Go to file
Alex Yatskov 1a3feb047a Update README 2023-12-30 14:32:26 -08:00
eb@c9d1015f9e . 2017-04-01 14:11:12 -07:00
img Add site metadata 2021-12-14 20:20:53 -08:00
jansson@bc5741fb1a Json export 2016-11-02 20:06:40 -07:00
.gitignore Cleanup 2016-11-02 20:40:44 -07:00
.gitmodules WIP 2016-12-03 18:24:56 -08:00
CMakeLists.txt Don't install eb with make 2019-06-26 22:49:47 +02:00
LICENSE update license file 2019-04-28 11:48:18 -07:00
README.md Update README 2023-12-30 14:32:26 -08:00
book.c Update max definition size to 1MB from 10KB 2018-05-21 17:53:41 -07:00
book.h updating license 2017-04-23 12:32:22 -07:00
convert.c updating license 2017-04-23 12:32:22 -07:00
convert.h updating license 2017-04-23 12:32:22 -07:00
hooks.c updating license 2017-04-23 12:32:22 -07:00
hooks.h updating license 2017-04-23 12:32:22 -07:00
main.c Fix argument requirement in getopt_long 2019-06-26 23:22:20 +02:00
util.h updating license 2017-04-23 12:32:22 -07:00

README.md

Zero-EPWING

Note: this project is no longer maintained. Please see this post for more information.

Zero-EPWING is a tool built to export easy to process JSON formatted UTF-8 data from dictionaries in EPWING format. This is a terrible format for many reasons, some of which are outlined below:

  • It is based on a closed and undocumented standard.
  • Not well supported as it isn't used anywhere else in the world.
  • The only library for parsing this format, libeb, is abandoned.
  • Data is stored in an inconsistent manner, with lots of duplication.
  • Text data is represented using the annoying EUC-JP encoding.
  • Characters which cannot be encoded are represented by image bitmaps.

Most applications that parse EPWING data traditionally use libeb to perform dictionary searches in place; dealing with quirks in the format and libeb output is just part of the process. Zero-EPWING takes a different approach -- extract all the data and output it an sane intermediate format, like JSON. As everyone knows how to parse JSON, it is trivial to take this intermediate data and store it in a reasonable, industry standard representation.

Building

Prepare your development environment by making sure the following tools are set up:

Once your system is configured, follow the steps below to create builds:

  1. Clone the repository by executing
    git clone --recurse-submodules https://git.foosoft.net/alex/zero-epwing
    
  2. Prepare the project. From the project root directory, execute
    cmake . -Bbuild && cmake --build build --
    
  3. Find the executable in the build directory.

Usage

Zero-EPWING takes a single parameter, the directory of the EPWING dictionary to dump. It also supports the following optional flags:

  • --entries (-e): output dictionary entry data (most common option).
  • --fonts (-f): output output font bitmap data (useful for OCR).
  • --markup (-m): markup the output with as much metadata as possible.
  • --positions (-s): output page and offset data for each entry.
  • --pretty (-p): output pretty-printed JSON (useful for debugging).

Upon loading and processing the requested EPWING data, Zero-EPWING will output a UTF-8 encoded JSON file to stdout. Diagnostic information about errors will be printed to stderr. Serious errors will result in this application returning a non-zero exit code. A sample of the JSON dictionary entry data output is pretty-printed below for reference.

{
    "charCode": "jisx0208",
    "discCode": "epwing",
    "subbooks": [
        {
            "title": "大辞泉",
            "copyright": "CD-ROM版大辞泉 1997年4月10日 第1版発行\n\n監  松村 明\n発行者 鈴木俊彦\n発行所...",
            "entries": [
                {
                    "heading": "亜",
                    "text": "亜\nア\nぐ\n部首二\n総画数\nコード区点..."
                },
                {
                    "heading": "あ",
                    "text": "あ\n{{w_50275}}\n{{w_50035}}五十音図ア行の第一音。五母音の一。後舌の開母音..."
                }
            ]
        }
    ]
}

You may have noticed the unusual-looking double curly bracket markers such as {{w_50035}}. Remember what I mentioned about certain characters being represented by image files? There are two graphical fonts sets in each dictionary: narrow and wide. Both of these font sets are available in four sizes: 24, 30, 36, and 48 pixels. Whenever a character cannot be encoded as text, a glyph is used in its place. These font indices cannot be converted directly to characters, differ from one dictionary to another, and must be manually mapped to Unicode character tables. Zero-EWPING has no facility to map these font glyphs to Unicode by itself, and instead places inline markers in the form of {{w_xxxx}} and {{n_xxxx}} in the output, specifying the referenced indices of the wide or narrow fonts respectively.

The bitmaps for these font glyphs can be dumped by executing this application with the --fonts command line argument.