derive_builder pattern is applied successfully.
This is mostly an exercise to understand the derive_builder pattern. It required a few tips to get it working, but in the end, it's actually what I want. I also learned a lot about how the Executor pattern, the Results<> object, error mapping, and futures interact in this code. This is going to be incredibly useful long-term, as long as I still keep this project "live" in my head.
This commit is contained in:
parent
e0c463f9fc
commit
1b8e1c067d
|
@ -14,6 +14,9 @@ readme = "./README.org"
|
|||
friendly_id = "0.3.0"
|
||||
thiserror = "1.0.20"
|
||||
derive_builder = "0.9.0"
|
||||
lazy_static = "1.4.0"
|
||||
regex = "1.3.9"
|
||||
slug = "0.1.4"
|
||||
tokio = { version = "0.2.22", features = ["rt-threaded", "blocking"] }
|
||||
serde = { version = "1.0.116", features = ["derive"] }
|
||||
serde_json = "1.0.56"
|
||||
|
|
|
@ -13,6 +13,6 @@ pub enum NoteStoreError {
|
|||
NotFound,
|
||||
|
||||
/// All other errors from the database.
|
||||
#[error(transparent)]
|
||||
#[error("Sqlx")]
|
||||
DBError(#[from] sqlx::Error),
|
||||
}
|
||||
|
|
|
@ -1,4 +1,3 @@
|
|||
use chrono::{DateTime, Utc};
|
||||
mod errors;
|
||||
mod row_structs;
|
||||
mod store;
|
||||
|
@ -9,6 +8,7 @@ pub use crate::store::NoteStore;
|
|||
|
||||
#[cfg(test)]
|
||||
mod tests {
|
||||
use chrono;
|
||||
use super::*;
|
||||
use tokio;
|
||||
|
||||
|
@ -38,12 +38,12 @@ mod tests {
|
|||
#[tokio::test(threaded_scheduler)]
|
||||
async fn fetching_unfound_page_by_title_works() {
|
||||
let title = "Nonexistent Page";
|
||||
let now = chrono::Utc::now();
|
||||
let _now = chrono::Utc::now();
|
||||
let storagepool = fresh_inmemory_database().await;
|
||||
let newpageresult = storagepool.get_page_by_title(&title).await;
|
||||
|
||||
assert!(newpageresult.is_ok(), "{:?}", newpage);
|
||||
let (newpage, newnotes) = newpageresult.unwrap();
|
||||
assert!(newpageresult.is_ok(), "{:?}", newpageresult);
|
||||
let (newpage, _newnotes) = newpageresult.unwrap();
|
||||
|
||||
assert_eq!(newpage.title, title, "{:?}", newpage.title);
|
||||
assert_eq!(newpage.slug, "nonexistent-page");
|
||||
|
|
|
@ -1,5 +1,5 @@
|
|||
use chrono::{DateTime, Utc};
|
||||
use derive_builder;
|
||||
use derive_builder::Builder;
|
||||
use serde::{Deserialize, Serialize};
|
||||
use sqlx::{self, FromRow};
|
||||
|
||||
|
@ -32,12 +32,13 @@ pub struct NewPage {
|
|||
pub slug: String,
|
||||
pub title: String,
|
||||
pub note_id: i64,
|
||||
#[builder(default = "chrono::Utc::now()")]
|
||||
#[builder(default = r#"chrono::Utc::now()"#)]
|
||||
pub creation_date: DateTime<Utc>,
|
||||
#[builder(default = "chrono::Utc::now()")]
|
||||
#[builder(default = r#"chrono::Utc::now()"#)]
|
||||
pub updated_date: DateTime<Utc>,
|
||||
#[builder(default = "chrono::Utc::now()")]
|
||||
#[builder(default = r#"chrono::Utc::now()"#)]
|
||||
pub lastview_date: DateTime<Utc>,
|
||||
#[builder(default = r#"None"#)]
|
||||
pub deleted_date: Option<DateTime<Utc>>,
|
||||
}
|
||||
|
||||
|
@ -45,28 +46,43 @@ pub struct NewPage {
|
|||
pub struct NewNote {
|
||||
pub uuid: String,
|
||||
pub content: String,
|
||||
