.HWP File Format – Paragraphs and controls

Record tree

PARA_HEADER             level 0
  PARA_TEXT             level 1
  PARA_CHAR_SHAPE       level 1
  PARA_LINE_SEG         level 1
  PARA_RANGE_TAG        level 1   (optional)
  CTRL_HEADER 'tbl '    level 1   (one per extended control in the text)
    TABLE               level 2
    LIST_HEADER         level 2   (one per cell)
    PARA_HEADER         level 2   (the cell's paragraphs)
      PARA_TEXT         level 3
      …

HWPTAG_PARA_HEADER

24 bytes (table 58); files before 5.0.3.2 have the first 22 (the size in the specification’s table 5). Newer versions may make the record longer; readers should use the record size and ignore bytes they do not know. The break kinds are table 59. When the text length is 0, a reader makes the paragraph from a paragraph break alone.

HWPTAG_PARA_TEXT and control characters

The text is UTF-16LE. Code units 0 to 31 are control characters of three kinds: “char” controls take one unit, like an ordinary character; “inline” and “extended” controls take eight units (16 bytes): the code, six units of data (an extended control gives the four-character control ID there and points to a CTRL_HEADER child record), and the code again.

41 00 42 00 0D 00

The text of a paragraph “AB”: A, B and the paragraph break (13). An empty paragraph’s text is just 0D 00, and some writers then leave out the PARA_TEXT record.

HWPTAG_PARA_CHAR_SHAPE and HWPTAG_PARA_LINE_SEG

PARA_CHAR_SHAPE is a list of pairs of UINT32, as many as the header’s count: the position in the text where a run starts, and the index of the HWPTAG_CHAR_SHAPE it uses. The first run starts at 0.

PARA_LINE_SEG is the writer’s line layout, 36 bytes per line:

This is a cache of how the writer laid out each line; there is one entry per line segment, as many as the paragraph header says. Lengths are in HWPUNIT, 7,200 to the inch (1/100 point).

HWPTAG_PARA_RANGE_TAG

Marks ranges of text, such as highlighting and proofreading marks; unlike character shapes, ranges may overlap. One 12-byte entry per range, as many as the paragraph header says (table 63):

The kinds are not listed in the specification.

Control IDs

A CTRL_HEADER starts with a four-character ID, stored as a UINT32 whose bytes read backwards: tbl is stored 20 6C 62 74.

Drawing objects (gso ) have a second ID in their SHAPE_COMPONENT record saying which kind they are: $lin line, $rec rectangle, $ell ellipse, $arc arc, $pol polygon, $cur curve, $pic picture, $ole OLE object, $con group (container).

Field IDs

Fields (control code 3) have IDs starting with %:

Proofreading (revision) marks have IDs starting %%*: %%*d delete, %%*a attach, %%*C clipping, %%*S sawtooth, %%*T thinking, %%*P praise, %%*L line, %%*c simple change, %%*h hyperlink, %%*A line attach, %%*i line link, %%*t line transfer, %%*r move right, %%*l move left, %%*n transfer, %%*e simple insert, and %%mr change. These are the IDs of the specification’s table 128. The field control’s fields are on the controls page.