searchengines-web/MeiliSearch
searchengines-web/MeiliSearch
1
0
Fork 0
mirror of https://github.com/meilisearch/MeiliSearch synced 2024-11-22 12:54:26 +01:00
Code Issues Releases Wiki Activity

Reintroduce a simple HTTP server

Browse Source
This commit is contained in:
Kerollmops 2020-05-31 17:48:13 +02:00
parent 2a10b2275e
commit a26553c90a
No known key found for this signature in database
GPG Key ID: 92ADA4E935E71FA4
9 changed files with 1458 additions and 83 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1048
Cargo.lock generated
View File

File diff suppressed because it is too large Load Diff

5
Cargo.toml
View File

@ -26,5 +26,10 @@ smallvec = "1.4.0"
structopt = { version = "0.3.14", default-features = false }
tempfile = "3.1.0"
# http server
serde = { version = "1.0", features = ["derive"] }
tokio = { version = "0.2.15", features = ["full"] }
warp = "0.2.2"
[profile.release]
debug = true

1
public/bulma.min.css vendored Normal file
View File

File diff suppressed because one or more lines are too long

199
public/index.html Normal file
View File

@ -0,0 +1,199 @@
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<link rel="stylesheet" href="/bulma.min.css">
<script type="text/javascript" src="/jquery-3.4.1.min.js"></script>
<script type="text/javascript" src="/papaparse.min.js"></script>
<title>The daugt</title>
<style>
em {
color: hsl(204, 86%, 25%);
font-style: inherit;
background-color: hsl(204, 86%, 88%);
}
#results {
max-width: 900px;
margin: 20px auto 0 auto;
padding: 0;
}
.notification {
display: flex;
justify-content: center;
}
.level-left {
margin-right: 50px;
}
.document {
padding: 20px 20px;
background-color: #f5f5f5;
border-radius: 4px;
margin-bottom: 20px;
display: flex;
}
.document ol {
flex: 0 0 75%;
max-width: 75%;
padding: 0;
margin: 0;
}
.document .image {
max-width: 25%;
flex: 0 0 25%;
padding-left: 30px;
box-sizing: border-box;
}
.document .image img {
width: 100%;
}
.field {
list-style-type: none;
display: flex;
flex-wrap: wrap;
}
.field:not(:last-child) {
margin-bottom: 7px;
}
.attribute {
flex: 0 0 35%;
max-width: 35%;
text-align: right;
padding-right: 10px;
box-sizing: border-box;
text-transform: uppercase;
color: rgba(0,0,0,.7);
}
.content {
max-width: 65%;
flex: 0 0 65%;
box-sizing: border-box;
padding-left: 10px;
color: rgba(0,0,0,.9);
}
</style>
</head>
<body>
<section class="hero is-light">
<div class="hero-body">
<div class="container">
<h1 class="title">
Welcome to daugt
</h1>
<h2 class="subtitle">
This dashboard will help you check the search results with ease.
</h2>
</div>
</div>
</section>
<section class="hero container">
<div class="notification" style="border-radius: 0 0 4px 4px;">
<nav class="level">
<!-- Left side -->
<div class="level-left">
<div class="level-item">
<div class="field has-addons has-addons-right">
<input id="search" class="input" type="text" autofocus placeholder="e.g. George Clooney">
</div>
</div>
</div>
<!-- Right side -->
<nav class="level-right">
<div class="level-item has-text-centered">
<div>
<p class="heading">Documents</p>
<p id="count" class="title">25</p>
</div>
</div>
<div class="level-item has-text-centered">
<div>
<p class="heading">Time Spent</p>
<p id="time" class="title">4ms</p>
</div>
</div>
</nav>
</nav>
</div>
</section>
<section>
<ol id="results" class="content">
<!-- documents matching requests -->
</ol>
</section>
</body>
<script>
var request = null;
$('#search').on('input', function () {
var query = $(this).val();
request = $.ajax({
type: "POST",
url: "query",
contentType: 'application/json',
data: JSON.stringify({ 'query': query }),
contentType: 'application/json',
success: function (data, textStatus, request) {
let httpResults = Papa.parse(data, { header: true, skipEmptyLines: true });
results.innerHTML = '';
let timeSpent = request.getResponseHeader('Time-Ms');
let numberOfDocuments = httpResults.data.length;
count.innerHTML = `${numberOfDocuments}`;
time.innerHTML = `${timeSpent}ms`;
for (element of httpResults.data) {
const elem = document.createElement('li');
elem.classList.add("document");
const ol = document.createElement('ol');
for (const prop in element) {
const field = document.createElement('li');
field.classList.add("field");
const attribute = document.createElement('div');
attribute.classList.add("attribute");
attribute.innerHTML = prop;
const content = document.createElement('div');
content.classList.add("content");
content.innerHTML = element[prop];
field.appendChild(attribute);
field.appendChild(content);
ol.appendChild(field);
}
elem.appendChild(ol);
results.appendChild(elem)
}
},
beforeSend: function () {
if (request !== null) {
request.abort();
}
},
});
});
</script>
</html>

