use CanonicalizedUdfPath everywhere (#28010)

this fixes a bug where actions can't always look up a Reference because the reference always holds a UdfPath that lacks a `.js` extension, while some ways of constructing the same UdfPath end up with a `.js` extension. See regression test `test_schedule_by_string`.

To do this, we should make `Reference` always hold the canonical version of each UdfPath, i.e. CanonicalizedUdfPath. An alternative would be to make a new type `CanonicalizedReference` but I don't think that's worth it. Once Reference held canonical paths, Resource needed to change as well. And then I went down the rabbit hole, ending up with the following conclusions:

- Developers can refer to `basic:insertObject` and `basic.js:insertObject` and they mean exactly the same thing.
- We already always store canonical paths in system tables like `_modules` and `_scheduled_jobs` and `_cron_job_logs` etc.
- If the developer sends a module path `basic` or `basic.js` into Convex, they expect them to be stored the same and have the same meaning, because that's almost true today (it's true if we end up storing it).
- Almost all internal code should use canonical UDF paths.
- If we want to display without the `.js` extension in error messages or virtual tables, we can use `.strip()` to remove it from the canonical representation. We should not choose to display it based on user input.
- I fixed the error message `Couldn't resolve api.engine.events.js.batchRegisteredMutations` to remove the `.js` regardless of user input.

GitOrigin-RevId: 606712e7a2426781bba9f517c9345eb578caf633

Author: ldanilek

crates/application/src/api.rs

@@ -7,7 +7,7 @@ use async_trait::async_trait;
 use bytes::Bytes;
 use common::{
     components::{
-        ComponentFunctionPath,
+        CanonicalizedComponentFunctionPath,
         ComponentId,
     },
     pause::PauseClient,
@@ -99,7 +99,7 @@ pub trait ApplicationApi: Send + Sync {
         host: &str,
         request_id: RequestId,
         identity: Identity,
-        path: ComponentFunctionPath,
+        path: CanonicalizedComponentFunctionPath,
         args: Vec<JsonValue>,
         caller: FunctionCaller,
         ts: ExecuteQueryTimestamp,
@@ -111,7 +111,7 @@ pub trait ApplicationApi: Send + Sync {
         host: &str,
         request_id: RequestId,
         identity: Identity,
-        path: ComponentFunctionPath,
+        path: CanonicalizedComponentFunctionPath,
         args: Vec<JsonValue>,
         caller: FunctionCaller,
         // Identifier used to make this mutation idempotent.
@@ -123,7 +123,7 @@ pub trait ApplicationApi: Send + Sync {
         host: &str,
         request_id: RequestId,
         identity: Identity,
-        path: ComponentFunctionPath,
+        path: CanonicalizedComponentFunctionPath,
         args: Vec<JsonValue>,
         caller: FunctionCaller,
     ) -> anyhow::Result<Result<RedactedActionReturn, RedactedActionError>>;
@@ -133,7 +133,7 @@ pub trait ApplicationApi: Send + Sync {
         host: &str,
         request_id: RequestId,
         identity: Identity,
-        path: ComponentFunctionPath,
+        path: CanonicalizedComponentFunctionPath,
         args: Vec<JsonValue>,
         caller: FunctionCaller,
     ) -> anyhow::Result<Result<FunctionReturn, FunctionError>>;
@@ -218,7 +218,7 @@ impl<RT: Runtime> ApplicationApi for Application<RT> {
         _host: &str,
         request_id: RequestId,
         identity: Identity,
-        path: ComponentFunctionPath,
+        path: CanonicalizedComponentFunctionPath,
         args: Vec<JsonValue>,
         caller: FunctionCaller,
         ts: ExecuteQueryTimestamp,
@@ -242,7 +242,7 @@ impl<RT: Runtime> ApplicationApi for Application<RT> {
         _host: &str,
         request_id: RequestId,
         identity: Identity,
-        path: ComponentFunctionPath,
+        path: CanonicalizedComponentFunctionPath,
         args: Vec<JsonValue>,
         caller: FunctionCaller,
         // Identifier used to make this mutation idempotent.
@@ -270,7 +270,7 @@ impl<RT: Runtime> ApplicationApi for Application<RT> {
         _host: &str,
         request_id: RequestId,
         identity: Identity,
-        path: ComponentFunctionPath,
+        path: CanonicalizedComponentFunctionPath,
         args: Vec<JsonValue>,
         caller: FunctionCaller,
     ) -> anyhow::Result<Result<RedactedActionReturn, RedactedActionError>> {
@@ -288,7 +288,7 @@ impl<RT: Runtime> ApplicationApi for Application<RT> {
         _host: &str,
         request_id: RequestId,
         identity: Identity,
-        path: ComponentFunctionPath,
+        path: CanonicalizedComponentFunctionPath,
         args: Vec<JsonValue>,
         caller: FunctionCaller,
     ) -> anyhow::Result<Result<FunctionReturn, FunctionError>> {

