.HWP File Format – Tables, pictures and drawings

Object controls

The drawing objects, pictures, OLE objects and groups are generic shape objects: the specification says their ID is written twice. pyhwp reads gso  as the ID in the control header and the specific ID ($pic and so on) at the start of the shape component. The records nest like this:

CTRL_HEADER 'gso '                level 1  common object properties (+ caption)
  SHAPE_COMPONENT                 level 2  '$rec', element properties, line, fill, text box
    LIST_HEADER                   level 3  (text box only) and its paragraphs
    SHAPE_COMPONENT_RECTANGLE     level 3  the shape's own fields
CTRL_HEADER 'tbl '                level 1  common object properties
  TABLE                           level 2  table properties
  LIST_HEADER + cell properties   level 2  one per cell, followed by its paragraphs

Common object properties

The data of the object’s CTRL_HEADER (table 69), followed by a caption list if the object has a caption (table 68).

46 + 2 × len bytes.

Caption

A caption is a list header followed by these fields (tables 71 and 72), then its paragraphs:

14 bytes; the specification’s table 71 counts 12.

HWPTAG_TABLE

22 + 2r + 10z bytes. The meaning of the row sizes is not spelt out; pyhwp names the array rowcols (cells per row), and reads the zones from version 5.0.0.7.

Cells

Each cell is a list header whose record goes on with these 26 bytes (tables 79 and 80), followed by the cell’s paragraphs:

HWPTAG_SHAPE_COMPONENT

Every drawing object (table 81) is its element properties, then an 11-byte line, then fill information, then, if it holds text, a 12-byte text box with a list header: 23 + n + n2 bytes. The element properties start with the object’s UINT32 control ID (table 82), then (table 83):

Rendering information

Each matrix is a 3 × 3 matrix of doubles whose last row is always 0, 0, 1, so only the first two rows are stored: six 8-byte doubles (table 85). 50 + 96 × cnt bytes. The order of the six elements is not stated.

Line

Text box

Drawing shapes

Each shape’s own record is a child of its shape component.

HWPTAG_SHAPE_COMPONENT_LINE

HWPTAG_SHAPE_COMPONENT_RECTANGLE

HWPTAG_SHAPE_COMPONENT_ELLIPSE

Properties (table 97): bit 0 the interval (the distance between two points on the circle) must be recalculated when it becomes an arc; bit 1 it has become an arc; bits 2–9 the kind of arc (values not listed).

HWPTAG_SHAPE_COMPONENT_ARC

HWPTAG_SHAPE_COMPONENT_POLYGON

An INT16 number of points cnt, then INT32[cnt] X coordinates, then INT32[cnt] Y coordinates: 2 + 8 × cnt bytes (table 99).

HWPTAG_SHAPE_COMPONENT_CURVE

As the polygon, then BYTE[cnt − 1] segment types, one per segment: 0 straight line, 1 curve (table 103).

HWPTAG_EQEDIT

The equation control eqed has the common object properties and this record (table 105):

The script language is compatible with EQN scripts; Hancom describes it in a separate equation specification, not covered here. The version and font strings are probably each preceded by their own length, as the total 16 + 6 × len suggests, but the table does not show it.

HWPTAG_SHAPE_COMPONENT_PICTURE

A picture is the common properties, the shape component and this record (tables 106 and 107):

Table 116 adds further picture properties: HWPUNIT × 2, the size of the image when the picture was created, and an INT8 image transparency, 9 bytes. Where they sit in the record is not stated.

Picture effects

A UINT32 saying which effects are present – shadow, glow (“neon”), soft edge and reflection – then each present effect (table 108). The bit for each effect is not given.

The shadow’s fields are typed in the order INT32, then four floats, INT32, four floats, INT32, colour.

Colour

The total is given as 4 + m (4 to 12) + n, which suggests that only the fields for the colour type are stored; the specification does not say which, nor list the colour type values.

HWPTAG_SHAPE_COMPONENT_OLE

The properties are typed UINT16, but bits 16–21 need a wider field; how they are stored is not explained.

Charts

A chart is stored in the text as an OLE object; its contents are an OLE compound file whose top storage can hold two streams:

• Contents – Hangul’s own chart format, the only stream in charts from Hangul Neo and older. Hancom describes it in a separate chart specification (in Korean): a sequence of chart objects, each a long ID, a long stored type ID, the stored name (char*) and stored version (int) – left out when the same type ID has already appeared – and the object’s data, which may contain further chart objects. The tree starts at a VtChart object with Backdrop, DataGrid, Footnote, Legend, Plot, PrintInformation and Title objects; the chart specification lists the properties of 63 object types and 52 tables of constants.
• OOXMLChartContents – from Hangul 2018, an Office Open XML chart: the root element is chartSpace, as in the DrawingML chart part of ECMA-376 and ISO/IEC 29500. Hangul 2018 also writes Contents as a fallback, converted to the old chart format.

The HWPTAG_CHART_DATA record is 2 bytes; its meaning is not given. The chart objects, their properties and constants are on the charts page.

HWPTAG_SHAPE_COMPONENT_CONTAINER

A group (table 120) is the common properties, then a WORD number of objects n and UINT32[n] control IDs of the grouped objects (table 121), then the shape records of each object in turn. Drawing objects, OLE objects, pictures and other groups can be grouped.

HWPTAG_VIDEO_DATA

A video (the specification heads it HWPTAG_VIDEO_TDATA) is the common properties and (table 123) an INT32 video type, 0 local or 1 web (table 124), then:

The file header’s bit 16 says a document contains video controls.

Not described

Text art (HWPTAG_SHAPE_COMPONENT_TEXTART), form objects (HWPTAG_FORM_OBJECT), memo list headers (HWPTAG_MEMO_LIST) and HWPTAG_SHAPE_COMPONENT_UNKNOWN (36 bytes) are named in the specification without a layout; their fields are unknown here.

Sources

• Hangul document file format 5.0, revision 1.3 – Hancom’s specification (PDF, in Korean), section 4.3.9
• Hangul document file format: charts, revision 1.2 – Hancom’s chart specification (PDF, in Korean)
• pyhwp – open-source reader, for the control ID of generic shapes (Python)