2
public/jquery-3.4.1.min.js vendored Normal file
View File

File diff suppressed because one or more lines are too long

7
public/papaparse.min.js vendored Executable file
View File

File diff suppressed because one or more lines are too long

78
src/bin/search.rs
View File

@ -2,16 +2,9 @@ use std::io::{self, Write};
use std::path::PathBuf;
use std::time::Instant;
use cow_utils::CowUtils;
use fst::{Streamer, IntoStreamer};
use heed::types::*;
use heed::{EnvOpenOptions, Database};
use levenshtein_automata::LevenshteinAutomatonBuilder;
use roaring::RoaringBitmap;
use heed::EnvOpenOptions;
use structopt::StructOpt;
use mega_mini_indexer::alphanumeric_tokens;
use mega_mini_indexer::BEU32;
use mega_mini_indexer::{Index, BEU32};
#[derive(Debug, StructOpt)]
#[structopt(name = "mm-indexer", about = "The server side of the daugt project.")]
@ -35,78 +28,27 @@ fn main() -> anyhow::Result<()> {
.max_dbs(5)
.open(opt.database)?;
let main = env.create_poly_database(None)?;
let postings_ids: Database<Str, ByteSlice> = env.create_database(Some("postings-ids"))?;
let documents: Database<OwnedType<BEU32>, ByteSlice> = env.create_database(Some("documents"))?;
let index = Index::new(&env)?;
let before = Instant::now();
let rtxn = env.read_txn()?;
let headers = match main.get::<_, Str, ByteSlice>(&rtxn, "headers")? {
let documents_ids = index.search(&rtxn, &opt.query)?;
let headers = match index.headers(&rtxn)? {
Some(headers) => headers,
None => return Ok(()),
};
let fst = match main.get::<_, Str, ByteSlice>(&rtxn, "words-fst")? {
Some(bytes) => fst::Set::new(bytes)?,
None => return Ok(()),
};
// Building these factories is not free.
let lev0 = LevenshteinAutomatonBuilder::new(0, true);
let lev1 = LevenshteinAutomatonBuilder::new(1, true);
let lev2 = LevenshteinAutomatonBuilder::new(2, true);
let words: Vec<_> = alphanumeric_tokens(&opt.query).collect();
let number_of_words = words.len();
let dfas = words.into_iter().enumerate().map(|(i, word)| {
let word = word.cow_to_lowercase();
let is_last = i + 1 == number_of_words;
let dfa = match word.len() {
0..=4 => if is_last { lev0.build_prefix_dfa(&word) } else { lev0.build_dfa(&word) },
5..=8 => if is_last { lev1.build_prefix_dfa(&word) } else { lev1.build_dfa(&word) },
_ => if is_last { lev2.build_prefix_dfa(&word) } else { lev2.build_dfa(&word) },
};
(word, dfa)
});
let before = Instant::now();
let mut intersect_result: Option<RoaringBitmap> = None;
for (word, dfa) in dfas {
let before = Instant::now();
let mut union_result = RoaringBitmap::default();
let mut stream = fst.search(dfa).into_stream();
while let Some(word) = stream.next() {
let word = std::str::from_utf8(word)?;
if let Some(ids) = postings_ids.get(&rtxn, word)? {
let right = RoaringBitmap::deserialize_from(ids)?;
union_result.union_with(&right);
}
}
eprintln!("union for {:?} took {:.02?}", word, before.elapsed());
intersect_result = match intersect_result.take() {
Some(mut left) => {
let before = Instant::now();
let left_len = left.len();
left.intersect_with(&union_result);
eprintln!("intersect between {:?} and {:?} took {:.02?}",
left_len, union_result.len(), before.elapsed());
Some(left)
},
None => Some(union_result),
};
}
let mut stdout = io::stdout();
stdout.write_all(&headers)?;
let total_length = intersect_result.as_ref().map_or(0, |x| x.len());
for id in intersect_result.unwrap_or_default().iter().take(20) {
if let Some(content) = documents.get(&rtxn, &BEU32::new(id))? {
for id in &documents_ids {
if let Some(content) = index.documents.get(&rtxn, &BEU32::new(*id))? {
stdout.write_all(&content)?;
}
}
eprintln!("Took {:.02?} to find {} documents", before.elapsed(), total_length);
eprintln!("Took {:.02?} to find {} documents", before.elapsed(), documents_ids.len());
Ok(())
}