crates/application/src/application_function_runner/mod.rs

@@ -20,7 +20,6 @@ use common::{
     components::{
         CanonicalizedComponentFunctionPath,
         ComponentDefinitionPath,
-        ComponentFunctionPath,
         ComponentId,
         ComponentPath,
     },
@@ -737,7 +736,7 @@ impl<RT: Runtime> ApplicationFunctionRunner<RT> {
     pub async fn retry_mutation(
         &self,
         request_id: RequestId,
-        path: ComponentFunctionPath,
+        path: CanonicalizedComponentFunctionPath,
         arguments: Vec<JsonValue>,
         identity: Identity,
         mutation_identifier: Option<SessionRequestIdentifier>,
@@ -768,7 +767,7 @@ impl<RT: Runtime> ApplicationFunctionRunner<RT> {
     async fn _retry_mutation(
         &self,
         request_id: RequestId,
-        path: ComponentFunctionPath,
+        path: CanonicalizedComponentFunctionPath,
         arguments: Vec<JsonValue>,
         identity: Identity,
         mutation_identifier: Option<SessionRequestIdentifier>,
@@ -787,7 +786,6 @@ impl<RT: Runtime> ApplicationFunctionRunner<RT> {
                 }))
             },
         };
-        let path = path.canonicalize();
         let udf_path_string = (!path.udf_path.is_system()).then_some(path.udf_path.to_string());
 
         let mut backoff = Backoff::new(
@@ -1047,7 +1045,7 @@ impl<RT: Runtime> ApplicationFunctionRunner<RT> {
     pub async fn run_action(
         &self,
         request_id: RequestId,
-        path: ComponentFunctionPath,
+        path: CanonicalizedComponentFunctionPath,
         arguments: Vec<JsonValue>,
         identity: Identity,
         caller: FunctionCaller,
@@ -1065,12 +1063,11 @@ impl<RT: Runtime> ApplicationFunctionRunner<RT> {
             },
         };
         let context = ExecutionContext::new(request_id.clone(), &caller);
-        let canonicalized_path = path.canonicalize();
         let usage_tracking = FunctionUsageTracker::new();
         let start = self.runtime.monotonic_now();
         let completion_result = self
             .run_action_no_udf_log(
-                canonicalized_path.clone(),
+                path.clone(),
                 arguments.clone(),
                 identity.clone(),
                 caller.clone(),
@@ -1083,7 +1080,7 @@ impl<RT: Runtime> ApplicationFunctionRunner<RT> {
             Err(e) => {
                 self.function_log.log_action_system_error(
                     &e,
-                    canonicalized_path,
+                    path,
                     arguments,
                     identity.into(),
                     start,
@@ -1866,7 +1863,7 @@ impl<RT: Runtime> ApplicationFunctionRunner<RT> {
     pub async fn run_query_at_ts(
         &self,
         request_id: RequestId,
-        path: ComponentFunctionPath,
+        path: CanonicalizedComponentFunctionPath,
         args: Vec<JsonValue>,
         identity: Identity,
         ts: Timestamp,
@@ -1899,7 +1896,7 @@ impl<RT: Runtime> ApplicationFunctionRunner<RT> {
     async fn run_query_at_ts_inner(
         &self,
         request_id: RequestId,
-        path: ComponentFunctionPath,
+        path: CanonicalizedComponentFunctionPath,
         args: Vec<JsonValue>,
         identity: Identity,
         ts: Timestamp,
@@ -1920,13 +1917,12 @@ impl<RT: Runtime> ApplicationFunctionRunner<RT> {
                 });
             },
         };
-        let canonicalized = path.canonicalize();
         let usage_tracker = FunctionUsageTracker::new();
         let result = self
             .cache_manager
             .get(
                 request_id,
-                canonicalized,
+                path,
                 args,
                 identity.clone(),
                 ts,
@@ -2000,7 +1996,7 @@ impl<RT: Runtime> ActionCallbacks for ApplicationFunctionRunner<RT> {
     async fn execute_query(
         &self,
         identity: Identity,
-        path: ComponentFunctionPath,
+        path: CanonicalizedComponentFunctionPath,
         args: Vec<JsonValue>,
         context: ExecutionContext,
     ) -> anyhow::Result<FunctionResult> {
@@ -2026,7 +2022,7 @@ impl<RT: Runtime> ActionCallbacks for ApplicationFunctionRunner<RT> {
     async fn execute_mutation(
         &self,
         identity: Identity,
-        path: ComponentFunctionPath,
+        path: CanonicalizedComponentFunctionPath,
         args: Vec<JsonValue>,
         context: ExecutionContext,
     ) -> anyhow::Result<FunctionResult> {
@@ -2054,7 +2050,7 @@ impl<RT: Runtime> ActionCallbacks for ApplicationFunctionRunner<RT> {
     async fn execute_action(
         &self,
         identity: Identity,
-        path: ComponentFunctionPath,
+        path: CanonicalizedComponentFunctionPath,
         args: Vec<JsonValue>,
         context: ExecutionContext,
     ) -> anyhow::Result<FunctionResult> {
@@ -2160,7 +2156,7 @@ impl<RT: Runtime> ActionCallbacks for ApplicationFunctionRunner<RT> {
         &self,
         identity: Identity,
         scheduling_component: ComponentId,
-        scheduled_path: ComponentFunctionPath,
+        scheduled_path: CanonicalizedComponentFunctionPath,
         udf_args: Vec<JsonValue>,
         scheduled_ts: UnixTimestamp,
         context: ExecutionContext,

