Return None instead of raising exceptions for missing items.

Change some methods to return None rather than raising exceptions
when an item is missing.

Author: sebpuetz

src/embeddings.rs

@@ -111,7 +111,7 @@ impl PyEmbeddings {
     ///
     /// If the word is not known, its representation is approximated
     /// using subword units.
-    fn embedding(&self, word: &str) -> PyResult<Py<PyArray1<f32>>> {
+    fn embedding(&self, word: &str) -> Option<Py<PyArray1<f32>>> {
         let embeddings = self.embeddings.borrow();
 
         use EmbeddingsWrap::*;
@@ -120,16 +120,13 @@ impl PyEmbeddings {
             NonView(e) => e.embedding(word),
         };
 
-        match embedding {
-            Some(embedding) => {
-                let gil = pyo3::Python::acquire_gil();
-                Ok(embedding.into_owned().into_pyarray(gil.python()).to_owned())
-            }
-            None => Err(exceptions::KeyError::py_err("Unknown word and n-grams")),
-        }
+        embedding.map(|e| {
+            let gil = pyo3::Python::acquire_gil();
+            e.into_owned().into_pyarray(gil.python()).to_owned()
+        })
     }
 
-    fn embedding_with_norm(&self, word: &str) -> PyResult<Py<PyTuple>> {
+    fn embedding_with_norm(&self, word: &str) -> Option<Py<PyTuple>> {
         let embeddings = self.embeddings.borrow();
 
         use EmbeddingsWrap::*;
@@ -138,22 +135,15 @@ impl PyEmbeddings {
             NonView(e) => e.embedding_with_norm(word),
         };
 
-        match embedding_with_norm {
-            Some(embedding_with_norm) => {
-                let gil = pyo3::Python::acquire_gil();
-                let py = gil.python();
-                Ok((
-                    embedding_with_norm.embedding.into_owned().into_pyarray(py),
-                    embedding_with_norm.norm,
-                )
-                    .into_py(py))
-            }
-            None => Err(exceptions::KeyError::py_err("Unknown word and n-grams")),
-        }
+        embedding_with_norm.map(|e| {
+            let gil = pyo3::Python::acquire_gil();
+            let embedding = e.embedding.into_owned().into_pyarray(gil.python());
+            (embedding, e.norm).into_py(gil.python())
+        })
     }
 
     /// Copy the entire embeddings matrix.
-    fn matrix_copy(&self) -> PyResult<Py<PyArray2<f32>>> {
+    fn matrix_copy(&self) -> Py<PyArray2<f32>> {
         let embeddings = self.embeddings.borrow();
 
         use EmbeddingsWrap::*;
@@ -175,7 +165,7 @@ impl PyEmbeddings {
             },
         };
         let gil = pyo3::Python::acquire_gil();
-        Ok(matrix.into_pyarray(gil.python()).to_owned())
+        matrix.into_pyarray(gil.python()).to_owned()
     }
 
     /// Embeddings metadata.

src/vocab.rs

@@ -23,20 +23,16 @@ impl PyVocab {
 
 #[pymethods]
 impl PyVocab {
-    fn item_to_indices(&self, key: String) -> PyResult<PyObject> {
+    fn item_to_indices(&self, key: String) -> Option<PyObject> {
         let embeds = self.embeddings.borrow();
 
-        embeds
-            .vocab()
-            .idx(key.as_str())
-            .map(|idx| {
-                let gil = pyo3::Python::acquire_gil();
-                match idx {
-                    WordIndex::Word(idx) => [idx].to_object(gil.python()),
-                    WordIndex::Subword(indices) => indices.to_object(gil.python()),
-                }
-            })
-            .ok_or_else(|| exceptions::KeyError::py_err("Unknown word or n-grams"))
+        embeds.vocab().idx(key.as_str()).map(|idx| {
+            let gil = pyo3::Python::acquire_gil();
+            match idx {
+                WordIndex::Word(idx) => [idx].to_object(gil.python()),
+                WordIndex::Subword(indices) => indices.to_object(gil.python()),
+            }
+        })
     }
 }
 

tests/test_embedding.py

@@ -31,6 +31,12 @@ def test_embeddings_with_norms():
             embedding_with_norm) == 3, "The number of values returned by 'iter_with_norm()' does not match!"
 
 
+def test_embeddings_with_norms_oov():
+    embeds = finalfusion.Embeddings(
+        "tests/embeddings.fifu")
+    assert embeds.embedding_with_norm("Something out of vocabulary") is None
+
+
 def test_embeddings():
     embeds = finalfusion.Embeddings(
         "tests/embeddings.fifu")
@@ -49,6 +55,12 @@ def test_embeddings():
             embedding_with_norm) == 2, "The number of values returned by normal iterator does not match!"
 
 
+def test_embeddings_oov():
+    embeds = finalfusion.Embeddings(
+        "tests/embeddings.fifu")
+    assert embeds.embedding("Something out of vocabulary") is None
+
+
 def test_norms():
     embeds = finalfusion.Embeddings(
         "tests/embeddings.fifu")

tests/test_vocab.py

@@ -0,0 +1,9 @@
+import finalfusion
+import pytest
+
+
+def test_embeddings_with_norms_oov():
+    embeds = finalfusion.Embeddings(
+        "tests/embeddings.fifu")
+    vocab = embeds.vocab()
+    assert vocab.item_to_indices("Something out of vocabulary") is None