Copied from "ptksdict-1.1.6.zip", "\share\doc\Format-desc.txt"

# $RCSfile: Format-desc.txt,v $
# $Author: swaj $
# $Revision: 1.9 $
#
# Copyright (c) Alexey Semenoff 2001-2006. All rights reserved.
# Distributed under GNU Public License.
#

Sdict file structure
====================

Foreword
---------

File contains the following sections:

1. Header
2. Dictionary information like title, copyright and version.
3. Short index
4. Full index
5. Articles



uint16_t, uint32_t are little endian;
utf-32le is also little endian.

Articles, title, copyright, version are organized as a units.
Unit is universal storage container and looks like;

struct {
uint32_t record_length;
utf8 record;
}



Header
------

Structure, 43 (0x2b) byte length:

+--------+------------+-----------+--------------------------------------------+
| Offset | Len, bytes | Content   |              Description                   |
+--------+------------+-----------+--------------------------------------------+
| 0x0    | 4          | uint8_t[] | Signature, 'sdct'                          |
| 0x4    | 3          | uint8_t[] | Input language                             |
| 0x7    | 3          | uint8_t[] | Output language                            |
| 0xa    | 1          | uint8_t   | Compression method            : (bytes 0-3)|
|        |            |           |             and index levels  : (bytes 4-7)|
| 0xb    | 4          | uint32_t  | Amount of words                            |
| 0xf    | 4          | uint32_t  | Length of short index                      |
| 0x13   | 4          | uint32_t  | Offset of 'title' unit                     |
| 0x17   | 4          | uint32_t  | Offset of 'copyright' unit                 |
| 0x1b   | 4          | uint32_t  | Offset of 'version' unit                   |
| 0x1f   | 4          | uint32_t  | Offset of short index                      |
| 0x23   | 4          | uint32_t  | Offset of full index                       |
| 0x27   | 4          | uint32_t  | Offset of articles                         |
+--------+------------+-----------+--------------------------------------------+
<p/>
'short index', 'full index' and 'articles' are offsets from begin of the file.
Compression methods are '0' - none, '1' - gzip (Zlib), '2' -
bzip2. If some compression defined, the following sections expected
to be compressed: Short index, Articles.

Index levels value means how many short index levels are used. By
default it contains 0x3X which means 3 levels.



Note!  The only 3 levels are supported in all components, the other
ones are still experimental!



Dictionary information
----------------------

There are 3 sections here: 'title', 'copyright' and 'version' stored
as 3 units.
Offsets 'title unit', 'copyright unit' and 'version unit' are from
begin of the file.
There is no strict order of storing dictionary information, default
order is 'title', 'copyright', 'version'.



Short index
-----------

Short index is the set of records:

struct {
utf-32le[3] short_word;
uint32_t word_pointer;
}

thus size of each element is 12 (0xc) bytes.

Amount of records stored in header->'Length of short index'

word_pointer points to the whole word from 'full index' and it is
relative against begin of 'full index' section, not begin of the
file.



Full index
----------

Full index is set of the following records:

struct {
uint16_t next_word;
uint16_t previous_word;
uint32_t article_pointer;
utf8[]   word;
}

next_word and previous_word are relative against begin of the
record. article pointer points to article from 'articles' section
and it is relative against begin of 'articles' section, not begin
of the file.



Articles
--------

Articles are set of units, see unit description in foreword
chapter.
 *
Copied from "ptksdict-1.1.6.zip", "\share\dicts\README":
# $RCSfile: README,v $
# $Author: swaj $
# $Revision: 1.9 $
#
# Copyright (c) Alexey Semenoff 2001-2006. All rights reserved.
# Distributed under GNU Public License.
#
 *
HOW TO CREATE YOUR OWN DICTIONARY
 *
Look an examples in sample*.txt.
 *
 *
Every item looks like WORD___ARTICLE, no "\r", "\n" inside the article,
both WORD and ARTICLE are utf8-encoded text.
 *
Additionaly the following HTML-like tags can be used:
	<br>                    - "\n"
	<p>                     - "\n"
	<b> ... </b>            - use bold font
	<i> ... </i>            - italic font
	<u> ... </u>            - underline
	<l>, <li> ... , </l>    - list, like <ul><li> ...
	<r>word</r>             - reference to other word, like <a href="word">word</a>
	<t>trans</t>            - transcription	<t>trans</t>
	<f>forms</f>            - word forms	<f>forms</f>