crates/application/src/lib.rs

@@ -43,7 +43,6 @@ use common::{
         CanonicalizedComponentFunctionPath,
         CanonicalizedComponentModulePath,
         ComponentDefinitionPath,
-        ComponentFunctionPath,
         ComponentId,
         ComponentPath,
     },
@@ -848,7 +847,7 @@ impl<RT: Runtime> Application<RT> {
     pub async fn read_only_udf(
         &self,
         request_id: RequestId,
-        path: ComponentFunctionPath,
+        path: CanonicalizedComponentFunctionPath,
         args: Vec<JsonValue>,
         identity: Identity,
         caller: FunctionCaller,
@@ -862,7 +861,7 @@ impl<RT: Runtime> Application<RT> {
     pub async fn read_only_udf_at_ts(
         &self,
         request_id: RequestId,
-        path: ComponentFunctionPath,
+        path: CanonicalizedComponentFunctionPath,
         args: Vec<JsonValue>,
         identity: Identity,
         ts: Timestamp,
@@ -934,7 +933,7 @@ impl<RT: Runtime> Application<RT> {
     pub async fn mutation_udf(
         &self,
         request_id: RequestId,
-        path: ComponentFunctionPath,
+        path: CanonicalizedComponentFunctionPath,
         args: Vec<JsonValue>,
         identity: Identity,
         // Identifier used to make this mutation idempotent.
@@ -1000,7 +999,7 @@ impl<RT: Runtime> Application<RT> {
     pub async fn action_udf(
         &self,
         request_id: RequestId,
-        name: ComponentFunctionPath,
+        name: CanonicalizedComponentFunctionPath,
         args: Vec<JsonValue>,
         identity: Identity,
         caller: FunctionCaller,
@@ -1126,7 +1125,7 @@ impl<RT: Runtime> Application<RT> {
     pub async fn any_udf(
         &self,
         request_id: RequestId,
-        path: ComponentFunctionPath,
+        path: CanonicalizedComponentFunctionPath,
         args: Vec<JsonValue>,
         identity: Identity,
         caller: FunctionCaller,
@@ -1147,7 +1146,7 @@ impl<RT: Runtime> Application<RT> {
         // rare enough to disregard it.
         let mut tx_type = self.begin(identity.clone()).await?;
 
