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 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.
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.
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 structural header, beside the styling one: pageBreaks grows the
between-rows spacer by the first row's height and draws a copy of that
row — colgroup included, so the columns line up — at the foot of the
gap, which is the top of the continuation. Round-trips as ODF
table:table-header-rows and Word's w:tblHeader; LibreOffice keeps both.
Word's "don't allow row to break" needs nothing: a row that fits a page
is already an atomic leaf here.
Neither word processor passes it into a cell — LibreOffice puts cell text in
its Table Contents style, Word's table styles carry w:spacing after=0 — so the
space filled the cell and vertical alignment had no slack left to move the text
in, which read as the alignment being ignored. The space above is LibreOffice's
alone, so only the add model zeroes that too. Both exporters write the zeros
onto the paragraph, and only when the default style has spacing at all.
The stripe began one row late: regionsAt counted the second body row as the
first band. Probed on a five-row fixture with a red band1Horz — LibreOffice
shades row 0 where the style paints no header region, and the row under the
header where it does.
The corpus cannot see this (no fixture uses the editor's own registry), so it
stayed at 739; checked by eye and by the unit test's inverted expectations. A
document with a table style assigned re-bands on its next open.
w:tblInd means the distance to the cell's text up to Word 2010, so a table
indented 0 hangs its left cell margin into the page margin and its first column
lines up with the body. We drew every such table 2.2mm right of LibreOffice.
Import subtracts it, marginLeft may now be negative, and both exports carry the
edge as it stands.
Word 2013 (compatibilityMode 15) redefined it as the edge — probed both ways
against soffice on one file, 29.6mm against 31.75mm. Corpus 795 -> 761.
w:tblStylePr — header row, banded rows, first/last column — was never read, so a
table styled in Word arrived as a bare grid. Import bakes each area into the cells
now (fill, borders, run marks): the file's own style has no registry entry to
re-derive regions from, and the bake is what a re-export carries.
A band's box is the whole banded region, so its insideH is the line between two
band rows; banding counts from the first body row, as in Word and LibreOffice.
Corpus 803 -> 795, page counts unchanged.
fo:padding defaults to 0 in ODF, but an undeclared side was left open and
fell through to Word's implicit 0.19cm, indenting every cell of a table
whose file names no cell style — a fixture's figure captions by 1.9mm.
Two ways a table's height was off. Its space above was a margin, so it
collapsed against the paragraph above instead of adding to it — now it rides
--space-before like a paragraph's. And every cell border paints a whole pixel
however thin the file declares it, where LibreOffice reserves the declared
width; the excess now comes off the cell's padding-bottom, 0.25mm a row.
Both round-tripped through save/load already but had no way in: the
header/footer bar gains a Chapter button next to page number/count, and the
table toolbar three buttons for the cell's vertical alignment.
ODF puts fo:margin-top/-bottom on the table style and LibreOffice honours
them like a paragraph's; we dropped both, so every callout table in a fixture sat
half a line too tight and the document ran a page short. With them the Math
Guide paginates to LibreOffice's 63 pages exactly.
Two measured deviations, both worth ~1mm a row on a fixture: a cell's first block
got the space above that LibreOffice adds nowhere in a cell, and a repeating
header row rendered centred and bold because it is a <th> and the UA says so.
Pagination's page-top-drop detection now skips cell leaves, or the space that
is never there would read as a decoration to add back.
Import read style:vertical-align but neither exporter wrote it, so a
mid-cell callout label jumped to the row's top on every save. odf-kit's
CellOptions and docx's TableCell both take it natively; DOCX import maps
w:vAlign in as well.
ODF style:vertical-align (middle/bottom) on a table cell, which every
LibreOffice callout table uses for its Tip/Note label. Import and render
only: odf-kit dedupes cell styles by property, so writing it back needs a
pass of its own — noted in docs/architecture/tables.md.
w:tcMar and an ODF cell style's own fo:padding were both dropped: the
table's value (in ODF the first cell's) stood for every cell.
cellPadding is now an attr on tableCell/tableHeader too, measured against
the table's rather than the editor default, so only a real disagreement
carries one. --cell-pad inherits, so the cell's own declaration wins for
its own td and the CSS is unchanged.
Checked against LibreOffice: a 1.2cm cell in a 0.3cm table puts its text
at 332.2pt where the grid column starts at 297.6.
The root file is read at every session start and had grown to 87 KB, so every
session paid for knowledge it did not use. Move the per-directory detail into
nested CLAUDE.md files, which load when a file in that directory is touched,
and the five deep-dive topics into docs/. The root keeps the commands, the hard
rules and an index of where the rest lives.
Content moved verbatim — the probed Word/LibreOffice behaviour, sentinel order
and shared constants are what the code cannot say. CI fails the root over 120
lines, since a prose budget alone is what let it grow.
Steffen Becker