#[builder(default = "note")]
|
||||
#[builder(default = r#""note".to_string()"#)]
|
||||
pub notetype: String,
|
||||
#[builder(default = "chrono::Utc::now()")]
|
||||
#[builder(default = r#"chrono::Utc::now()"#)]
|
||||
pub creation_date: DateTime<Utc>,
|
||||
#[builder(default = "chrono::Utc::now()")]
|
||||
#[builder(default = r#"chrono::Utc::now()"#)]
|
||||
pub updated_date: DateTime<Utc>,
|
||||
#[builder(default = "chrono::Utc::now()")]
|
||||
#[builder(default = r#"chrono::Utc::now()"#)]
|
||||
pub lastview_date: DateTime<Utc>,
|
||||
#[builder(default = r#"None"#)]
|
||||
pub deleted_date: Option<DateTime<Utc>>,
|
||||
}
|
||||
|
||||
#[derive(Clone, Serialize, Deserialize, Debug, FromRow)]
|
||||
pub(crate) struct JustSlugs {
|
||||
pub slug: String,
|
||||
}
|
||||
|
||||
#[derive(Clone, Serialize, Deserialize, Debug, FromRow)]
|
||||
pub struct JustTitles {
|
||||
title: String,
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
mod tests {
|
||||
use super::*;
|
||||
|
||||
#[test]
|
||||
fn can_build_new_page() {
|
||||
fn can_build_new_note() {
|
||||
let now = chrono::Utc::now();
|
||||
let newnote = NewNoteBuilder::default();
|
||||
assert!((newnote.creation_date - now).num_minutes() < 1.0);
|
||||
assert!((newnote.updated_date - now).num_minutes() < 1.0);
|
||||
assert!((newnote.lastview_date - now).num_minutes() < 1.0);
|
||||
let newnote = NewNoteBuilder::default()
|
||||
.uuid("foo".to_string())
|
||||
.content("bar".to_string())
|
||||
.build()
|
||||
.unwrap();
|
||||
assert!((newnote.creation_date - now).num_minutes() < 1);
|
||||
assert!((newnote.updated_date - now).num_minutes() < 1);
|
||||
assert!((newnote.lastview_date - now).num_minutes() < 1);
|
||||
assert!(newnote.deleted_date.is_none());
|
||||
}
|
||||
}
|
||||
|
|
|
@ -1,38 +1,87 @@
|
|||
SELECT parent_uuid, uuid, content, notetype, nature, position FROM (
|
||||
-- This is undoubtedly one of the more complex bits of code I've
|
||||
-- written recently, and I do wish there had been macros because
|
||||
-- there's a lot of hand-written, copy-pasted code here around the
|
||||
-- basic content of a note; it would have been nice to be able to DRY
|
||||
-- that out.
|
||||
|
||||
WITH RECURSIVE children(
|
||||
parent_id,
|
||||
parent_uuid, id,
|
||||
uuid,
|
||||
content,
|
||||
notetype,
|
||||
creation_date,
|
||||
updated_date,
|
||||
lastview_date,
|
||||
deleted_date,
|
||||
cycle
|
||||
) AS (
|
||||
-- This expression creates a table, 'notetree', that contains all of
|
||||
-- the notes nested under a page. Each entry in the table includes
|
||||
-- the note's parent's internal and external ids so that applications
|
||||
-- can build an actual tree out of a vec of these things.
|
||||
|
||||
SELECT
|
||||
notes.id,
|
||||
notes.uuid,
|
||||
notes.id,
|
||||
notes.uuid,
|
||||
notes.content,
|
||||
notes.notetype, 'page', 0, ','||notes.id||','
|
||||
FROM notes INNER JOIN pages
|
||||
ON pages.note_id = notes.id
|
||||
WHERE pages.id = ?
|
||||
AND notes.notetype="page"
|
||||
-- TODO: Extensive testing to validate that the nodes are delivered
|
||||
-- *in nesting order* to the client.
