Conversion flag

client/src/bastionlab/polars/policy.py

@@ -131,17 +131,24 @@ class Policy:
 
     safe_zone: Rule
     unsafe_handling: UnsafeAction
-    savable: bool
+    savable: bool = True
+    convertable: bool = False
 
     def serialize(self) -> str:
         if self.savable:
             savable_str = "true"
         else:
             savable_str = "false"
-        return f'{{"safe_zone":{self.safe_zone.serialize()},"unsafe_handling":{self.unsafe_handling.serialize()},"savable":{savable_str}}}'
+
+        if self.convertable:
+            convertable_str = "true"
+        else:
+            convertable_str = "false"
+
+        return f'{{"safe_zone":{self.safe_zone.serialize()},"unsafe_handling":{self.unsafe_handling.serialize()},"savable":{savable_str},"convertable":{convertable_str}}}'
 
 
-DEFAULT_POLICY = Policy(Aggregation(10), Review(), True)
+DEFAULT_POLICY = Policy(Aggregation(10), Review(), True, True)
 """
 Default BastionLab Client Policy `Policy(Aggregation(10), Review(), True)`
 """

docs/docs/tutorials/data_conversion.ipynb

@@ -151,7 +151,9 @@
     "from bastionlab.polars.policy import Policy, TrueRule, Log\n",
     "\n",
     "df = pl.read_csv(\"titanic.csv\")\n",
-    "policy = Policy(safe_zone=TrueRule(), unsafe_handling=Log(), savable=False)\n",
+    "policy = Policy(\n",
+    "    safe_zone=TrueRule(), unsafe_handling=Log(), savable=False, convertable=True\n",
+    ")\n",
     "rdf = client.polars.send_df(df.limit(100), policy=policy)\n",
     "\n",
     "rdf"
@@ -459,7 +461,7 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "Python 3",
+   "display_name": "Python 3.8.10 64-bit",
    "language": "python",
    "name": "python3"
   },
@@ -473,12 +475,12 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.8.10 (default, Jun 22 2022, 20:18:18) \n[GCC 9.4.0]"
+   "version": "3.8.10"
   },
   "orig_nbformat": 4,
   "vscode": {
    "interpreter": {
-    "hash": "916dbcbb3f70747c44a77c7bcd40155683ae19c65e1c03b4aa3499c5328201f1"
+    "hash": "e7370f93d1d0cde622a1f8e1c04877d8463912d04d973331ad4851f04de6915a"
    }
   }
  },

server/bastionlab_conversion/src/converter.rs

@@ -37,6 +37,11 @@ impl ConversionService for Converter {
     ) -> Result<Response<Reference>, Status> {
         self.sess_manager.verify_request(&request)?;
         let identifier = &request.get_ref().identifier;
+        let df_artifact = self.polars.get_df_artifact(identifier)?;
+
+        if !df_artifact.policy.check_convertable() {
+            return Err(Status::unknown("Dataframe is not convertable"));
+        }
 
         let df = self.polars.get_df_unchecked(&identifier)?;
 

server/bastionlab_polars/src/access_control.rs

@@ -8,6 +8,7 @@ pub struct Policy {
     safe_zone: Rule,
     unsafe_handling: UnsafeAction,
     savable: bool,
+    convertable: bool,
 }
 
 impl Policy {
@@ -26,6 +27,7 @@ impl Policy {
             safe_zone: Rule::AtLeastNOf(2, vec![self.safe_zone.clone(), other.safe_zone.clone()]),
             unsafe_handling: self.unsafe_handling.merge(other.unsafe_handling),
             savable: self.savable && other.savable,
+            convertable: self.convertable && other.convertable,
         }
     }
 
@@ -34,12 +36,17 @@ impl Policy {
             safe_zone: Rule::True,
             unsafe_handling: UnsafeAction::Log,
             savable: true,
+            convertable: true,
         }
     }
 
     pub fn check_savable(&self) -> bool {
         return self.savable;
     }
+
+    pub fn check_convertable(&self) -> bool {
+        return self.convertable;
+    }
 }
 
 #[derive(Debug, Clone, PartialEq, Serialize, Deserialize)]

server/bastionlab_polars/src/composite_plan.rs

@@ -119,11 +119,14 @@ impl CompositePlan {
                         "Could not apply with_row_count: no input data frame",
                     ))?;
                     let df = frame.df.with_row_count(&name, Some(0)).map_err(|e| {
-                        Status::invalid_argument(format!("Error while running with_row_count: {}", e))
+                        Status::invalid_argument(format!(
+                            "Error while running with_row_count: {}",
+                            e
+                        ))
                     })?;
                     let stats = frame.stats;
                     stack.push(StackFrame { df, stats });
-            }
+                }
             }
         }
 

server/bastionlab_polars/src/lib.rs

@@ -59,7 +59,7 @@ pub struct DelayedDataFrame {
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct DataFrameArtifact {
     dataframe: DataFrame,
-    policy: Policy,
+    pub policy: Policy,
     fetchable: VerificationResult,
     blacklist: Vec<String>,
     query_details: String,
@@ -251,6 +251,19 @@ Reason: {}",
         })
     }
 
