server: WIP tantivy, cache slurps, use shared::compute_color,

server/Cargo.toml

@@ -13,6 +13,7 @@ async-graphql = { version = "6.0.11", features = ["log"] }
 async-graphql-rocket = "6.0.11"
 async-trait = "0.1.81"
 build-info = "0.0.38"
+cacher = {git = "https://git-private.z.xinu.tv/wathiede/cacher"}
 css-inline = "0.13.0"
 glog = "0.1.0"
 html-escape = "0.2.13"

server/Rocket.toml

@@ -1,6 +1,8 @@
 [release]
 address = "0.0.0.0"
 port = 9345
+newsreader_database_url = "postgres://newsreader@nixos-07.h.xinu.tv/newsreader"
+newsreader_tantivy_db_path = "../target/database/newsreader"
 
 [debug]
 address = "0.0.0.0"
@@ -8,3 +10,5 @@ port = 9345
 # Uncomment to make it production like.
 #log_level = "critical"
 newsreader_database_url = "postgres://newsreader@nixos-07.h.xinu.tv/newsreader"
+newsreader_tantivy_db_path = "../target/database/newsreader"
+slurp_cache_path = "/net/nasx/x/letterbox/slurp"

server/sql/all-posts.sql

@@ -0,0 +1,10 @@
+SELECT
+    site,
+    title,
+    summary,
+    link,
+    date,
+    is_read,
+    uid,
+    id
+FROM post

server/src/bin/server.rs

@@ -18,18 +18,14 @@ use rocket::{
     Response, State,
 };
 use rocket_cors::{AllowedHeaders, AllowedOrigins};
-use serde::Deserialize;
 use server::{
+    config::Config,
     error::ServerError,
     graphql::{Attachment, GraphqlSchema, Mutation, QueryRoot},
     nm::{attachment_bytes, cid_attachment_bytes},
 };
 use sqlx::postgres::PgPool;
-
-#[derive(Deserialize)]
-struct Config {
-    newsreader_database_url: String,
-}
+use tantivy::{Index, IndexWriter};
 
 #[get("/refresh")]
 async fn refresh(nm: &State<Notmuch>) -> Result<Json<String>, Debug<NotmuchError>> {
@@ -170,6 +166,122 @@ fn graphiql() -> content::RawHtml<String> {
     content::RawHtml(GraphiQLSource::build().endpoint("/api/graphql").finish())
 }
 