115
src/bin/serve.rs Normal file
View File

@ -0,0 +1,115 @@
use std::net::SocketAddr;
use std::path::PathBuf;
use std::str::FromStr;
use std::time::Instant;
use heed::EnvOpenOptions;
use serde::Deserialize;
use structopt::StructOpt;
use warp::{Filter, http::Response};
use mega_mini_indexer::{BEU32, Index};
#[derive(Debug, StructOpt)]
#[structopt(name = "mmi", about = "The server side of the mmi project.")]
struct Opt {
/// The database path where the LMDB database is located.
/// It is created if it doesn't already exist.
#[structopt(long = "db", parse(from_os_str))]
database: PathBuf,
/// The maximum size the database can take on disk. It is recommended to specify
/// the whole disk space (value must be a multiple of a page size).
#[structopt(long = "db-size", default_value = "107374182400")] // 100 GB
database_size: usize,
/// The ip and port on which the database will listen for HTTP requests.
#[structopt(short = "l", long, default_value = "127.0.0.1:9700")]
http_listen_addr: String,
}
#[tokio::main]
async fn main() -> anyhow::Result<()> {
let opt = Opt::from_args();
std::fs::create_dir_all(&opt.database)?;
let env = EnvOpenOptions::new()
.map_size(opt.database_size)
.max_dbs(10)
.open(&opt.database)?;
let index = Index::new(&env)?;
// We run and wait on the HTTP server
// Expose an HTML page to debug the search in a browser
let dash_html_route = warp::filters::method::get()
.and(warp::filters::path::end())
.map(|| warp::reply::html(include_str!("../../public/index.html")));
let dash_bulma_route = warp::filters::method::get()
.and(warp::path!("bulma.min.css"))
.map(|| Response::builder()
.header("content-type", "text/css; charset=utf-8")
.body(include_str!("../../public/bulma.min.css"))
);
let dash_jquery_route = warp::filters::method::get()
.and(warp::path!("jquery-3.4.1.min.js"))
.map(|| Response::builder()
.header("content-type", "application/javascript; charset=utf-8")
.body(include_str!("../../public/jquery-3.4.1.min.js"))
);
let dash_papaparse_route = warp::filters::method::get()
.and(warp::path!("papaparse.min.js"))
.map(|| Response::builder()
.header("content-type", "application/javascript; charset=utf-8")
.body(include_str!("../../public/papaparse.min.js"))
);
#[derive(Deserialize)]
struct QueryBody {
query: String,
}
let env_cloned = env.clone();
let query_route = warp::filters::method::post()
.and(warp::path!("query"))
.and(warp::body::json())
.map(move |query: QueryBody| {
let before_search = Instant::now();
let rtxn = env_cloned.read_txn().unwrap();
let documents_ids = index.search(&rtxn, &query.query).unwrap();
let mut body = Vec::new();
if let Some(headers) = index.headers(&rtxn).unwrap() {
// We write the headers
body.extend_from_slice(headers);
for id in documents_ids {
if let Some(content) = index.documents.get(&rtxn, &BEU32::new(id)).unwrap() {
body.extend_from_slice(&content);
}
}
}
Response::builder()
.header("Content-Type", "text/csv")
.header("Time-Ms", before_search.elapsed().as_millis().to_string())
.body(String::from_utf8(body).unwrap())
});
let routes = dash_html_route
.or(dash_bulma_route)
.or(dash_jquery_route)
.or(dash_papaparse_route)
.or(query_route);
let addr = SocketAddr::from_str(&opt.http_listen_addr).unwrap();
eprintln!("listening on http://{}", addr);
warp::serve(routes).run(addr).await;
Ok(())
}

