mirror of
https://github.com/xiaoyifang/goldendict-ng.git
synced 2024-11-24 00:14:06 +00:00
31 lines
1.7 KiB
Markdown
31 lines
1.7 KiB
Markdown
## Index file
|
|
|
|
Each index file have 4 sections.
|
|
|
|
1. `IdxHeader`
|
|
2. `ExtraInfo` (Being used but unnamed in source code)
|
|
3. `Chunks`
|
|
4. `BtreeIndex`
|
|
|
|
The `IdxHeader` are 32bits blocks of various meta info of the index. The most important info are `chunksOffset` and `indexRootOffset` pointing to the starting offset of `BtreeIndex` and `Chunks`.
|
|
|
|
Some dicts only have one `ExtraInfo`: the `dictionaryName` which is an uint32 size of a string followed and the string.
|
|
|
|
Each chunk contains uint32 size of uncompressed data, uint32 size of zlib compressed data, and the zlib compressed data.
|
|
|
|
The `Chunks` maybe used by both `IdxHeader` and `BtreeIndex`.
|
|
|
|
By adding new a new chunk to `Chunks` and store an offset to `IdxHeader`, `ExtraInfo` can store arbitrary long information.
|
|
|
|
`BtreeIndex` is a zlib compressed typical btree implementation in which each Node will include `word` info and a `offset` that pointing to corresponding `chunk`'s position.
|
|
|
|
Note that a `chunk` only includes necessary data to find an article, and it does not contain the `word`.
|
|
|
|
The exact data in `chunk` is decided and interpreted by dictionary implementations. For example, the starting and ending position of an article in a dictionary file.
|
|
|
|
## What's under the hood after a word is queried?
|
|
|
|
After typing a word into the search box and press enter, the embedded browser will load `gdlookup://localhost?word=<wantted word>`. This url will be handled by Qt webengine's Url Scheme handler. The returned html page will be composed in the ArticleMaker which will initiate some DataRequest on dictionary formats. Resource files will be requested via `bres://` or `qrc://` which will went through a similar process.
|
|
|
|
TODO: other subsystems.
|