| title | date | tags | categories | |||
|---|---|---|---|---|---|---|
使用Rust、Actix-web和MongoDB构建简单博客网站-04 |
2021-08-22 |
|
rust |
距离上次更新有一年多了，本次文章计划是介绍使用Autowired库以及重构下项目结构。
本文完整源码见 GitHub Repo: https://github.com/nintha/demo-myblog
autowired GitHub地址：https://github.com/nintha/autowired-rs
web开发中有些是需要单例使用的组件，比如 mongodb 的 client。之前我们是直接用 once_cell 进行了一次简单的包装，但是如果这样的组件数量多了的话，维护起来也是一件头疼的事情。所以，我们将使用 autowired 库进行一些封装处理。
mongodb 相关的依赖先升级到新版本：
bson = "1.2"
mongodb = "1.2"加入autowired 依赖：
autowired = "0.1.8"
inventory = "0.1"这里引入的inventory 库是一个用于在main函数前执行一些逻辑的工具，autowired 库的实现依赖它。
// middleware/mongodb.rs
pub const DB_NAME: &str = "myblog";
#[bean(option)]
fn build_mongodb_client() -> Option<Client> {
let config = myblog_config();
let client = Runtime::new()
.unwrap()
.block_on(Client::with_uri_str(&config.mongodb_uri));
log::info!("build mongodb client, uri={}", config.mongodb_uri);
client.ok()
}这里我们对一个返回值是Option<Client>的函数标注了属性#[bean(option)]，这样返回的对象会被 autowired 自动管理。
// main.rs
#[actix_web::main]
async fn main() -> anyhow::Result<()> {
autowired::setup_submitted_beans();
// 其他逻辑代码
}我们需要在程序开头的时候执行 autowired::setup_submitted_beans() ，用于初始化所有的组件。
// article/service.rs
#[derive(Default, autowired::Component)]
pub struct ArticleService {
mongodb: Autowired<mongodb::Client>,
}
impl MongodbCrudService<Article> for ArticleService {
fn table(&self) -> Collection {
self.mongodb
.database(DB_NAME)
.collection(Article::TABLE_NAME)
}
}使用之前注册的单例组件也是比较简单的。这个例子是定义一个获取Collection实例的方法，由于Autowired<Client> 可以调用 Deref::deref获取到&Client，后续的使用就很自然了。可以看到，使用autowired 有一个小小的的限制，由于内部会把实例用Arc<T>进行一次封装，导致组件功能都需要支持通过&self 进行调用。
为了方便后续的开发和功能的扩展，我们把MongoDB的基本CRUD操作封装在一个Trait里面
// common/service.rs
use crate::common::{struct_into_document, CursorIntoVec};
use async_trait::async_trait;
use bson::oid::ObjectId;
use bson::Document;
use mongodb::Collection;
use serde::de::DeserializeOwned;
use serde::Serialize;
/// type `T` is the record data type
///
/// Eg: it's the type `Article` for the `article` collection
#[async_trait(?Send)]
pub trait MongodbCrudService<T>
where
T: 'static + DeserializeOwned + Serialize,
{
fn table(&self) -> Collection;
async fn list_with_filter(&self, filter: Document) -> mongodb::error::Result<Vec<T>> {
let cursor = self.table().find(Some(filter), None).await?;
Ok(cursor.into_vec::<T>().await)
}
/// return inserted id
async fn save(&self, record: &T) -> anyhow::Result<String> {
let d: Document = struct_into_document(record).ok_or_else(|| {
anyhow!("[MongodbCrudService::save] Failed to convert struct into document")
})?;
let rs = self.table().insert_one(d, None).await?;
let inserted_id: String = rs
.inserted_id
.as_object_id()
.map(ObjectId::to_hex)
.ok_or_else(|| anyhow!("[MongodbCrudService::save] Failed to get inserted id"))?;
Ok(inserted_id)
}
/// return modified count
async fn update_by_oid(&self, oid: ObjectId, record: &T) -> anyhow::Result<i64> {
let filter = doc! {"_id": oid};
let d: Document = struct_into_document(record).ok_or_else(|| {
anyhow!("[MongodbCrudService::update_by_oid] Failed to convert struct into document")
})?;
let update = doc! {"$set": d};
let result = self.table().update_one(filter, update, None).await?;
Ok(result.modified_count)
}
/// return deleted count
async fn remove_by_oid(&self, oid: ObjectId) -> anyhow::Result<i64> {
let filter = doc! {"_id": oid};
let result = self.table().delete_one(filter, None).await?;
Ok(result.deleted_count)
}
}MongodbCrudService 定义了五个方法，除了 fn table(&self) -> Collection; 没有默认的实现，剩下的CRUD4个方法都有默认的实现。使用的时候只需要声明具体的集合就可以了。
// article/handler.rs
use actix_web::{web, HttpRequest};
use autowired::Autowired;
use bson::oid::ObjectId;
use bson::Document;
use crate::article::service::ArticleService;
use crate::article::ArticleQuery;
use crate::common::*;
use super::Article;
use crate::common::service::MongodbCrudService;
const ARTICLE_SERVICE: Autowired<ArticleService> = Autowired::new();
pub async fn save_article(article: web::Json<Article>) -> RespResult {
let article: Article = article.into_inner();
let id = ARTICLE_SERVICE.save(&article).await?;
log::info!("save_article, id={}", id);
Resp::ok(id).to_json_result()
}
pub async fn list_article(query: web::Json<ArticleQuery>) -> RespResult {
let query = query.into_inner();
// 构造查询参数
let mut filter: Document = doc! {};
if query._id.is_some() {
filter.insert("_id", query._id.unwrap());
}
// 关键字模糊查询
if !query.keyword.is_empty() {
filter.insert(
"$or",
bson::Bson::Array(vec![
doc! {"title": {"$regex": & query.keyword, "$options": "i"}}.into(),
doc! {"author": {"$regex": & query.keyword, "$options": "i"}}.into(),
doc! {"content": {"$regex": & query.keyword, "$options": "i"}}.into(),
]),
);
}
let list = ARTICLE_SERVICE.list_with_filter(filter).await?;
Resp::ok(list).to_json_result()
}
pub async fn update_article(req: HttpRequest, article: web::Json<Article>) -> RespResult {
let id = req.match_info().get("id").unwrap_or("");
let oid = ObjectId::with_string(id).map_err(|e| {
log::error!("update_article, can't parse id to ObjectId, {:?}", e);
BusinessError::ValidationError("id".to_owned())
})?;
let effect = ARTICLE_SERVICE
.update_by_oid(oid, &article.into_inner())
.await?;
log::info!("update article, id={}, effect={}", id, effect);
Resp::ok(effect).to_json_result()
}
pub async fn remove_article(req: HttpRequest) -> RespResult {
let id = req.match_info().get("id").unwrap_or("");
if id.is_empty() {
return Err(BusinessError::ValidationError("id".to_owned()));
}
let oid = ObjectId::with_string(id).map_err(|e| {
log::error!("remove_article, can't parse id to ObjectId, {:?}", e);
BusinessError::ValidationError("id".to_owned())
})?;
let deleted = ARTICLE_SERVICE.remove_by_oid(oid).await?;
log::info!("delete article, id={}, effect={}", id, deleted);
Resp::ok(deleted).to_json_result()
}具体使用的时候，我们只需要关注业务逻辑，对输入输出参数进行一些逻辑校验和格式的转换，具体和MongoDB的交互就被屏蔽了，后续操作其他模型数据的时候也不需要写重复的代码。