+#[rocket::post("/create-news-db")]
+fn create_news_db(config: &State<Config>) -> Result<String, Debug<ServerError>> {
+    std::fs::remove_dir_all(&config.newsreader_tantivy_db_path).map_err(ServerError::from)?;
+    std::fs::create_dir_all(&config.newsreader_tantivy_db_path).map_err(ServerError::from)?;
+    use tantivy::schema::*;
+    let mut schema_builder = Schema::builder();
+    schema_builder.add_text_field("site", STRING | STORED);
+    schema_builder.add_text_field("title", TEXT | STORED);
+    schema_builder.add_text_field("summary", TEXT);
+    schema_builder.add_text_field("link", STRING | STORED);
+    schema_builder.add_date_field("date", FAST);
+    schema_builder.add_bool_field("is_read", FAST);
+    schema_builder.add_text_field("uid", STRING | STORED);
+    schema_builder.add_i64_field("id", FAST);
+
+    let schema = schema_builder.build();
+    Index::create_in_dir(&config.newsreader_tantivy_db_path, schema).map_err(ServerError::from)?;
+    Ok(format!(
+        "DB created in {}\n",
+        config.newsreader_tantivy_db_path
+    ))
+}
+
+#[rocket::post("/reindex-news-db")]
+async fn reindex_news_db(
+    pool: &State<PgPool>,
+    config: &State<Config>,
+) -> Result<String, Debug<ServerError>> {
+    use tantivy::{doc, Term};
+
+    let start_time = std::time::Instant::now();
+    let pool: &PgPool = pool;
+
+    let index =
+        Index::open_in_dir(&config.newsreader_tantivy_db_path).map_err(ServerError::from)?;
+    let mut index_writer = index.writer(50_000_000).map_err(ServerError::from)?;
+    let schema = index.schema();
+    let site = schema.get_field("site").map_err(ServerError::from)?;
+    let title = schema.get_field("title").map_err(ServerError::from)?;
+    let summary = schema.get_field("summary").map_err(ServerError::from)?;
+    let link = schema.get_field("link").map_err(ServerError::from)?;
+    let date = schema.get_field("date").map_err(ServerError::from)?;
+    let is_read = schema.get_field("is_read").map_err(ServerError::from)?;
+    let uid = schema.get_field("uid").map_err(ServerError::from)?;
+    let id = schema.get_field("id").map_err(ServerError::from)?;
+
+    let rows = sqlx::query_file!("sql/all-posts.sql")
+        .fetch_all(pool)
+        .await
+        .map_err(ServerError::from)?;
+
+    let total = rows.len();
+    for (i, r) in rows.into_iter().enumerate() {
+        if i % 10_000 == 0 {
+            info!(
+                "{i}/{total} processed, elapsed {:.2}s",
+                start_time.elapsed().as_secs_f32()
+            );
+        }
+        let id_term = Term::from_field_text(uid, &r.uid);
+        index_writer.delete_term(id_term);
+        index_writer
+            .add_document(doc!(
+                site =>  r.site.expect("UNKOWN_SITE"),
+                title =>  r.title.expect("UNKOWN_TITLE"),
+                // TODO: clean and extract text from HTML
+                summary => r.summary.expect("UNKNOWN_SUMMARY"),
+                link => r.link.expect("link"),
+                date => tantivy::DateTime::from_primitive(r.date.expect("date")),
+                is_read => r.is_read.expect("is_read"),
+                uid => r.uid,
+                id => r.id as i64,
+            ))
+            .map_err(ServerError::from)?;
+    }
+
+    index_writer.commit().map_err(ServerError::from)?;
+
+    info!("took {:.2}s to reindex", start_time.elapsed().as_secs_f32());
+    Ok(format!(
+        "DB openned in {}\n",
+        config.newsreader_tantivy_db_path
+    ))
+}
+
+#[rocket::get("/search-news-db")]
+fn search_news_db(
+    index: &State<tantivy::Index>,
+    reader: &State<tantivy::IndexReader>,
+) -> Result<String, Debug<ServerError>> {
+    use tantivy::{collector::TopDocs, query::QueryParser, Document, TantivyDocument};
+
+    let searcher = reader.searcher();
+    let schema = index.schema();
+    let site = schema.get_field("site").map_err(ServerError::from)?;
+    let title = schema.get_field("title").map_err(ServerError::from)?;
+    let summary = schema.get_field("summary").map_err(ServerError::from)?;
+    let query_parser = QueryParser::for_index(&index, vec![site, title, summary]);
+
+    let query = query_parser
+        .parse_query("grapheme")
+        .map_err(ServerError::from)?;
+    let top_docs = searcher
+        .search(&query, &TopDocs::with_limit(10))
+        .map_err(ServerError::from)?;
+    let mut results = vec![];
+    info!("search found {} docs", top_docs.len());
+    for (_score, doc_address) in top_docs {
+        let retrieved_doc: TantivyDocument =
+            searcher.doc(doc_address).map_err(ServerError::from)?;
+        results.push(format!("{}", retrieved_doc.to_json(&schema)));
+    }
+
+    Ok(format!("{}", results.join(" ")))
+}
+
 #[rocket::get("/graphql?<query..>")]
 async fn graphql_query(schema: &State<GraphqlSchema>, query: GraphQLQuery) -> GraphQLResponse {
     query.execute(schema.inner()).await
@@ -183,7 +295,6 @@ async fn graphql_request(
     request.execute(schema.inner()).await
 }
 
