[BUG] Reenable HTML viz hooks for np.ndarray and PIL Images

daft/series.py

@@ -97,7 +97,7 @@ def from_pylist(data: list, name: str = "list_series", pyobj: str = "allow") ->
             raise ValueError(f"pyobj: expected either 'allow', 'disallow', or 'force', but got {pyobj})")
 
         if pyobj == "force":
-            pys = PySeries.from_pylist(name, data)
+            pys = PySeries.from_pylist(name, data, pyobj=pyobj)
             return Series._from_pyseries(pys)
 
         try:
@@ -106,7 +106,7 @@ def from_pylist(data: list, name: str = "list_series", pyobj: str = "allow") ->
         except pa.lib.ArrowInvalid:
             if pyobj == "disallow":
                 raise
-            pys = PySeries.from_pylist(name, data)
+            pys = PySeries.from_pylist(name, data, pyobj=pyobj)
             return Series._from_pyseries(pys)
 
     @classmethod

daft/viz/html_viz_hooks.py

@@ -14,13 +14,13 @@
 _VIZ_HOOKS_REGISTRY = {}
 
 
-def register_viz_hook(klass: type[HookClass], hook: Callable[[HookClass], str]):
+def register_viz_hook(klass: type[HookClass], hook: Callable[[object], str]):
     """Registers a visualization hook that returns the appropriate HTML for
     visualizing a specific class in HTML"""
     _VIZ_HOOKS_REGISTRY[klass] = hook
 
 
-def get_viz_hook(val: HookClass) -> Callable[[HookClass], str] | None:
+def get_viz_hook(val: object) -> Callable[[object], str] | None:
     for klass in _VIZ_HOOKS_REGISTRY:
         if isinstance(val, klass):
             return _VIZ_HOOKS_REGISTRY[klass]
@@ -45,7 +45,7 @@ def get_viz_hook(val: HookClass) -> Callable[[HookClass], str] | None:
 
 if HAS_PILLOW:
 
-    def _viz_pil_image(val: PIL.Image.Image):
+    def _viz_pil_image(val: PIL.Image.Image) -> str:
         img = val.copy()
         img.thumbnail((128, 128))
         bio = io.BytesIO()
@@ -57,7 +57,7 @@ def _viz_pil_image(val: PIL.Image.Image):
 
 if HAS_NUMPY:
 
-    def _viz_numpy(val: np.ndarray):
+    def _viz_numpy(val: np.ndarray) -> str:
         return f"&ltnp.ndarray<br>shape={val.shape}<br>dtype={val.dtype}&gt"
 
     register_viz_hook(np.ndarray, _viz_numpy)

src/array/ops/repr.rs

@@ -216,15 +216,43 @@ impl_array_html_value!(ListArray);
 impl_array_html_value!(FixedSizeListArray);
 impl_array_html_value!(StructArray);
 impl_array_html_value!(ExtensionArray);
-
-#[cfg(feature = "python")]
-impl_array_html_value!(crate::datatypes::PythonArray);
-
 impl_array_html_value!(DateArray);
 impl_array_html_value!(DurationArray);
 impl_array_html_value!(TimestampArray);
 impl_array_html_value!(EmbeddingArray);
 
+#[cfg(feature = "python")]
+impl crate::datatypes::PythonArray {
+    pub fn html_value(&self, idx: usize) -> String {
+        use pyo3::prelude::*;
+
+        let val = self.get(idx);
+
+        let custom_viz_hook_result: Option<String> = Python::with_gil(|py| {
+            // Find visualization hooks for this object's class
+            let pyany = val.into_ref(py);
+            let get_viz_hook = py
+                .import("daft.viz.html_viz_hooks")?
+                .getattr("get_viz_hook")?;
+            let hook = get_viz_hook.call1((pyany,))?;
+
+            if hook.is_none() {
+                Ok(None)
+            } else {
+                hook.call1((pyany,))?.extract()
+            }
+        })
+        .unwrap();
+
+        match custom_viz_hook_result {
+            None => html_escape::encode_text(&self.str_value(idx).unwrap())
+                .into_owned()
+                .replace('\n', "<br />"),
+            Some(result) => result,
+        }
+    }
+}
+
 impl<T> DataArray<T>
 where
     T: DaftNumericType,

src/python/series.rs

@@ -31,19 +31,22 @@ impl PySeries {
 
     // This ingests a Python list[object] directly into a Rust PythonArray.
     #[staticmethod]
-    pub fn from_pylist(name: &str, pylist: &PyAny) -> PyResult<Self> {
+    pub fn from_pylist(name: &str, pylist: &PyAny, pyobj: &str) -> PyResult<Self> {
         let vec_pyobj: Vec<PyObject> = pylist.extract()?;
         let py = pylist.py();
-        let dtype = infer_daft_dtype_for_sequence(&vec_pyobj, py)?;
+
+        let dtype = match pyobj {
+            "force" => DataType::Python,
+            "allow" => infer_daft_dtype_for_sequence(&vec_pyobj, py)?.unwrap_or(DataType::Python),
+            "disallow" => panic!("Cannot create a Series from a pylist and being strict about only using Arrow types by setting pyobj=disallow"),
+            _ => panic!("Unsupported pyobj behavior when creating Series from pylist: {}", pyobj)
+        };
         let arrow_array: Box<dyn arrow2::array::Array> =
             Box::new(PseudoArrowArray::<PyObject>::from_pyobj_vec(vec_pyobj));
         let field = Field::new(name, DataType::Python);
 
