Search is working now

This commit is contained in:
Vincent Prouillet 2018-03-20 21:27:33 +01:00
parent ddf8970ad8
commit b0363298f5
7 changed files with 2116 additions and 37 deletions

1927
Cargo.lock generated Normal file

File diff suppressed because it is too large Load diff

View file

@ -4,7 +4,7 @@ version = "0.1.0"
authors = ["Vincent Prouillet <prouillet.vincent@gmail.com>"]
[dependencies]
elasticlunr-rs = "1"
elasticlunr-rs = "2"
ammonia = "1"
lazy_static = "1"

View file

@ -2,15 +2,17 @@ extern crate elasticlunr;
#[macro_use]
extern crate lazy_static;
extern crate ammonia;
#[macro_use]
extern crate errors;
extern crate content;
use std::collections::{HashMap, HashSet};
use std::path::PathBuf;
use elasticlunr::Index;
use elasticlunr::{Index, Language};
use content::Section;
use errors::Result;
pub const ELASTICLUNR_JS: &'static str = include_str!("elasticlunr.min.js");
@ -33,16 +35,23 @@ lazy_static! {
}
/// Returns the generated JSON index with all the documents of the site added
/// Returns the generated JSON index with all the documents of the site added using
/// the language given
/// Errors if the language given is not available in Elasticlunr
/// TODO: is making `in_search_index` apply to subsections of a `false` section useful?
pub fn build_index(sections: &HashMap<PathBuf, Section>) -> String {
let mut index = Index::new(&["title", "body"]);
pub fn build_index(sections: &HashMap<PathBuf, Section>, lang: &str) -> Result<String> {
let language = match Language::from_code(lang) {
Some(l) => l,
None => { bail!("Tried to build search index for language {} which is not supported", lang); }
};
let mut index = Index::with_language(language, &["title", "body"]);
for section in sections.values() {
add_section_to_index(&mut index, section);
}
index.to_json()
Ok(index.to_json())
}
fn add_section_to_index(index: &mut Index, section: &Section) {

View file

@ -21,13 +21,12 @@ extern crate search;
extern crate tempdir;
use std::collections::HashMap;
use std::fs::{remove_dir_all, copy};
use std::fs::{create_dir_all, remove_dir_all, copy};
use std::mem;
use std::path::{Path, PathBuf};
use glob::glob;
use tera::{Tera, Context};
use walkdir::WalkDir;
use sass_rs::{Options as SassOptions, OutputStyle, compile_file};
use errors::{Result, ResultExt};
@ -522,10 +521,10 @@ impl Site {
pub fn build_search_index(&self) -> Result<()> {
// index first
create_file(
&self.output_path.join("search_index.js"),
&self.output_path.join(&format!("search_index.{}.js", self.config.default_language)),
&format!(
"window.searchIndex = {};",
search::build_index(&self.sections)
search::build_index(&self.sections, &self.config.default_language)?
),
)?;

View file

@ -1,3 +1,30 @@
.search-container {
display: inline-block;
position: relative;
width: 300px;
input {
width: 100%;
}
}
.search-results {
display: none;
position: absolute;
background: white;
color: black;
padding: 1rem;
width: 100%;
box-shadow: 2px 2px 2px 0 rgba(0, 0, 0, 0.5);
max-height: 500px;
overflow: auto;
&__items {
list-style: none;
}
&__item {
margin-bottom: 1rem;
font-size: 0.9rem;
}
}

