Allow tables to be created without proving a URL

Author: swallez

crates/store/re_protos/proto/rerun/v1alpha1/cloud.proto

@@ -863,9 +863,9 @@ message CreateTableEntryRequest {
 
   // Information about the table to register.
   //
-  // This must be encoded message of one one of the following supported types:
+  // This must be encoded message of one of the following supported types:
   // - LanceTable
-  google.protobuf.Any provider_details = 2;
+  optional google.protobuf.Any provider_details = 2;
 
   // Schema of the table to create
   rerun.common.v1alpha1.Schema schema = 3;

crates/store/re_protos/src/v1alpha1/rerun.cloud.v1alpha1.ext.rs

@@ -761,7 +761,7 @@ impl TryFrom<crate::cloud::v1alpha1::CreateDatasetEntryResponse> for CreateDatas
 pub struct CreateTableEntryRequest {
     pub name: String,
     pub schema: Schema,
-    pub provider_details: ProviderDetails,
+    pub provider_details: Option<ProviderDetails>,
 }
 
 impl TryFrom<CreateTableEntryRequest> for crate::cloud::v1alpha1::CreateTableEntryRequest {
@@ -770,7 +770,10 @@ impl TryFrom<CreateTableEntryRequest> for crate::cloud::v1alpha1::CreateTableEnt
         Ok(Self {
             name: value.name,
             schema: Some((&value.schema).try_into()?),
-            provider_details: Some((&value.provider_details).try_into()?),
+            provider_details: value
+                .provider_details
+                .map(|d| (&d).try_into())
+                .transpose()?,
         })
     }
 }
@@ -789,12 +792,10 @@ impl TryFrom<crate::cloud::v1alpha1::CreateTableEntryRequest> for CreateTableEnt
                     "schema"
                 ))?
                 .try_into()?,
-            provider_details: ProviderDetails::try_from(&value.provider_details.ok_or(
-                missing_field!(
-                    crate::cloud::v1alpha1::CreateTableEntryRequest,
-                    "provider_details"
-                ),
-            )?)?,
+            provider_details: value
+                .provider_details
+                .map(|v| ProviderDetails::try_from(&v))
+                .transpose()?,
         })
     }
 }

crates/store/re_protos/src/v1alpha1/rerun.cloud.v1alpha1.rs

@@ -816,12 +816,11 @@ where
     pub async fn create_table_entry(
         &mut self,
         name: &str,
-        url: &Url,
+        url: Option<Url>,
         schema: SchemaRef,
     ) -> ApiResult<TableEntry> {
-        let provider_details = ProviderDetails::LanceTable(LanceTable {
-            table_url: url.clone(),
-        });
+        let provider_details =
+            url.map(|url| ProviderDetails::LanceTable(LanceTable { table_url: url }));
         let request = CreateTableEntryRequest {
             name: name.to_owned(),
             schema: schema.as_ref().clone(),

crates/store/re_redap_client/src/connection_client.rs

@@ -144,7 +144,7 @@ pub async fn create_dataset_tests(service: impl RerunCloudService) {
     let create_table_request = CreateTableEntryRequest {
         name: table_name.to_owned(),
         schema: schema.clone(),
-        provider_details,
+        provider_details: Some(provider_details),
     }
     .try_into()
     .expect("Unable to create table request");

crates/store/re_redap_tests/src/tests/create_dataset.rs

@@ -28,7 +28,7 @@ pub async fn create_table_entry(service: impl RerunCloudService) {
     let create_table_request = CreateTableEntryRequest {
         name: table_name.to_owned(),
         schema: schema.clone(),
-        provider_details,
+        provider_details: Some(provider_details),
     }
     .try_into()
     .expect("Unable to create table request");

crates/store/re_redap_tests/src/tests/create_table.rs

@@ -275,7 +275,7 @@ async fn create_table_entry(
             CreateTableEntryRequest {
                 name: table_name.to_owned(),
                 schema: schema.clone(),
-                provider_details,
+                provider_details: Some(provider_details),
             }
             .try_into()
             .unwrap(),