86
src/lib.rs
View File

@ -1,7 +1,14 @@
use std::collections::HashMap;
use std::hash::BuildHasherDefault;
use std::time::Instant;
use cow_utils::CowUtils;
use fst::{IntoStreamer, Streamer};
use fxhash::FxHasher32;
use heed::types::*;
use heed::{PolyDatabase, Database};
use levenshtein_automata::LevenshteinAutomatonBuilder;
use roaring::RoaringBitmap;
use slice_group_by::StrGroupBy;
pub type FastMap4<K, V> = HashMap<K, V, BuildHasherDefault<FxHasher32>>;
@ -14,3 +21,82 @@ pub fn alphanumeric_tokens(string: &str) -> impl Iterator<Item = &str> {
let is_alphanumeric = |s: &&str| s.chars().next().map_or(false, char::is_alphanumeric);
string.linear_group_by_key(|c| c.is_alphanumeric()).filter(is_alphanumeric)
}
#[derive(Clone)]
pub struct Index {
pub main: PolyDatabase,
pub postings_ids: Database<Str, ByteSlice>,
pub documents: Database<OwnedType<BEU32>, ByteSlice>,
}
impl Index {
pub fn new(env: &heed::Env) -> heed::Result<Index> {
let main = env.create_poly_database(None)?;
let postings_ids = env.create_database(Some("postings-ids"))?;
let documents = env.create_database(Some("documents"))?;
Ok(Index {
main,
postings_ids,
documents,
})
}
pub fn headers<'t>(&self, rtxn: &'t heed::RoTxn) -> heed::Result<Option<&'t [u8]>> {
self.main.get::<_, Str, ByteSlice>(rtxn, "headers")
}
pub fn search(&self, rtxn: &heed::RoTxn, query: &str) -> anyhow::Result<Vec<DocumentId>> {
let fst = match self.main.get::<_, Str, ByteSlice>(rtxn, "words-fst")? {
Some(bytes) => fst::Set::new(bytes)?,
None => return Ok(Vec::new()),
};
// Building these factories is not free.
let lev0 = LevenshteinAutomatonBuilder::new(0, true);
let lev1 = LevenshteinAutomatonBuilder::new(1, true);
let lev2 = LevenshteinAutomatonBuilder::new(2, true);
let words: Vec<_> = alphanumeric_tokens(query).collect();
let number_of_words = words.len();
let dfas = words.into_iter().enumerate().map(|(i, word)| {
let word = word.cow_to_lowercase();
let is_last = i + 1 == number_of_words;
let dfa = match word.len() {
0..=4 => if is_last { lev0.build_prefix_dfa(&word) } else { lev0.build_dfa(&word) },
5..=8 => if is_last { lev1.build_prefix_dfa(&word) } else { lev1.build_dfa(&word) },
_ => if is_last { lev2.build_prefix_dfa(&word) } else { lev2.build_dfa(&word) },
};
(word, dfa)
});
let mut intersect_result: Option<RoaringBitmap> = None;
for (word, dfa) in dfas {
let before = Instant::now();
let mut union_result = RoaringBitmap::default();
let mut stream = fst.search(dfa).into_stream();
while let Some(word) = stream.next() {
let word = std::str::from_utf8(word)?;
if let Some(ids) = self.postings_ids.get(rtxn, word)? {
let right = RoaringBitmap::deserialize_from(ids)?;
union_result.union_with(&right);
}
}
eprintln!("union for {:?} took {:.02?}", word, before.elapsed());
intersect_result = match intersect_result.take() {
Some(mut left) => {
let before = Instant::now();
let left_len = left.len();
left.intersect_with(&union_result);
eprintln!("intersect between {:?} and {:?} took {:.02?}",
left_len, union_result.len(), before.elapsed());
Some(left)
},
None => Some(union_result),
};
}
Ok(intersect_result.unwrap_or_default().iter().take(20).collect())
}
}