MattssonandClaude Fable 5.1
7a30f623ba
* fix(invoices): draft stamp out of flow, page-break control, whole-word wrapping, English units, multi-line descriptions
A user reported five things about the invoice PDF after the English
translation shipped:
- The yellow draft banner pushed the whole document down, so a draft
previewed differently from the final invoice. It is now a small stamp in
the page's top margin (absolute, fixed), repeated on every page, out of
the flow.
- Table rows, totals, the payment box and the notice boxes could split
across a page break, and a section heading could be left alone at the
bottom of a page. Those blocks now carry wrap={false}; headings and the
table header carry minPresenceAhead.
- react-pdf hyphenated Swedish words with English patterns ("Septem-ber").
Descriptions, notes, notices and the footer now wrap whole words.
- "st" printed verbatim on an English invoice. The editor's known units map
to English labels at render time; user-typed units print as stored.
- Descriptions were single-line inputs, so a user could never choose where
a line breaks. The editor field is now an auto-growing textarea; the PDF
and the on-screen views keep the line breaks; the Peppol item name
collapses them (single-line field).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BuYbae3WWwYzBucTUVayoW
* fix(invoices): keep long tokens printable and oversize text splittable in the invoice PDF
Skeptic findings on the first commit:
- A word wider than its column was dropped from the page or overprinted the
quantity column, because the no-hyphenation callback gave react-pdf no
break point inside it. Words up to 16 characters still wrap whole; longer
tokens (URLs, e-mail addresses, references) break after separators and
every 16 characters.
- wrap={false} on a text row or notes taller than a page clipped everything
past the page edge. Line descriptions and notes are now kept together only
while a line estimate says they fit; past that they split.
- A multi-line description reached the periodisering voucher text and broke
the SIE export (one record per line). The accrual builder and the Peppol
item name share toSingleLine(); the SIE writer collapses line breaks in
quoted text as a format guard.
- The English no-number draft stamp ended 1.3pt below the top margin.
Tests lay the document out with @react-pdf/layout and assert that no text
node ends past the page edge and that every line's ink stays inside its
column, for 80-line descriptions and notes, a 3000-character description,
and four long tokens.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BuYbae3WWwYzBucTUVayoW
* fix(invoices): count rendered lines and cap the kept-together budget so no font can clip a row
Skeptic cycle 2: with the bundled Source Serif 4 font (13.7pt per line)
and a description of 20 source lines of wide glyphs, the estimate said
"fits" while the chunked token wrapping produced 58 rendered lines; the
non-splittable row then ran off the page. The estimate now counts the
chunks a long token is broken into, and the cap is 12 lines: at 20pt per
line that is under a third of the page for any font a company can pick.
The page-edge test now measures absolute positions (box.top is
parent-relative) and covers the bundled serif case through
prepareInvoiceFont().
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BuYbae3WWwYzBucTUVayoW
* fix(invoices): keep words whole by estimated ink width, not a character cap
Skeptic cycle 2 (regression): the flat 16-character cap gave react-pdf a
break point inside ordinary Swedish compounds (Fastighetsskötse-l,
Företagsförsäkri-ng), which the breaker used whenever it filled the line
better. Words are now kept whole whenever a rough per-glyph width estimate
says they fit the column; only a token wider than the column gets parts,
after separators and where the column is full. The page-fit estimate uses
the same widths.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BuYbae3WWwYzBucTUVayoW
* fix(invoices): estimate word width from real Helvetica metrics
Skeptic cycle 3: the hand-rolled per-glyph estimate over-counted ordinary
lowercase (6.5pt for a 5.56pt glyph), so 24 to 28 character compounds
(Momskompensationsansökan, Mervärdesskattedeklarationen) were still handed
a break point while they fit the column, and under-counted rare glyphs
(æ, œ, Cyrillic) so a token of those could overflow. The estimate now uses
Helvetica advances measured through react-pdf's own metrics
(lib/invoices/pdf-glyph-widths.ts) with a 10% margin for the bundled
fonts, and counts any glyph outside Helvetica at the widest Latin advance.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BuYbae3WWwYzBucTUVayoW
* docs(invoices): state the bundled-font width margin accurately
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BuYbae3WWwYzBucTUVayoW
* fix(invoices): review round: SIE backslash escaping, text-row budget, ROT/RUT box keep-together
- CodeQL: the SIE quoted-text escaper left backslashes alone while backslash
is the escape character. It now writes a literal backslash as two, and the
parser unescapes both that and the escaped quote, so text round-trips.
- CodeRabbit: a free-text row renders at full table width but was budgeted
at the description column, so it could split unnecessarily. It uses the
full-width budget now.