         let data_array = DataArray::<PythonType>::new(field.into(), arrow_array)?;
-        let series = match dtype {
-            Some(dtype) => data_array.cast(&dtype)?,
-            None => data_array.into_series(),
-        };
+        let series = data_array.cast(&dtype)?;
         Ok(series.into())
     }
 

tests/dataframe/test_logical_type.py

@@ -32,7 +32,7 @@ def test_image_type_df(from_pil_imgs) -> None:
     ]
     if from_pil_imgs:
         data = [Image.fromarray(arr, mode="RGB") if arr is not None else None for arr in data]
-    df = daft.from_pydict({"index": np.arange(len(data)), "image": Series.from_pylist(data, pyobj="force")})
+    df = daft.from_pydict({"index": np.arange(len(data)), "image": Series.from_pylist(data, pyobj="allow")})
 
     image_expr = col("image")
     if not from_pil_imgs:

tests/dataframe/test_repr.py

@@ -2,7 +2,9 @@
 
 import re
 
+import numpy as np
 import pandas as pd
+from PIL import Image
 
 import daft
 
@@ -176,3 +178,56 @@ def test_repr_with_html_string():
     for i in range(3):
         assert f"<div>body{i}</div>" in non_html_table
         assert f"<tr><td>&lt;div&gt;body{i}&lt;/div&gt;</td></tr>" in html_table
+
+
+class MyObj:
+    def __repr__(self) -> str:
+        return "myobj-custom-repr"
+
+
+def test_repr_html_custom_hooks():
+    img = Image.fromarray(np.ones((3, 3)).astype(np.uint8))
+    arr = np.ones((3, 3))
+
+    df = daft.from_pydict(
+        {
+            "objects": daft.Series.from_pylist([MyObj() for _ in range(3)], pyobj="force"),
+            "np": daft.Series.from_pylist([arr for _ in range(3)], pyobj="force"),
+            "pil": daft.Series.from_pylist([img for _ in range(3)], pyobj="force"),
+        }
+    )
+    df.collect()
+
+    assert (
+        df.__repr__()
+        == """+-------------------+-------------+----------------------------------+
+| objects           | np          | pil                              |
+| Python            | Python      | Python                           |
++-------------------+-------------+----------------------------------+
+| myobj-custom-repr | [[1. 1. 1.] | <PIL.Image.Image image mode=L... |
+|                   |  [1. 1. 1.] |                                  |
+|                   |  [1. ...    |                                  |
++-------------------+-------------+----------------------------------+
+| myobj-custom-repr | [[1. 1. 1.] | <PIL.Image.Image image mode=L... |
+|                   |  [1. 1. 1.] |                                  |
+|                   |  [1. ...    |                                  |
++-------------------+-------------+----------------------------------+
+| myobj-custom-repr | [[1. 1. 1.] | <PIL.Image.Image image mode=L... |
+|                   |  [1. 1. 1.] |                                  |
+|                   |  [1. ...    |                                  |
++-------------------+-------------+----------------------------------+
+
+(Showing first 3 of 3 rows)"""
+    )
+
+    html_repr = df._repr_html_()
+
+    # Assert that MyObj is correctly displayed in html repr (falls back to __repr__)
+    assert "myobj-custom-repr" in html_repr
+
+    # Assert that PIL viz hook correctly triggers in html repr
+    assert 'alt="<PIL.Image.Image image mode=L size=3x3' in html_repr
+    assert '<img style="max-height:128px;width:auto" src="data:image/png;base64,' in html_repr
+
+    # Assert that numpy array viz hook correctly triggers in html repr
+    assert "<td>&ltnp.ndarray<br>shape=(3, 3)<br>dtype=float64&gt</td><td>" in html_repr

tests/series/test_image.py

@@ -178,7 +178,7 @@ def test_image_pil_inference(fixed_shape, mode):
             if arr is not None:
                 arr[..., -1] = 255
     imgs = [Image.fromarray(arr, mode=mode) if arr is not None else None for arr in arrs]
-    s = Series.from_pylist(imgs, pyobj="force")
+    s = Series.from_pylist(imgs, pyobj="allow")
     assert s.datatype() == DataType.image(mode)
     out = s.to_pylist()
     if num_channels == 1:
@@ -206,7 +206,12 @@ def test_image_pil_inference_mixed():
         else None
         for arr in arrs
     ]
+
+    # Forcing should still create Python Series
     s = Series.from_pylist(imgs, pyobj="force")
+    assert s.datatype() == DataType.python()
+
+    s = Series.from_pylist(imgs, pyobj="allow")
     assert s.datatype() == DataType.image()
     out = s.to_pylist()
     arrs[3] = np.expand_dims(arrs[3], axis=-1)