-
 #[rocket::main]
 async fn main() -> Result<(), Box<dyn Error>> {
     glog::new()
@@ -213,6 +324,9 @@ async fn main() -> Result<(), Box<dyn Error>> {
         .mount(
             shared::urls::MOUNT_POINT,
             routes![
+                create_news_db,
+                reindex_news_db,
+                search_news_db,
                 original,
                 refresh,
                 show_pretty,
@@ -229,14 +343,26 @@ async fn main() -> Result<(), Box<dyn Error>> {
         .attach(AdHoc::config::<Config>());
 
     let config: Config = rkt.figment().extract()?;
+    if !std::fs::exists(&config.slurp_cache_path)? {
+        info!("Creating slurp cache @ '{}'", &config.slurp_cache_path);
+        std::fs::create_dir_all(&config.slurp_cache_path)?;
+    }
     let pool = PgPool::connect(&config.newsreader_database_url).await?;
+    let tantivy_newsreader_index = Index::open_in_dir(&config.newsreader_tantivy_db_path)?;
+    let tantivy_newsreader_reader = tantivy_newsreader_index.reader()?;
     let schema = Schema::build(QueryRoot, Mutation, EmptySubscription)
         .data(Notmuch::default())
+        .data(config)
         .data(pool.clone())
         .extension(async_graphql::extensions::Logger)
         .finish();
 
-    let rkt = rkt.manage(schema).manage(pool).manage(Notmuch::default());
+    let rkt = rkt
+        .manage(schema)
+        .manage(pool)
+        .manage(Notmuch::default())
+        .manage(tantivy_newsreader_index)
+        .manage(tantivy_newsreader_reader);
     //.manage(Notmuch::with_config("../notmuch/testdata/notmuch.config"))
 
     rkt.launch().await?;

server/src/config.rs

@@ -0,0 +1,7 @@
+use serde::Deserialize;
+#[derive(Deserialize)]
+pub struct Config {
+    pub newsreader_database_url: String,
+    pub newsreader_tantivy_db_path: String,
+    pub slurp_cache_path: String,
+}

server/src/error.rs

@@ -1,6 +1,8 @@
 use std::{convert::Infallible, str::Utf8Error, string::FromUtf8Error};
 
 use mailparse::MailParseError;
+use tantivy::TantivyError;
+use tantivy::query::QueryParserError;
 use thiserror::Error;
 
 use crate::TransformError;
@@ -29,6 +31,10 @@ pub enum ServerError {
     StringError(String),
     #[error("invalid url: {0}")]
     UrlParseError(#[from] url::ParseError),
+    #[error("tantivy error: {0}")]
+    TantivyError(#[from] TantivyError),
+    #[error("tantivy query parse error: {0}")]
+    QueryParseError(#[from] QueryParserError),
     #[error("impossible: {0}")]
     InfaillibleError(#[from] Infallible),
 }

server/src/graphql.rs

@@ -8,7 +8,7 @@ use notmuch::Notmuch;
 use serde::{Deserialize, Serialize};
 use sqlx::postgres::PgPool;
 
-use crate::{newsreader, nm, Query};
+use crate::{config::Config, newsreader, nm, Query};
 
 /// # Number of seconds since the Epoch
 pub type UnixTime = isize;
@@ -384,6 +384,7 @@ impl QueryRoot {
     async fn thread<'ctx>(&self, ctx: &Context<'ctx>, thread_id: String) -> Result<Thread, Error> {
         let nm = ctx.data_unchecked::<Notmuch>();
         let pool = ctx.data_unchecked::<PgPool>();
+        let config = ctx.data_unchecked::<Config>();
         let debug_content_tree = ctx
             .look_ahead()
             .field("messages")
@@ -392,7 +393,7 @@ impl QueryRoot {
             .exists();
         // TODO: look at thread_id and conditionally load newsreader
         if newsreader::is_newsreader_thread(&thread_id) {
-            Ok(newsreader::thread(pool, thread_id).await?)
+            Ok(newsreader::thread(config, pool, thread_id).await?)
         } else {
             Ok(nm::thread(nm, thread_id, debug_content_tree).await?)
         }

server/src/lib.rs

@@ -1,14 +1,16 @@
+pub mod config;
 pub mod error;
 pub mod graphql;
 pub mod newsreader;
 pub mod nm;
 