|
||||
|
||||
SELECT
|
||||
parent_id,
|
||||
parent_uuid,
|
||||
note_id,
|
||||
note_uuid,
|
||||
content,
|
||||
position,
|
||||
notetype,
|
||||
creation_date,
|
||||
updated_date,
|
||||
lastview_date,
|
||||
deleted_date
|
||||
|
||||
FROM (
|
||||
|
||||
WITH RECURSIVE notetree(
|
||||
parent_id,
|
||||
parent_uuid,
|
||||
id,
|
||||
uuid,
|
||||
content,
|
||||
position,
|
||||
notetype,
|
||||
creation_date,
|
||||
updated_date,
|
||||
lastview_date,
|
||||
deleted_date,
|
||||
cycle) AS
|
||||
|
||||
-- ROOT expression
|
||||
SELECT
|
||||
notes.id,
|
||||
notes.uuid,
|
||||
notes.id,
|
||||
notes.uuid,
|
||||
notes.content,
|
||||
0, -- Root notes are always in position 0
|
||||
notes.notetype,
|
||||
notes.creation_date,
|
||||
notes.updated_date,
|
||||
notes.lastview_date,
|
||||
notes.deleted_date,
|
||||
','||notes.id||',' -- Cycle monitor
|
||||
FROM notes
|
||||
INNER JOIN pages ON pages.note_id = notes.id
|
||||
WHERE
|
||||
pages.slug = ? AND notes.notetype = "root"
|
||||
|
||||
-- RECURSIVE expression
|
||||
UNION SELECT
|
||||
notetree.id,
|
||||
notetree.uuid,
|
||||
notes.id,
|
||||
notes.uuid,
|
||||
notes.content,
|
||||
note_relationships.position,
|
||||
notes.notetype,
|
||||
notes.creation_date,
|
||||
notes.updated_date,
|
||||
notes.lastview_date,
|
||||
notes.deleted_date,
|
||||
notetree.cycle||notes.id||','
|
||||
FROM notes
|
||||
INNER JOIN note_relationships ON notes.id = note_relationships.note_id
|
||||
-- For a given ID in the level of notetree in *this* recursion,
|
||||
-- we want each note's branches one level down.
|
||||
INNER JOIN notetree ON note_relationships.parent_id = notetree.id
|
||||
-- And we want to make sure there are no cycles. There shouldn't
|
||||
-- be; we're supposed to prevent those. But you never know.
|
||||
WHERE
|
||||
notetree.cycle NOT LIKE '%,'||notes.id||',%'
|
||||
ORDER BY note_relationships.position);
|
||||
|
||||
UNION
|
||||
SELECT note_relationships.parent_id, notes.id,
|
||||
notes.content, notes.notetype, note_relationships.nature,
|
||||
note_relationships.position,
|
||||
children.cycle||notes.id||','
|
||||
FROM notes
|
||||
INNER JOIN note_relationships ON notes.id = note_relationships.note_id
|
||||
INNER JOIN children ON note_relationships.parent_id = children.id
|
||||
WHERE children.cycle NOT LIKE '%,'||notes.id||',%'
|
||||
ORDER BY note_relationships.position)
|
||||
SELECT * from children);</code>
|
||||
|
|
|
@ -1,11 +1,12 @@
|
|||
use crate::errors::NoteStoreError;
|
||||
use crate::row_structs::{RawNote, RawPage};
|
||||
use chrono;
|
||||
use friendly_id;
|
||||
use crate::row_structs::{JustSlugs, NewNote, NewPage, RawNote, RawPage};
|
||||
use lazy_static::lazy_static;
|
||||
use regex::Regex;
|
||||
use slug::slugify;
|
||||
use sqlx;
|
||||
use sqlx::{
|
||||
sqlite::{Sqlite, SqlitePool},
|
||||
Done, Executor,
|
||||
Executor,
|
||||
};
|
||||
use std::sync::Arc;
|
||||
|
||||
|
@ -26,7 +27,7 @@ impl NoteStore {
|
|||
// to its original empty form. Do not use unless you
|
||||
// really, really want that to happen.