-        let canonicalized_path = path.clone().canonicalize();
+        let canonicalized_path = path.clone();
         let Some(analyzed_function) = ModuleModel::new(&mut tx_type)
             .get_analyzed_function(&canonicalized_path)
             .await?
@@ -2025,7 +2024,7 @@ impl<RT: Runtime> Application<RT> {
             component: ComponentPath::TODO(),
             udf_path: CanonicalizedUdfPath::new(module_path, function_name),
         };
-        let arguments = parse_udf_args(&path.clone().into(), args)?;
+        let arguments = parse_udf_args(&path, args)?;
         let (result, log_lines) = match analyzed_function.udf_type {
             UdfType::Query => self
                 .runner

crates/application/src/tests/components.rs

@@ -14,7 +14,7 @@ use common::{
         ComponentType,
     },
     components::{
-        ComponentFunctionPath,
+        CanonicalizedComponentFunctionPath,
         ComponentId,
         ComponentPath,
         Reference,
@@ -209,7 +209,7 @@ bar.invokeAction = async (requestId, argsStr) => {
     let action_return = application
         .action_udf(
             RequestId::new(),
-            ComponentFunctionPath {
+            CanonicalizedComponentFunctionPath {
                 component: ComponentPath::test_user(),
                 udf_path: "foo:bar".parse()?,
             },

crates/application/src/tests/cron_jobs.rs

@@ -6,7 +6,7 @@ use std::{
 
 use common::{
     components::{
-        ComponentFunctionPath,
+        CanonicalizedComponentFunctionPath,
         ComponentId,
         ComponentPath,
     },
@@ -72,12 +72,12 @@ async fn create_cron_job(
         "key".to_string(),
         serde_json::Value::String("value".to_string()),
     );
-    let path = ComponentFunctionPath {
+    let path = CanonicalizedComponentFunctionPath {
         component: ComponentPath::test_user(),
         udf_path: "basic:insertObject".parse()?,
     };
     let cron_spec = CronSpec {
-        udf_path: path.udf_path.clone().canonicalize(),
+        udf_path: path.udf_path.clone(),
         udf_args: parse_udf_args(&path, vec![JsonValue::Object(map)])?,
         cron_schedule: CronSchedule::Interval { seconds: 60 },
     };

crates/application/src/tests/mutation.rs

@@ -1,7 +1,7 @@
 use anyhow::Context;
 use common::{
     components::{
-        ComponentFunctionPath,
+        CanonicalizedComponentFunctionPath,
         ComponentPath,
     },
     knobs::UDF_EXECUTOR_OCC_MAX_RETRIES,
@@ -33,7 +33,7 @@ async fn insert_object(
     let result = application
         .mutation_udf(
             RequestId::new(),
-            ComponentFunctionPath {
+            CanonicalizedComponentFunctionPath {
                 component: ComponentPath::test_user(),
                 udf_path: "basic:insertObject".parse()?,
             },
@@ -57,7 +57,7 @@ async fn insert_and_count(
     let result = application
         .mutation_udf(
             RequestId::new(),
-            ComponentFunctionPath {
+            CanonicalizedComponentFunctionPath {
                 component: ComponentPath::test_user(),
                 udf_path: "basic:insertAndCount".parse()?,
             },

crates/application/src/tests/returns_validation.rs

@@ -1,6 +1,6 @@
 use common::{
     components::{
-        ComponentFunctionPath,
+        CanonicalizedComponentFunctionPath,
         ComponentPath,
     },
     pause::PauseClient,
@@ -33,7 +33,7 @@ async fn run_zero_arg_mutation(
     application
         .mutation_udf(
             RequestId::new(),
-            ComponentFunctionPath {
+            CanonicalizedComponentFunctionPath {
                 component: ComponentPath::test_user(),
                 udf_path: name.parse()?,
             },
@@ -54,7 +54,7 @@ async fn run_zero_arg_query(
     application
         .read_only_udf(
             RequestId::new(),
-            ComponentFunctionPath {
+            CanonicalizedComponentFunctionPath {
                 component: ComponentPath::test_user(),
                 udf_path: name.parse()?,
             },
@@ -73,7 +73,7 @@ async fn run_zero_arg_action(
     application
         .action_udf(
             RequestId::new(),
-            ComponentFunctionPath {
+            CanonicalizedComponentFunctionPath {
                 component: ComponentPath::test_user(),
                 udf_path: name.parse()?,
             },