-use std::{collections::HashMap, convert::Infallible, str::FromStr};
+use std::{collections::HashMap, convert::Infallible, str::FromStr, sync::Arc};
 
 use async_trait::async_trait;
+use cacher::{Cacher, FilesystemCacher};
 use css_inline::{CSSInliner, InlineError, InlineOptions};
 use linkify::{LinkFinder, LinkKind};
-use log::{error, warn};
+use log::{error, info, warn};
 use lol_html::{
     element, errors::RewritingError, html_content::ContentType, rewrite_str, text,
     RewriteStrSettings,
@@ -16,6 +18,7 @@ use lol_html::{
 use maplit::{hashmap, hashset};
 use scraper::{Html, Selector};
 use thiserror::Error;
+use tokio::sync::Mutex;
 use url::Url;
 
 use crate::newsreader::{extract_thread_id, is_newsreader_thread};
@@ -109,16 +112,17 @@ impl Transformer for InlineStyle {
             include_str!("custom.css"),
         );
         let inline_opts = InlineOptions {
-            inline_style_tags: false,
+            inline_style_tags: true,
             keep_style_tags: false,
-            keep_link_tags: false,
+            keep_link_tags: true,
             base_url: None,
-            load_remote_stylesheets: false,
+            load_remote_stylesheets: true,
             extra_css: Some(css.into()),
             preallocate_node_capacity: 32,
             ..InlineOptions::default()
         };
 
+        //info!("HTML:\n{html}");
         Ok(match CSSInliner::new(inline_opts).inline(&html) {
             Ok(inlined_html) => inlined_html,
             Err(err) => {
@@ -212,6 +216,7 @@ impl Transformer for AddOutlink {
 }
 
 struct SlurpContents {
+    cacher: Arc<Mutex<FilesystemCacher>>,
     site_selectors: HashMap<String, Vec<Selector>>,
 }
 
@@ -241,19 +246,26 @@ impl Transformer for SlurpContents {
         let Some(selectors) = self.get_selectors(&link) else {
             return Ok(html.to_string());
         };
+        let mut cacher = self.cacher.lock().await;
+        let body = if let Some(body) = cacher.get(link.as_str()) {
+            info!("cache hit for {link}");
+            String::from_utf8_lossy(&body).to_string()
+        } else {
             let body = reqwest::get(link.as_str()).await?.text().await?;
+            cacher.set(link.as_str(), body.as_bytes());
+            body
+        };
         let doc = Html::parse_document(&body);
 
         let mut results = Vec::new();
         for selector in selectors {
-            if let Some(frag) = doc.select(&selector).next() {
+            for frag in doc.select(&selector) {
                 results.push(frag.html())
-            } else {
-                warn!("couldn't find '{:?}' in {}", selector, link);
-                return Ok(html.to_string());
+                // TODO: figure out how to warn if there were no hits
+                //warn!("couldn't find '{:?}' in {}", selector, link);
             }
         }
-        Ok(results.join("<br><br>"))
+        Ok(results.join(""))
     }
 }
 
@@ -292,7 +304,7 @@ pub fn sanitize_html(
 ) -> Result<String, TransformError> {
     let inline_opts = InlineOptions {
         inline_style_tags: true,
-        keep_style_tags: false,
+        keep_style_tags: true,
         keep_link_tags: false,
         base_url: None,
         load_remote_stylesheets: false,
@@ -335,6 +347,30 @@ pub fn sanitize_html(
 
             el.set_attribute("src", &src)?;
 