- CodeRabbit: the ROT/RUT box was unconditionally kept on one page although
its per-line breakdown carries the (possibly multi-line) descriptions. It
uses the same keep-together estimate as rows and notes, and its line texts
wrap whole words.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BuYbae3WWwYzBucTUVayoW
---------
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
42 lines
3.1 KiB
TypeScript
42 lines
3.1 KiB
TypeScript
/**
|
|
* Helvetica advance widths in 1/1000 em, keyed by code point, as measured
|
|
* through react-pdf's own pdfkit metrics (the WinAnsi glyph set plus the
|
|
* common typographic extras). Used to estimate whether a word fits its
|
|
* column before react-pdf lays the text out; see approxWidthPt in
|
|
* pdf-template.tsx.
|
|
*
|
|
* Generated once with `new PDFDocument().font('Helvetica').fontSize(1000)
|
|
* .widthOfString(ch)` for U+0020..U+017F and a few U+20xx code points.
|
|
* Glyphs Helvetica does not carry are omitted and fall back to a
|
|
* conservative default at the call site.
|
|
*/
|
|
export const HELVETICA_WIDTHS: ReadonlyMap<number, number> = new Map<number, number>([
|
|
[32, 278], [33, 278], [34, 355], [35, 556], [36, 556], [37, 889], [38, 667], [39, 191],
|
|
[40, 333], [41, 333], [42, 389], [43, 584], [44, 278], [45, 333], [46, 278], [47, 278],
|
|
[48, 556], [49, 556], [50, 556], [51, 556], [52, 556], [53, 556], [54, 556], [55, 556],
|
|
[56, 556], [57, 556], [58, 278], [59, 278], [60, 584], [61, 584], [62, 584], [63, 556],
|
|
[64, 1015], [65, 667], [66, 667], [67, 722], [68, 722], [69, 667], [70, 611], [71, 778],
|
|
[72, 722], [73, 278], [74, 500], [75, 667], [76, 556], [77, 833], [78, 722], [79, 778],
|
|
[80, 667], [81, 778], [82, 722], [83, 667], [84, 611], [85, 722], [86, 667], [87, 944],
|
|
[88, 667], [89, 667], [90, 611], [91, 278], [92, 278], [93, 278], [94, 469], [95, 556],
|
|
[96, 333], [97, 556], [98, 556], [99, 500], [100, 556], [101, 556], [102, 278], [103, 556],
|
|
[104, 556], [105, 222], [106, 222], [107, 500], [108, 222], [109, 833], [110, 556], [111, 556],
|
|
[112, 556], [113, 556], [114, 333], [115, 500], [116, 278], [117, 556], [118, 500], [119, 722],
|
|
[120, 500], [121, 500], [122, 500], [123, 334], [124, 260], [125, 334], [126, 584],
|
|
[160, 278], [161, 333], [162, 556], [163, 556], [164, 556], [165, 556], [166, 260], [167, 556],
|
|
[168, 333], [169, 737], [170, 370], [171, 556], [172, 584], [174, 737], [175, 333], [176, 400],
|
|
[177, 584], [178, 333], [179, 333], [180, 333], [181, 556], [182, 537], [183, 278], [184, 333],
|
|
[185, 333], [186, 365], [187, 556], [188, 834], [189, 834], [190, 834], [191, 611],
|
|
[192, 667], [193, 667], [194, 667], [195, 667], [196, 667], [197, 667], [198, 1000], [199, 722],
|
|
[200, 667], [201, 667], [202, 667], [203, 667], [204, 278], [205, 278], [206, 278], [207, 278],
|
|
[208, 722], [209, 722], [210, 778], [211, 778], [212, 778], [213, 778], [214, 778], [215, 584],
|
|
[216, 778], [217, 722], [218, 722], [219, 722], [220, 722], [221, 667], [222, 667], [223, 611],
|
|
[224, 556], [225, 556], [226, 556], [227, 556], [228, 556], [229, 556], [230, 889], [231, 500],
|
|
[232, 556], [233, 556], [234, 556], [235, 556], [236, 278], [237, 278], [238, 278], [239, 278],
|
|
[240, 556], [241, 556], [242, 556], [243, 556], [244, 556], [245, 556], [246, 556], [247, 584],
|
|
[248, 611], [249, 556], [250, 556], [251, 556], [252, 556], [253, 500], [254, 556], [255, 500],
|
|
[338, 1000], [339, 944], [352, 667], [353, 500], [376, 500], [381, 611], [382, 500],
|
|
[8211, 556], [8212, 1000], [8216, 222], [8217, 222], [8220, 333], [8221, 333], [8226, 350],
|
|
[8230, 1000], [8240, 1000], [8364, 556], [8482, 1000],
|
|
])
|