|
||||
pub async fn reset_database(&self) -> NoteResult<()> {
|
||||
reset_databate(&*self.0).await
|
||||
reset_database(&*self.0).await.map_err(NoteStoreError::DBError)
|
||||
}
|
||||
|
||||
/// Fetch page by slug
|
||||
|
@ -39,6 +40,7 @@ impl NoteStore {
|
|||
pub async fn get_page_by_slug(&self, slug: &str) -> NoteResult<(RawPage, Vec<RawNote>)> {
|
||||
// let select_note_collection_for_root = include_str!("sql/select_note_collection_for_root.sql");
|
||||
let mut tx = self.0.begin().await?;
|
||||
let page = select_page_by_slug(&mut tx, slug).await?;
|
||||
// let notes = sqlx::query_as(select_note_collection_for_root)
|
||||
// .bind(page.note_id)
|
||||
// .fetch(&tx)
|
||||
|
@ -49,19 +51,17 @@ impl NoteStore {
|
|||
|
||||
pub async fn get_page_by_title(&self, title: &str) -> NoteResult<(RawPage, Vec<RawNote>)> {
|
||||
let mut tx = self.0.begin().await?;
|
||||
let page = match select_page_by_title(&mut tx, title) {
|
||||
Ok(page) => page,
|
||||
Err(sqlx::Error::NotFound) => {
|
||||
match create_page_for_title(&mut tx, title) {
|
||||
Ok(page) => page,
|
||||
Err(e) => return Err(e)
|
||||
}
|
||||
},
|
||||
Err(e) => return Err(e),
|
||||
};
|
||||
// Todo: Replace vec with the results of the CTE
|
||||
return Ok((page, vec![]))
|
||||
}
|
||||
let page = match select_page_by_title(&mut tx, title).await {
|
||||
Ok(page) => page,
|
||||
Err(sqlx::Error::RowNotFound) => match create_page_for_title(&mut tx, title).await {
|
||||
Ok(page) => page,
|
||||
Err(e) => return Err(NoteStoreError::DBError(e))
|
||||
},
|
||||
Err(e) => return Err(NoteStoreError::DBError(e)),
|
||||
};
|
||||
// Todo: Replace vec with the results of the CTE
|
||||
return Ok((page, vec![]));
|
||||
}
|
||||
}
|
||||
|
||||
// ___ _ _
|
||||
|
@ -79,14 +79,14 @@ async fn select_page_by_slug<'e, E>(executor: E, slug: &str) -> SqlResult<RawPag
|
|||
where
|
||||
E: 'e + Executor<'e, Database = Sqlite>,
|
||||
{
|
||||
let select_one_page_by_title_sql = concat!(
|
||||
let select_one_page_by_slug_sql = concat!(
|
||||
"SELECT id, title, slug, note_id, creation_date, updated_date, ",
|
||||
"lastview_date, deleted_date FROM pages WHERE slug=?;"
|
||||
);
|
||||
sqlx::query_as(select_one_page_by_slug_sql)
|
||||
Ok(sqlx::query_as(&select_one_page_by_slug_sql)
|
||||
.bind(&slug)
|
||||
.fetch_one(&mut executor)
|
||||
.await?
|
||||
.fetch_one(executor)
|
||||
.await?)
|
||||
}
|
||||
|
||||
async fn select_page_by_title<'e, E>(executor: E, title: &str) -> SqlResult<RawPage>
|
||||
|
@ -97,10 +97,10 @@ where
|
|||
"SELECT id, title, slug, note_id, creation_date, updated_date, ",
|
||||
"lastview_date, deleted_date FROM pages WHERE title=?;"
|
||||
);
|
||||
sqlx::query_as(select_one_page_by_title_sql)
|
||||
Ok(sqlx::query_as(&select_one_page_by_title_sql)
|
||||
.bind(&title)
|
||||
.fetch_one(&mut executor)
|
||||
.await?
|
||||
.fetch_one(executor)
|
||||
.await?)
|
||||
}
|
||||
|
||||
async fn reset_database<'e, E>(executor: E) -> SqlResult<()>
|
||||
|
@ -108,42 +108,63 @@ where
|
|||
E: 'e + Executor<'e, Database = Sqlite>,
|
||||
{
|
||||
let initialize_sql = include_str!("sql/initialize_database.sql");
|
||||
sqlx::query(initialize_sql).execute(&*self.0).await?
