plume-model: refactor Searcher to have its own DbPool #809

Closed
igalic wants to merge 6 commits from igalic/Plume:refactor/extract-searcher-with-dbpool into main
Owner

This PR is another attempt at fixing #799

Instead of relying on a separate actor model, as described, we make use of the fact that DbPool is an Arc, and can be arbitrarily cloned, so we add it to Searcher. Further more, and db_pool.get() acts like sending a message to an Actor: if there are enough connections, the Pool gives us one, if not, tough luck!

this way, we don't need to pass along a conn into the searcher functions.
This should make splitting PlumeRocket up into its components a little easier.

This PR is another attempt at fixing #799 Instead of relying on a separate actor model, as described, we make use of the fact that DbPool is an `Arc`, and can be arbitrarily cloned, so we add it to `Searcher`. Further more, and `db_pool.get()` *acts* like sending a message to an Actor: if there are enough connections, the Pool gives us one, if not, tough luck! this way, we don't need to pass along a conn into the searcher functions. This should make splitting PlumeRocket up into its components a little easier.
kiwii approved these changes 2020-07-23 18:35:10 +00:00
kiwii left a comment
Owner

👌

👌
@ -175,15 +186,16 @@ impl Searcher {
let lang = schema.get_field("lang").unwrap();
let license = schema.get_field("license").unwrap();
let conn = self.dbpool.get().unwrap();
Owner

I would prefer using ? instead of unwrap here.

I would prefer using `?` instead of `unwrap` here.
Author
Owner

can you briefly explain why?

we use .unwrap() in all other lines.

can you briefly explain why? we use `.unwrap()` in all other lines.
Owner

I think all the previous ones are safe: we now it can't panic because we control the search index schema, we know that these fields exist (even if Rust can't check this at comoile-time). Here, we may fail to get a DB connection from the pool for whatever reason, and I think we should make the error "go up", so that the web UI or the CLI can display a proper error.

I think all the previous ones are safe: we now it can't panic because we control the search index schema, we know that these fields exist (even if Rust can't check this at comoile-time). Here, we may fail to get a DB connection from the pool for whatever reason, and I think we should make the error "go up", so that the web UI or the CLI can display a proper error.
@ -222,24 +229,27 @@ impl Searcher {
let searcher = self.reader.searcher();
let res = searcher.search(&query.into_query(), &collector).unwrap();
let conn = self.dbpool.get().unwrap();
Owner

Same here.

Same here.
igalic reviewed 2020-07-24 14:49:47 +00:00
igalic left a comment
Author
Owner

👀

👀
@ -40,6 +40,7 @@ pub enum Error {
Io(std::io::Error),
MissingApProperty,
NotFound,
DbPool,
Author
Owner

you're not trying to convert the original diesel::r2d2::Error?

you're not trying to convert the original `diesel::r2d2::Error`?
Owner

It is a r2d2::Error, not the same as Diesel's one apparently. And I don't want to add another direct dependency in our Cargo.toml.

It is a `r2d2::Error`, not the same as Diesel's one apparently. And I don't want to add another direct dependency in our Cargo.toml.
Author
Owner

gonna have to ask somebody how the heck this is happening

gonna have to ask somebody how the heck this is happening
@ -177,1 +188,4 @@
let conn = match self.dbpool.get() {
Ok(c) => c,
Err(_) => return Err(Error::DbPool),
Author
Owner

that means we won't know exactly why the pool didn't give us a connection

that means we won't know exactly why the pool didn't give us a connection
igalic reviewed 2020-07-24 15:46:38 +00:00
igalic left a comment
Author
Owner

🙋🏻‍♀️

🙋🏻‍♀️
@ -31,0 +55,4 @@
///
/// The panic messages are normally explicit enough for a human to
/// understand how to fix the issue when they see it.
pub fn new(db_pool: DbPool) -> Arc<Self> {
Author
Owner

why are we returning Arc<Self>?
this means our code needs to know exactly how it's going to be used

what about cli? does that use an Arc?

i would use this function to only return a Self
there's no need to init_pool, we just want a copy

we can create an initialize method, which tries to call open and create methods in the appropriate order with the error recovery, if that makes sense here

but as it stands, for a new function, this does too much

why are we returning `Arc<Self>`? this means our code needs to know exactly how it's going to be used what about cli? does that use an Arc? i would use this function to only return a Self there's no need to init_pool, we just want a copy we can create an initialize method, which tries to call open and create *methods* in the appropriate order with the error recovery, if that makes sense here but as it stands, for a new function, this does too much
igalic force-pushed refactor/extract-searcher-with-dbpool from 9309ed279f to a5c81498b2 2020-07-24 20:12:17 +00:00 Compare
igalic added 1 commit 2020-07-25 08:05:45 +00:00
this will not compile, so now it's time to fix the rest
Author
Owner

closing in favour of #813

closing in favour of #813
igalic closed this pull request 2020-07-26 19:37:26 +00:00

Pull request closed

Sign in to join this conversation.
No reviewers
No milestone
No project
No assignees
2 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference: Plume/Plume#809
No description provided.