This PR merges the use of lunr search instead of the default mdbook search to enable support of JP characters.
69 lines
2.6 KiB
JavaScript
69 lines
2.6 KiB
JavaScript
// Modified version of this code:
|
||
// https://github.com/rust-lang/mdBook/issues/1081#issuecomment-2030246067
|
||
|
||
// Required for `elasticlunr.multiLanguage()` to work.
|
||
// This is something lunr.jp should implement
|
||
// but it doesn't.
|
||
// copied from: https://github.com/MihaiValentin/lunr-languages/blob/master/lunr.ja.js
|
||
elasticlunr.jp.wordCharacters = "一二三四五六七八九十百千万億兆一-龠々〆ヵヶぁ-んァ-ヴーア-ン゙a-zA-Za-zA-Z0-90-9";
|
||
|
||
var re_ENchar = new RegExp("[" + elasticlunr.wordCharacters + "]")
|
||
|
||
// Patch load method, which is used by searcher.js.
|
||
// This swaps the prebuilt index with a new index
|
||
// that supports JP keyword search.
|
||
// Since this happens on client side, it may increase load time.
|
||
elasticlunr.Index.load = function (index) {
|
||
// `index` contains the index built by mdbook during build.
|
||
// It also contains all raw page texts so we'll use them
|
||
// to rebuild a new index.
|
||
|
||
return elasticlunr(function () {
|
||
this.use(elasticlunr.multiLanguage('en', 'jp'));
|
||
|
||
var en_tokenizer = elasticlunr.tokenizer;
|
||
|
||
// first tokenize with the English tokenizer, and then
|
||
// only apply the Japanese tokenizer on tokens that
|
||
// don't look like English.
|
||
//
|
||
// Why should we not use just the JP tokenizer?
|
||
// Because a query like "they're" would be
|
||
// tokenized into "they" and "re" and
|
||
// EN stopWordFilter removes "they" and
|
||
// you're left with "re" which would match
|
||
// any words that start with re.
|
||
elasticlunr.tokenizer = function (obj) {
|
||
var result = [];
|
||
var en_tokens = en_tokenizer(obj);
|
||
for (var i = 0; i < en_tokens.length; i++) {
|
||
if (en_tokens[i].match(re_ENchar)) {
|
||
result.push(en_tokens[i]);
|
||
} else {
|
||
var jp_tokens = elasticlunr.jp.tokenizer(en_tokens[i]);
|
||
result.push.apply(result, jp_tokens);
|
||
}
|
||
}
|
||
return result;
|
||
};
|
||
|
||
// At least the `seperator` property is required for
|
||
// EN tokenizer to work correctly
|
||
for (var prop in en_tokenizer) {
|
||
elasticlunr.tokenizer[prop] = en_tokenizer[prop]
|
||
}
|
||
|
||
// fields to index
|
||
this.addField('title');
|
||
this.addField('body');
|
||
this.addField('breadcrumbs');
|
||
|
||
// Identify documents field
|
||
this.setRef('id');
|
||
|
||
// Build index of all pages
|
||
for (let key in index.documentStore.docs) {
|
||
this.addDoc(index.documentStore.docs[key]);
|
||
}
|
||
});
|
||
};
|