|
||||
sqlx::query(initialize_sql).execute(executor).await.map(|_| ())
|
||||
}
|
||||
|
||||
async fn get_note_collection_for_root<'e, E>(executor: E, root: i64) -> SqlResult<Vec<RawNotes>>
|
||||
async fn get_note_collection_for_root<'e, E>(executor: E, root: i64) -> SqlResult<Vec<RawNote>>
|
||||
where
|
||||
E: 'e + Executor<'e, Database = Sqlite>,
|
||||
{
|
||||
let select_note_collection_for_root = include_str!("sql/select_note_collection_for_root.sql");
|
||||
sqlx::query_as(select_note_collection_for_root)
|
||||
.fetch(&*self.0)
|
||||
.await?
|
||||
let select_note_collection_for_root = include_str!("sql/select_note_collection_from_root.sql");
|
||||
Ok(sqlx::query_as(&select_note_collection_for_root)
|
||||
.bind(&root)
|
||||
.fetch_all(executor)
|
||||
.await?)
|
||||
}
|
||||
|
||||
async fn insert_one_new_note<'e, E>(executor: E, note: &NewNote) -> SqlResult<i64> where
|
||||
async fn insert_one_new_note<'e, E>(executor: E, note: &NewNote) -> SqlResult<i64>
|
||||
where
|
||||
E: 'e + Executor<'e, Database = Sqlite>,
|
||||
{
|
||||
let insert_one_note_sql = concat!(
|
||||
"INSERT INTO notes ( ",
|
||||
" uuid, ",
|
||||
" content, ",
|
||||
" notetype, ",
|
||||
" creation_date, ",
|
||||
" updated_date, ",
|
||||
" lastview_date) ",
|
||||
"VALUES (?, ?, ?, ?, ?, ?);");
|
||||
let insert_one_note_sql = concat!(
|
||||
"INSERT INTO notes ( ",
|
||||
" uuid, ",
|
||||
" content, ",
|
||||
" notetype, ",
|
||||
" creation_date, ",
|
||||
" updated_date, ",
|
||||
" lastview_date) ",
|
||||
"VALUES (?, ?, ?, ?, ?, ?);"
|
||||
);
|
||||
|
||||
Ok(sqlx::query(insert_one_note_sql)
|
||||
.bind(¬e.uuid)
|
||||
.bind(¬e.content)
|
||||
.bind(¬e.note_type)
|
||||
.bind(¬e.creation_date)
|
||||
.bind(¬e.updated_date)
|
||||
.bind(¬e.lastview_date)
|
||||
.execute(&mut tx)
|
||||
.await?
|
||||
.last_insert_rowid())
|
||||
.bind(¬e.uuid)
|
||||
.bind(¬e.content)
|
||||
.bind(¬e.notetype)
|
||||
.bind(¬e.creation_date)
|
||||
.bind(¬e.updated_date)
|
||||
.bind(¬e.lastview_date)
|
||||
.execute(executor)
|
||||
.await?
|
||||
.last_insert_rowid())
|
||||
}
|
||||
|
||||
fn find_maximal_slug(slugs: &Vec<JustSlugs>) -> Option<u32> {
|
||||
lazy_static! {
|
||||
static ref RE_CAP_NUM: Regex = Regex::new(r"-(\d+)$").unwrap();
|
||||
}
|
||||
|
||||
if slugs.len() == 0 {
|
||||
return None;
|
||||
}
|
||||
|
||||
let mut slug_counters: Vec<u32> = slugs
|
||||
.iter()
|
||||
.filter_map(|slug| RE_CAP_NUM.captures(&slug.slug))
|
||||
.map(|cap| cap.get(1).unwrap().as_str().parse::<u32>().unwrap())
|
||||
.collect();
|
||||
slug_counters.sort_unstable();
|
||||
slug_counters.pop()
|
||||
}
|
||||
|
||||
// Given an initial string and an existing collection of slugs,
|
||||
|
@ -156,28 +177,19 @@ where
|
|||
lazy_static! {
|
||||
static ref RE_STRIP_NUM: Regex = Regex::new(r"-\d+$").unwrap();
|
||||
}
|
||||
lazy_static! {
|
||||
