.WPD File Format – Function codes

Byte ranges

In secondary merge files the end of a field is ^R (0x12) and a hard return; the document names the end-of-record character too, but it is lost from the published text. There is no start-of-field or start-of-record code.

Single-byte functions

0xB8 to 0xBB (octal 270–273) are not listed in the document. Spelling and grammar checkers need to allow for the hyphen codes A9–AE and the hard space A0.

Multiple-byte functions

A multiple-byte function is its code, its data and the code again. The length column is the whole function, both codes included; −1 means variable length, and a reader must scan for the closing code. Where data holds an “old value”, a writer may put zero; WordPerfect updates it. Margins and column positions are in character columns.

The document lists no codes above F8. libwpd treats F9–FE as variable length.

Header and footer

D1 old-def half-lines-used FF FF left-margin right-margin text… FF half-lines def D1

The definition byte’s low two bits are the type: 0 header A, 1 header B, 2 footer A, 3 footer B. Its high six bits say where it occurs: 0 never, 1 all pages, 2 odd pages, 4 even pages. The low two bits of the old definition byte must also be correct.

Footnotes and endnotes

E2 def a b c d old-footnote-lines lines-page-1 … lines-page-n pages FF left right text… E2

Definition byte: bit 0 is 0 for numbers, 1 for characters; bit 1 is 0 for a footnote, 1 for an endnote. With numbers, a b is the note number; with characters, a is how many and b the character. c d is the note’s number of lines. Both pairs are 14-bit numbers in two 7-bit bytes, high first. An endnote has only a null between d and FF.

The 74 bytes of footnote options (E3), counted from 1:

Paragraph numbering

Each definition byte of E5 and EE has the numbering style in its low nibble and the punctuation in its high nibble.

Marked text and generated lists

The definition/information byte of E7, E8 and E9 has the definition in its high nibble and the information in its low nibble: 0 table of contents (level 0–4), 1 index (E9 only; 0 no concordance file, 1 a file name is included), 2 list (list 0–4), 3 table of authorities (E9 only; section 0–15).

Definition bytes for contents, index and lists (contents uses all five, one per level; index and lists only the first): 0 no page numbers, 1 page number after the text after two spaces, 2 page number in brackets after one space, 3 page number flush right, 4 flush right with dot leader. For a table of authorities only the first byte counts, with bits xxxBxxDU: B a blank line between authorities, D dot leaders, U underlining allowed.

Tab reset

F1 old-tab-table(32) old-type-table(20) new-tab-table(32) new-type-table(20) F1

Each bit of a tab table is one column, bits 0 to 250. Each nibble of a type table is the type of the corresponding set bit (the fifth nibble for the fifth tab), up to 40 tabs; the rest are type 0.

Password protection

WordPerfect Corporation’s document describes unencrypted files only. libwpd recognises a protected 4.2 file by its first four bytes FE FF 61 61, followed by a 16-bit checksum of the password; the scrambled text starts at offset 6. No description of the encryption was ever made available by WordPerfect or its later owners (WordPerfect for DOS Updated); libwpd’s method was worked out independently. It is not secure and is easily reversed. libwpd handles a protected 4.2 file like this:

1. The password is turned to upper case (letters a to z only).
2. The checksum is built over the password’s characters: start at 0; for each character, rotate the 16-bit value right by one bit and XOR it with the character shifted left eight bits. A file is protected when its stored checksum is not 0, and the password is right when the checksums match.
3. The first 6 bytes (the marker and the checksum) are not encrypted.
4. The byte at position i after those 6 is XORed with the password character i mod length and with the mask (length + 1 + i) mod 256.

Recognising a 4.2 file

Without a header, libwpd checks the whole file: bytes below 0xC0 are accepted; 0xFF on its own rejects the file; each code from 0xC0 must be closed by the same code after its fixed length, or later for variable-length codes. Text that is valid UTF-8 is skipped, and a file with no WordPerfect functions at all is not accepted, so plain text is not mistaken for a document. libwpd tries the Macintosh 1.x layout first, which uses the same code ranges but different lengths.

Sources

• WordPerfect File Information, Version 4.2 – WordPerfect Corporation, 1984–1986; the format’s own description, kept on the “WordPerfect for DOS Updated” site
• WordPerfect for DOS Updated: software and file formats – notes that no documentation of encrypted files was ever available from WordPerfect or its later owners
• libwpd source – WP42FileStructure, WP42Parser, WP42Heuristics, WP42*Group and WPDocument.cpp