landscape document).
Per-section paper rides the same fix: an HfSet carries its own format and
orientation, ODF cloning the section's page layout as it already did for
the margins and DOCX writing it into each w:sectPr.
The citation is an inline atom holding the whole record, which is what ODF
writes on text:bibliography-mark — so no database travels beside the file.
The list is a new index family: one row per source, in document order and
without page numbers, as LibreOffice prints one.
DOCX cites by tag and keeps the sources in a custom-XML part the docx
package knows nothing about, so the post-pack pass mints the part, its
properties, both relationships and the content type itself.
References ▸ Index entry marks a place with the term the index should
list — under a key of its own where it belongs elsewhere — and the Index
in the same menu lists each term once, sorted case-insensitively, with
every page it was marked on.
Both formats write a place, not a range, so the entry is an inline atom:
ODF text:alphabetical-index-mark plus text:alphabetical-index, Word XE
plus INDEX. LibreOffice reads our file back unchanged. The bibliography,
the other half of that gap, is still open.
F5 (LibreOffice's key) opens an outline of the document's headings:
click one to jump, and move it up or down or promote/demote its level,
each carrying everything under the heading. Below it the tables,
pictures and bookmarks LibreOffice's Navigator also lists.
A move swaps with the adjacent chapter at the same level and stops at a
heading above it, so a subheading never reparents itself. The pane docks
right beside the comments one — the classic chrome's Tools chip owns the
left edge.
A .ott/.dotx is the built document with one label changed — ODF's
package media type, Word's content type for the document part — so the
picker offers both and the chosen extension decides which exporter runs.
Opening or saving also records the file in a recent list. The name is in
localStorage so the menu draws at once; the handle that reopens it is in
IndexedDB beside it. A file that has moved, or whose permission the
browser drops on reload, is reported and dropped from the list.
A handful of photos is the whole ~5 MB quota, and past it the document
stopped saving. The autosaved JSON now keeps an `idb:<key>` per picture
and the bytes live in their own store; the key is a sampled content
hash, so a picture used twice is stored once and each save sweeps what
the document no longer names.
loadDocument is async as a result and Editor.svelte awaits it. Where
IndexedDB is unavailable the data-URIs stay inline exactly as before.
A list inside a box was flattened to literal bullet/number runs and a
picture was dropped: the box XML is a hand-built string, and neither had
the package parts it needs.
applyTextBoxesDocx already holds the whole unzipped package, so it mints
them — a media entry with its relationship, and an abstractNum/num pair
above the package's own ids, the paragraphs carrying a sentinel until
those ids exist. Verified through LibreOffice and back through our own
importer, nesting, marker symbols and start values intact.
Steffen Becker