static ref RE_CAP_NUM: Regex = Regex::new(r"-(\d+)$").unwrap();
|
||||
}
|
||||
|
||||
let initial_slug = slugify::slugify(title);
|
||||
let sample_slug = RE_STRIP_NUM.replace_all(slug, "");
|
||||
let similar_slugs: Vec<JustSlugs> = sqlx::query("SELECT slug FROM pages WHERE slug LIKE '?%';")
|
||||
.bind(&sample_slug)
|
||||
.execute(executor)
|
||||
.await?;
|
||||
let slug_counters = similar_slugs
|
||||
.iter()
|
||||
.map(|slug| RE_CAPNUM.captures(slug.slug))
|
||||
.filter_map(|cap| cap.get(1).unwrap().parse::<u32>().unwrap())
|
||||
.collect();
|
||||
match slug_counters.len() {
|
||||
0 => Ok(initial_slug),
|
||||
_ => {
|
||||
slug_counters.sort_unstable();
|
||||
return Ok(format!("{}-{}", initial_slug, slug_counters.pop() + 1))
|
||||
}
|
||||
}
|
||||
let initial_slug = slugify(title);
|
||||
let sample_slug = RE_STRIP_NUM.replace_all(&initial_slug, "");
|
||||
let slug_finder_sql = "SELECT slug FROM pages WHERE slug LIKE '?%';";
|
||||
let similar_slugs: Vec<JustSlugs> = sqlx::query_as(&slug_finder_sql)
|
||||
.bind(&*sample_slug)
|
||||
.fetch_all(executor)
|
||||
.await?;
|
||||
let maximal_slug = find_maximal_slug(&similar_slugs);
|
||||
match maximal_slug {
|
||||
None => Ok(initial_slug),
|
||||
Some(max_slug) => Ok(format!("{}-{}", initial_slug, max_slug + 1)),
|
||||
}
|
||||
}
|
||||
|
||||
async fn insert_one_new_page<'e, E>(executor: E, page: &NewPage) -> SqlResult<i64>
|
||||
|
@ -185,30 +197,31 @@ where
|
|||
E: 'e + Executor<'e, Database = Sqlite>,
|
||||
{
|
||||
let insert_one_page_sql = concat!(
|
||||
"INSERT INTO pages ( ",
|
||||
" slug, ",
|
||||
" title, ",
|
||||
" note_id, ",
|
||||
" creation_date, ",
|
||||
" updated_date, ",
|
||||
" lastview_date) ",
|
||||
"VALUES (?, ?, ?, ?, ?, ?);");
|
||||
"INSERT INTO pages ( ",
|
||||
" slug, ",
|
||||
" title, ",
|
||||
" note_id, ",
|
||||
" creation_date, ",
|
||||
" updated_date, ",
|
||||
" lastview_date) ",
|
||||
"VALUES (?, ?, ?, ?, ?, ?);"
|
||||
);
|
||||
|
||||
Ok(sqlx::query(insert_one_page_sql)
|
||||
.bind(&page.slug)
|
||||
.bind(&page.title)
|
||||
.bind(&page.note_id)
|
||||
.bind(&page.creation_date)
|
||||
.bind(&page.updated_date)
|
||||
.bind(&page.lastview_date)
|
||||
.execute(&mut tx)
|
||||
.await?
|
||||
.last_insert_rowid())
|
||||
.bind(&page.slug)
|
||||
.bind(&page.title)
|
||||
.bind(&page.note_id)
|
||||
.bind(&page.creation_date)
|
||||
.bind(&page.updated_date)
|
||||
.bind(&page.lastview_date)
|
||||
.execute(executor)
|
||||
.await?
|
||||
.last_insert_rowid())
|
||||
}
|
||||
|
||||
|
||||
async fn create_page_for_title<'e, E>(executor: E, title: &str) -> SqlResult<RawPage> where
|
||||
async fn create_page_for_title<'e, E>(_executor: E, _title: &str) -> SqlResult<RawPage>
|
||||
where
|
||||
E: 'e + Executor<'e, Database = Sqlite>,
|
||||
{
|
||||
todo!()
|
||||
todo!()
|
||||
}
|
||||
|
|
Loading…
Reference in New Issue