+            Ok(())
+        }),
+        // Add https to href with //<domain name>
+        element!("link[href]", |el| {
+            info!("found link[href] {el:?}");
+            let mut href = el.get_attribute("href").expect("href was required");
+            if href.starts_with("//") {
+                warn!("adding https to {href}");
+                href.insert_str(0, "https:");
+            }
+
+            el.set_attribute("href", &href)?;
+
+            Ok(())
+        }),
+        // Add https to src with //<domain name>
+        element!("style[src]", |el| {
+            let mut src = el.get_attribute("src").expect("src was required");
+            if src.starts_with("//") {
+                src.insert_str(0, "https:");
+            }
+
+            el.set_attribute("src", &src)?;
+
             Ok(())
         }),
     ];

server/src/newsreader.rs

@@ -1,24 +1,26 @@
-use std::hash::{DefaultHasher, Hash, Hasher};
+use std::sync::Arc;
 
+use cacher::FilesystemCacher;
 use log::info;
 use maplit::hashmap;
 use scraper::Selector;
+use shared::compute_color;
 use sqlx::postgres::PgPool;
+use tokio::sync::Mutex;
 use url::Url;
 
-use crate::Query;
-
-const TAG_PREFIX: &'static str = "News/";
-const THREAD_PREFIX: &'static str = "news:";
-
 use crate::{
     compute_offset_limit,
+    config::Config,
     error::ServerError,
     graphql::{NewsPost, Tag, Thread, ThreadSummary},
-    AddOutlink, EscapeHtml, FrameImages, InlineStyle, SanitizeHtml, SlurpContents, StripHtml,
-    Transformer,
+    AddOutlink, EscapeHtml, FrameImages, InlineStyle, Query, SanitizeHtml, SlurpContents,
+    StripHtml, Transformer,
 };
 
+const TAG_PREFIX: &'static str = "News/";
+const THREAD_PREFIX: &'static str = "news:";
+
 pub fn is_newsreader_search(query: &str) -> bool {
     query.contains(TAG_PREFIX)
 }
@@ -128,11 +130,9 @@ pub async fn tags(pool: &PgPool, _needs_unread: bool) -> Result<Vec<Tag>, Server
     let tags = tags
         .into_iter()
         .map(|tag| {
-            let mut hasher = DefaultHasher::new();
-            tag.site.hash(&mut hasher);
-            let hex = format!("#{:06x}", hasher.finish() % (1 << 24));
             let unread = tag.unread.unwrap_or(0).try_into().unwrap_or(0);
             let name = format!("{TAG_PREFIX}{}", tag.site.expect("tag must have site"));
+            let hex = compute_color(&name);
             Tag {
                 name,
                 fg_color: "white".to_string(),
@@ -144,7 +144,11 @@ pub async fn tags(pool: &PgPool, _needs_unread: bool) -> Result<Vec<Tag>, Server
     Ok(tags)
 }
 
-pub async fn thread(pool: &PgPool, thread_id: String) -> Result<Thread, ServerError> {
+pub async fn thread(
+    config: &Config,
+    pool: &PgPool,
+    thread_id: String,
+) -> Result<Thread, ServerError> {
     let id = thread_id
         .strip_prefix(THREAD_PREFIX)
         .expect("news thread doesn't start with '{THREAD_PREFIX}'")
@@ -173,8 +177,10 @@ pub async fn thread(pool: &PgPool, thread_id: String) -> Result<Thread, ServerEr
     // TODO: add site specific cleanups. For example:
     // * Grafana does <div class="image-wrapp"><img class="lazyload>"<img src="/media/...>"</img></div>
     // * Some sites appear to be HTML encoded, unencode them, i.e. imperialviolent
+    let cacher = Arc::new(Mutex::new(FilesystemCacher::new(&config.slurp_cache_path)?));
     let body_tranformers: Vec<Box<dyn Transformer>> = vec![
         Box::new(SlurpContents {
+            cacher,
             site_selectors: hashmap![
                 "atmeta.com".to_string() => vec![
                     Selector::parse("div.entry-content").unwrap(),