diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
index 5f126e6..6a52db4 100644
--- a/CONTRIBUTING.md
+++ b/CONTRIBUTING.md
@@ -34,6 +34,12 @@ Once `h5grove` is installed in editable mode, run the tests with
 invoke test
 ```
 
+Available options (cf. `invoke --help test`):
+
+- `-c`, `--cov-lines`: display line numbers in coverage report
+- `-k [keyword]`, `--keyword [keyword]`: filter tests by keyword
+- `-v`, `--verbose`: enable verbose test output
+
 ### Benchmarks
 
 Benchmarks of different formats (`json`, `.npy`) and web server frameworks (`flask`, `tornado`) are run as part of the tests. Those benchmarks are based on [pytest-benchmark](https://pytest-benchmark.readthedocs.io/en/latest/).
diff --git a/h5grove/encoders.py b/h5grove/encoders.py
index f054e99..4d20fe1 100644
--- a/h5grove/encoders.py
+++ b/h5grove/encoders.py
@@ -111,10 +111,6 @@ def encode(content: Any, encoding: Optional[str] = "json") -> Response:
         )
 
     content_array = np.array(content, copy=False)
-    if not is_numeric_data(content_array):
-        raise QueryArgumentError(
-            f"Unsupported encoding {encoding} for non-numeric content"
-        )
 
     if encoding == "bin":
         return Response(
@@ -124,6 +120,11 @@ def encode(content: Any, encoding: Optional[str] = "json") -> Response:
             },
         )
 
+    if not is_numeric_data(content_array):
+        raise QueryArgumentError(
+            f"Unsupported encoding {encoding} for non-numeric content"
+        )
+
     if encoding == "csv":
         return Response(
             csv_encode(content_array),
diff --git a/tasks.py b/tasks.py
index 8e851e7..e84ac07 100644
--- a/tasks.py
+++ b/tasks.py
@@ -31,7 +31,7 @@ def lint(c):
     mypy(c)
 
 
-@task
+@task(optional=["verbose", "keyword", "cov-lines"])
 def test(c, verbose=False, keyword="", cov_lines=False):
     """Test without benchmark"""
     c.run(