crates/store/re_redap_tests/src/tests/update_entry.rs

@@ -12,6 +12,7 @@ use re_arrow_util::RecordBatchExt as _;
 use re_chunk_store::{Chunk, ChunkStore, ChunkStoreHandle};
 use re_log_encoding::ToTransport as _;
 use re_log_types::{EntityPath, EntryId, StoreId, StoreKind};
+use re_protos::cloud::v1alpha1::ext::LanceTable;
 use re_protos::cloud::v1alpha1::ext::{
     self, CreateDatasetEntryRequest, CreateDatasetEntryResponse, CreateTableEntryRequest,
     CreateTableEntryResponse, DataSource, DatasetDetails, EntryDetailsUpdate, ProviderDetails,
@@ -38,17 +39,39 @@ use tonic::{Code, Request, Response, Status};
 use crate::chunk_index::DatasetChunkIndexes;
 use crate::entrypoint::NamedPath;
 use crate::store::{ChunkKey, Dataset, InMemoryStore, Table};
+use re_tuid::Tuid;
 
-#[derive(Debug, Default)]
-pub struct RerunCloudHandlerSettings {}
+#[derive(Debug)]
+pub struct RerunCloudHandlerSettings {
+    storage_dir: tempfile::TempDir,
+}
+
+impl Default for RerunCloudHandlerSettings {
+    fn default() -> Self {
+        Self {
+            storage_dir: create_data_dir().expect("Failed to create data directory"),
+        }
+    }
+}
+
+fn create_data_dir() -> Result<tempfile::TempDir, crate::store::Error> {
+    Ok(tempfile::Builder::new().prefix("rerun-data-").tempdir()?)
+}
 
-#[derive(Default)]
 pub struct RerunCloudHandlerBuilder {
     settings: RerunCloudHandlerSettings,
-
     store: InMemoryStore,
 }
 
+impl Default for RerunCloudHandlerBuilder {
+    fn default() -> Self {
+        Self {
+            settings: Default::default(),
+            store: InMemoryStore::default(),
+        }
+    }
+}
+
 impl RerunCloudHandlerBuilder {
     pub fn new() -> Self {
         Self::default()
@@ -90,7 +113,6 @@ impl RerunCloudHandlerBuilder {
 const DUMMY_TASK_ID: &str = "task_00000000DEADBEEF";
 
 pub struct RerunCloudHandler {
-    #[expect(dead_code)]
     settings: RerunCloudHandlerSettings,
 
     store: tokio::sync::RwLock<InMemoryStore>,
@@ -1478,7 +1500,28 @@ impl RerunCloudService for RerunCloudHandler {
 
         let schema = Arc::new(request.schema);
 
-        let table = match &request.provider_details {
+        let details = match request.provider_details {
+            Some(details) => details,
+            None => {
+                // Create a directory in the storage directory. We use a tuid to avoid collisions
+                // and avoid any sanitization issue with the provided table name.
+                let table_path = self
+                    .settings
+                    .storage_dir
+                    .path()
+                    .join(format!("lance-{}", Tuid::new().to_string()));
+                ProviderDetails::LanceTable(LanceTable {
+                    table_url: url::Url::from_directory_path(table_path).map_err(|_| {
+                        Status::internal(format!(
+                            "Failed to create table directory in {:?}",
+                            self.settings.storage_dir.path()
+                        ))
+                    })?,
+                })
+            }
+        };
+
+        let table = match details {
             ProviderDetails::LanceTable(table) => {
                 store
                     .create_table_entry(table_name, &table.table_url, schema)

crates/store/re_server/src/rerun_cloud.rs

@@ -1669,7 +1669,7 @@ class CatalogClientInternal:
 
     def create_dataset(self, name: str) -> DatasetEntryInternal: ...
     def register_table(self, name: str, url: str) -> TableEntryInternal: ...
-    def create_table(self, name: str, schema: pa.Schema, url: str) -> TableEntryInternal: ...
+    def create_table(self, name: str, schema: pa.Schema, url: str | None) -> TableEntryInternal: ...
     def write_table(self, name: str, batches: pa.RecordBatchReader, insert_mode: TableInsertMode) -> None: ...
     def ctx(self) -> dfn.SessionContext: ...
 

rerun_py/rerun_bindings/rerun_bindings.pyi

@@ -289,7 +289,7 @@ def register_table(self, name: str, url: str) -> TableEntry:
 
         return TableEntry(self._internal.register_table(name, url))
 
-    def create_table(self, name: str, schema: pa.Schema, url: str) -> TableEntry:
+    def create_table(self, name: str, schema: pa.Schema, url: str | None) -> TableEntry:
         """
         Create and register a new table.