.HLP File Format – Text records

Windows Help file

The displayable records of |TOPIC keep their numbers and commands in LinkData1 and their text in LinkData2. Commands and strings take turns: a string (often empty), a command, a string, a command.

Compressed integers

Numbers in LinkData1 are shortened by a flag in the lowest bit of the first byte. Clear, the value fits one byte (or word); set, it takes twice that.

NameLow bit clearLow bit set
cshort1 byte: value / 2 − 0x402 bytes: value / 2 − 0x4000
cushort1 byte: value / 22 bytes: value / 2
clong2 bytes: value / 2 − 0x40004 bytes: value / 2 − 0x40000000
culong2 bytes: value / 24 bytes: value / 2

Record layout

LinkData1 starts as follows, by record type.

  1. All types: a clong, the topic size.
  2. Types 0x20 and 0x23: a cushort, TopicLength, the characters in this record. It is what the TOPICOFFSET count adds up.
  3. Type 0x23: a table header (u8 columns, u8 table type; if the type is 0 or 2, an i16 minimum width; then two i16 per column, the gap and the width).

Then paragraphs follow. Type 0x23 has a five-byte prefix for each cell (i16 column, i16 unknown, one zero byte); a column of −1 ends the table.

Paragraphs

Each paragraph starts with 4 bytes whose meaning is not known and a 16-bit bit field. For each bit that is set, one field follows, in this order:

BitFieldType
0x0001Unknownclong
0x0002Space abovecshort
0x0004Space belowcshort
0x0008Line spacingcshort
0x0010Left indentcshort
0x0020Right indentcshort
0x0040First-line indentcshort
0x0100Border3 bytes: flags, then a 16-bit width
0x0200Tab stopsA cshort count, then each stop as a cushort, followed by another cushort (the tab type) when bit 0x4000 of the stop is set

Bits 0x0400 and 0x0800 right-align and centre the paragraph. Metrics are in half-points or twips, depending on the font table’s type.

Formatting commands

After the paragraph fields the string and command stream begins. Reading stops at 0xFF; other commands are single bytes with these parameters.

CodeMeaningParameters
0x20{vfld} marker (Multimedia Viewer)u32 number
0x21{dtype} marker (Multimedia Viewer)u16 number
0x80Font changeu16 index into the descriptors
0x81Line breaknone
0x82End of paragraph (or table cell)none
0x83Tabnone
0x86, 0x87, 0x88Picture or embedded object: in line, left, rightsee below
0x89End of a hotspotnone
0x8BNon-breaking spacenone
0x8CNon-breaking hyphennone
0xC8, 0xCCMacro hotspotu16 length, then the macro text
0xE0–0xE3, 0xE6, 0xE7Popup (even) or jump (odd) to a topic. 0xE6 and 0xE7 have no underlineu32 target
0xEA, 0xEB, 0xEE, 0xEFPopup (even) or jump (odd) to another file or windowsee below
0xFFEnd of the commands for this paragraphnone

A jump’s target depends on the compiler. Windows 3.0 stores a topic number, which |TOMAP turns into a position. Windows 3.1 and later store the 32-bit hash of the context name, looked up in |CONTEXT.

Jumps to other files

0xEA and its neighbours are followed by a u16 length, a u8 type and a u32 hash of the context name, then, by type: nothing (0), a one-byte window number (1), or a file name (4), or a file name and a window name (6), each as a NUL-terminated string.

Pictures and embedded objects

0x86–0x88 are followed by a u8 type, then a clong size. Type 0x22 has a cushort hotspot count next. Then, for size bytes:

  • Type 3 or 0x22: an i16 that is 0 for a picture stored in |bmN, followed by a u16 number, or 1 when the picture is stored here.
  • Type 5: three i16 values, then a string. It starts ! for a button (label, macro), * for sound or video, and is otherwise a window from a DLL.