153
docs/static/search.js vendored
View file

@ -1,29 +1,141 @@
function formatSearchResultHeader(term, count) {
if (count === 0) {
return "No search results for '" + term + "'.";
}
function debounce(func, wait) {
var timeout;
return count + " search result" + count > 1 ? "s" : "" + " for '" + term + "':";
return function () {
var context = this;
var args = arguments;
clearTimeout(timeout);
timeout = setTimeout(function () {
timeout = null;
func.apply(context, args);
}, wait);
};
}
function formatSearchResultItem(term, item) {
console.log(item);
// Taken from mdbook
// The strategy is as follows:
// First, assign a value to each word in the document:
// Words that correspond to search terms (stemmer aware): 40
// Normal words: 2
// First word in a sentence: 8
// Then use a sliding window with a constant number of words and count the
// sum of the values of the words within the window. Then use the window that got the
// maximum sum. If there are multiple maximas, then get the last one.
// Enclose the terms in <b>.
function makeTeaser(body, terms) {
var TERM_WEIGHT = 40;
var NORMAL_WORD_WEIGHT = 2;
var FIRST_WORD_WEIGHT = 8;
var TEASER_MAX_WORDS = 30;
var stemmedTerms = terms.map(function (w) {
return elasticlunr.stemmer(w.toLowerCase());
});
var termFound = false;
var index = 0;
var weighted = []; // contains elements of ["word", weight, index_in_document]
// split in sentences, then words
var sentences = body.toLowerCase().split(". ");
for (var i in sentences) {
var words = sentences[i].split(" ");
var value = FIRST_WORD_WEIGHT;
for (var j in words) {
var word = words[j];
if (word.length > 0) {
for (var k in stemmedTerms) {
if (elasticlunr.stemmer(word).startsWith(stemmedTerms[k])) {
value = TERM_WEIGHT;
termFound = true;
}
}
weighted.push([word, value, index]);
value = NORMAL_WORD_WEIGHT;
}
index += word.length;
index += 1; // ' ' or '.' if last word in sentence
}
index += 1; // because we split at a two-char boundary '. '
}
if (weighted.length === 0) {
return body;
}
var windowWeights = [];
var windowSize = Math.min(weighted.length, TEASER_MAX_WORDS);
// We add a window with all the weights first
var curSum = 0;
for (var i = 0; i < windowSize; i++) {
curSum += weighted[i][1];
}
windowWeights.push(curSum);
for (var i = 0; i < weighted.length - windowSize; i++) {
curSum -= weighted[i][1];
curSum += weighted[i + windowSize][1];
windowWeights.push(curSum);
}
// If we didn't find the term, just pick the first window
var maxSumIndex = 0;
if (termFound) {
var maxFound = 0;
// backwards
for (var i = windowWeights.length - 1; i >= 0; i--) {
if (windowWeights[i] > maxFound) {
maxFound = windowWeights[i];
maxSumIndex = i;
}
}
}
var teaser = [];
var startIndex = weighted[maxSumIndex][2];
for (var i = maxSumIndex; i < maxSumIndex + windowSize; i++) {
var word = weighted[i];
if (startIndex < word[2]) {
// missing text from index to start of `word`
teaser.push(body.substring(startIndex, word[2]));
startIndex = word[2];
}
// add <em/> around search terms
if (word[1] === TERM_WEIGHT) {
teaser.push("<b>");
}
startIndex = word[2] + word[0].length;
teaser.push(body.substring(word[2], startIndex));
if (word[1] === TERM_WEIGHT) {
teaser.push("</b>");
}
}
teaser.push("…");
return teaser.join("");
}
function formatSearchResultItem(item, terms) {
return '<div class="search-results__item">'
+ item
+ `<a href="${item.ref}">${item.doc.title}</a>`
+ `<div>${makeTeaser(item.doc.body, terms)}</div>`
+ '</div>';
}
function initSearch() {
var $searchInput = document.getElementById("search");
var $searchResults = document.querySelector(".search-results");
var $searchResultsHeader = document.querySelector(".search-results__headers");
var $searchResultsItems = document.querySelector(".search-results__items");
var MAX_ITEMS = 10;
var options = {
bool: "AND",
expand: true,
teaser_word_count: 30,
limit_results: 30,
fields: {
title: {boost: 2},
body: {boost: 1},
@ -32,22 +144,25 @@ function initSearch() {
var currentTerm = "";
var index = elasticlunr.Index.load(window.searchIndex);
$searchInput.addEventListener("keyup", function() {
$searchInput.addEventListener("keyup", debounce(function() {
var term = $searchInput.value.trim();
if (!index || term === "" || term === currentTerm) {
if (term === currentTerm || !index) {
return;
}
$searchResults.style.display = term === "" ? "block" : "none";
$searchResults.style.display = term === "" ? "none" : "block";
$searchResultsItems.innerHTML = "";
if (term === "") {
return;
}
var results = index.search(term, options);
currentTerm = term;
$searchResultsHeader.textContent = searchResultText(term, results.length);
for (var i = 0; i < results.length; i++) {
for (var i = 0; i < Math.min(results.length, MAX_ITEMS); i++) {
var item = document.createElement("li");
item.innerHTML = formatSearchResult(results[i], term);
item.innerHTML = formatSearchResultItem(results[i], term.split(" "));
$searchResultsItems.appendChild(item);
}
});
}, 150));
}

View file

@ -18,15 +18,17 @@
<a class="white" href="{{ get_url(path="./documentation/_index.md") }}" class="nav-link">Docs</a>
<a class="white" href="{{ get_url(path="./themes/_index.md") }}" class="nav-link">Themes</a>
<a class="white" href="https://github.com/Keats/gutenberg" class="nav-link">GitHub</a>
<input id="search" type="search" placeholder="Search the docs">
<div class="search-container">
<input id="search" type="search" placeholder="Search the docs">
<div class="search-results">
<div class="search-results__items"></div>
</div>
</div>
</nav>
</header>
<div class="search-results">
<h2 class="search-results__header"></h2>
<div class="search-results__items"></div>
</div>
<div class="content {% block extra_content_class %}{% endblock extra_content_class %}">
{% block content %}
<div class="hero">
@ -101,7 +103,7 @@
</footer>
<script type="text/javascript" src="{{ get_url(path="elasticlunr.min.js", trailing_slash=false) }}"></script>
<script type="text/javascript" src="{{ get_url(path="search_index.js", trailing_slash=false) }}"></script>
<script type="text/javascript" src="{{ get_url(path="search_index.en.js", trailing_slash=false) }}"></script>
<script type="text/javascript" src="{{ get_url(path="search.js", trailing_slash=false) }}"></script>
</body>
</html>