+    pub fn get_df_artifact(&self, identifier: &str) -> Result<DataFrameArtifact, Status> {
+        let dfs = self.dataframes.read().unwrap();
+        Ok(dfs
+            .get(identifier)
+            .ok_or_else(|| {
+                Status::not_found(format!(
+                    "Could not find dataframe: identifier={}",
+                    identifier
+                ))
+            })?
+            .clone())
+    }
+
     pub fn get_df_unchecked(&self, identifier: &str) -> Result<DataFrame, Status> {
         let dfs = self.dataframes.read().unwrap();
         Ok(dfs

tests/test_conversion.py

@@ -32,7 +32,7 @@ def test_df_to_tensor_conv(self):
             }
         ).with_column((pl.col("a") * pl.col("b")).alias("c"))
 
-        rdf = client.polars.send_df(df, Policy(TrueRule(), Log(), False))
+        rdf = client.polars.send_df(df, Policy(TrueRule(), Log(), False, True))
 
         arr = rdf.to_array()
 
@@ -53,7 +53,7 @@ def test_split_remote_array_with_negs(self):
             }
         ).with_column((pl.col("a") * pl.col("b")).alias("c"))
 
-        rdf = client.polars.send_df(df, Policy(TrueRule(), Log(), False))
+        rdf = client.polars.send_df(df, Policy(TrueRule(), Log(), False, True))
         arr = rdf.to_array()
 
         with self.assertRaises(ValueError) as ve:
@@ -69,7 +69,7 @@ def test_split_remote_array(self):
             }
         ).with_column((pl.col("a") * pl.col("b")).alias("c"))
 
-        rdf = client.polars.send_df(df, Policy(TrueRule(), Log(), False))
+        rdf = client.polars.send_df(df, Policy(TrueRule(), Log(), False, True))
 
         arr = rdf.to_array()
 

tests/test_dtypes.py

@@ -43,7 +43,7 @@ def test_dtypes(self):
 
         for df in frames:
             rdf = client.polars.send_df(
-                df, policy=Policy(TrueRule(), Log(), savable=False)
+                df, policy=Policy(TrueRule(), Log(), savable=False, convertable=True)
             )
             df2 = rdf.select(pl.all()).collect().fetch()
 
@@ -67,7 +67,7 @@ def test_mixed_types_dataframe(self):
 
         for df in frames:
             rdf = client.polars.send_df(
-                df, policy=Policy(TrueRule(), Log(), savable=False)
+                df, policy=Policy(TrueRule(), Log(), savable=False, convertable=True)
             )
             df2 = rdf.select(pl.all()).collect().fetch()
 

tests/test_queries.py

@@ -29,7 +29,10 @@ def testingdf(self):
         connection = Connection("localhost", 50056)
         client = connection.client
         policy = Policy(
-            safe_zone=Aggregation(min_agg_size=1), unsafe_handling=Log(), savable=False
+            safe_zone=Aggregation(min_agg_size=1),
+            unsafe_handling=Log(),
+            savable=False,
+            convertable=True,
         )
         rdf = client.polars.send_df(df, policy)
         self.assertNotEqual(rdf, None)
@@ -40,7 +43,10 @@ def testingquery(self):
         connection = Connection("localhost", 50056)
         client = connection.client
         policy = Policy(
-            safe_zone=Aggregation(min_agg_size=1), unsafe_handling=Log(), savable=False
+            safe_zone=Aggregation(min_agg_size=1),
+            unsafe_handling=Log(),
+            savable=False,
+            convertable=True,
         )
         rdf = client.polars.send_df(df, policy)
         per_class_rates = (
@@ -57,7 +63,12 @@ def testingquery2(self):
         df = pl.read_csv("titanic.csv").limit(50)
         connection = Connection("localhost", 50056)
         client = connection.client
-        policy = Policy(safe_zone=Aggregation(1), unsafe_handling=Log(), savable=False)
+        policy = Policy(
+            safe_zone=Aggregation(1),
+            unsafe_handling=Log(),
+            savable=False,
+            convertable=True,
+        )
         rdf = client.polars.send_df(df, policy)
         per_sex_rates = (
             rdf.select([pl.col("Sex"), pl.col("Survived")])

tests/test_training.py

@@ -63,7 +63,9 @@ def test_df_to_tensor_conv(self):
         client = connection.client
         get_covid_dataset()
         df = pl.read_csv("covid.csv").limit(200)
-        policy = Policy(safe_zone=TrueRule(), unsafe_handling=Log(), savable=True)
+        policy = Policy(
+            safe_zone=TrueRule(), unsafe_handling=Log(), savable=True, convertable=True
+        )
         rdf = client.polars.send_df(df, policy=policy, sanitized_columns=["Name"])
         rdf = rdf.drop(
             [