Backport bug fixes for a 4.2.1 release (#2827)

* DOC: remove unused section

* Disable uri-reference format check in jsonsschema (#2771)

* Disable uri-reference format check. Consistently use same validator across codebase

* Remove validation in SchemaInfo as not used anywhere and it referenced the wrong jsonschema draft

* Add compatibility for older jsonschema versions

* Improve comments

* Simplify validate_jsonschema

* Replace `iteritems` with `items` due to pandas deprecation (#2683)

* Add changelog entry.

* Bump version.

* Run black and flake8.

* Pin selenium in CI.

Co-authored-by: Jake VanderPlas <jakevdp@google.com>
Co-authored-by: Stefan Binder <binder_stefan@outlook.com>
Co-authored-by: Joel Ostblom <joelostblom@users.noreply.github.com>

.github/workflows/build.yml

@@ -24,6 +24,7 @@ jobs:
       run: |
         python -m pip install --upgrade pip
         pip install .[dev]
+        pip install "selenium<4.3.0"
         pip install altair_saver
     - name: Test with pytest
       run: |

.github/workflows/docbuild.yml

@@ -19,10 +19,9 @@ jobs:
       run: |
         python -m pip install --upgrade pip
         pip install .[dev]
+        pip install "selenium<4.3.0"
         pip install altair_saver
         pip install -r doc/requirements.txt
     - name: Run docbuild
       run: |
         cd doc && make ${{ matrix.build-type }}
-
-        

altair/__init__.py

@@ -1,5 +1,5 @@
 # flake8: noqa
-__version__ = "4.2.0"
+__version__ = "4.2.1"
 
 from .vegalite import *
 from . import examples

altair/utils/core.py

@@ -314,7 +314,7 @@ def to_list_if_array(val):
         else:
             return val
 
-    for col_name, dtype in df.dtypes.iteritems():
+    for col_name, dtype in df.dtypes.items():
         if str(dtype) == "category":
             # XXXX: work around bug in to_json for categorical types
             # https://github.com/pydata/pandas/issues/10778

altair/utils/display.py

@@ -4,10 +4,9 @@
 from typing import Callable, Dict
 import uuid
 
-from jsonschema import validate
-
 from .plugin_registry import PluginRegistry
 from .mimebundle import spec_to_mimebundle
+from .schemapi import validate_jsonschema
 
 
 # ==============================================================================
@@ -126,7 +125,7 @@ def _validate(self):
         # type: () -> None
         """Validate the spec against the schema."""
         schema_dict = json.loads(pkgutil.get_data(*self.schema_path).decode("utf-8"))
-        validate(self.spec, schema_dict)
+        validate_jsonschema(self.spec, schema_dict)
 
     def _repr_mimebundle_(self, include=None, exclude=None):
         """Return a MIME bundle for display in Jupyter frontends."""

altair/utils/schemapi.py

@@ -6,10 +6,11 @@
 import json
 
 import jsonschema
+import jsonschema.exceptions
 import numpy as np
 import pandas as pd
 
-
+JSONSCHEMA_VALIDATOR = jsonschema.Draft7Validator
 # If DEBUG_MODE is True, then schema objects are converted to dict and
 # validated at creation time. This slows things down, particularly for
 # larger specs, but leads to much more useful tracebacks for the user.
@@ -39,6 +40,24 @@ def debug_mode(arg):
         DEBUG_MODE = original
 
 
+def validate_jsonschema(spec, schema, resolver=None):
+    # We don't use jsonschema.validate as this would validate the schema itself.
+    # Instead, we pass the schema directly to the validator class. This is done for
+    # two reasons: The schema comes from Vega-Lite and is not based on the user
+    # input, therefore there is no need to validate it in the first place. Furthermore,
+    # the "uri-reference" format checker fails for some of the references as URIs in
+    # "$ref" are not encoded,
+    # e.g. '#/definitions/ValueDefWithCondition<MarkPropFieldOrDatumDef,
+    # (Gradient|string|null)>' would be a valid $ref in a Vega-Lite schema but
+    # it is not a valid URI reference due to the characters such as '<'.
+    validator = JSONSCHEMA_VALIDATOR(
+        schema, format_checker=JSONSCHEMA_VALIDATOR.FORMAT_CHECKER, resolver=resolver
+    )
+    error = jsonschema.exceptions.best_match(validator.iter_errors(spec))
+    if error is not None:
+        raise error
+
+
 def _subclasses(cls):
     """Breadth-first sequence of all classes which inherit from cls."""
     seen = set()
@@ -150,7 +169,7 @@ class SchemaBase(object):
     _schema = None
     _rootschema = None
     _class_is_valid_at_instantiation = True
-    _validator = jsonschema.Draft7Validator
+    _validator = JSONSCHEMA_VALIDATOR
 
     def __init__(self, *args, **kwds):
         # Two valid options for initialization, which should be handled by
@@ -440,9 +459,7 @@ def validate(cls, instance, schema=None):
         if schema is None:
             schema = cls._schema
         resolver = jsonschema.RefResolver.from_schema(cls._rootschema or cls._schema)
-        return jsonschema.validate(
-            instance, schema, cls=cls._validator, resolver=resolver
-        )
+        return validate_jsonschema(instance, schema, resolver=resolver)
 
     @classmethod
     def resolve_references(cls, schema=None):
@@ -461,7 +478,7 @@ def validate_property(cls, name, value, schema=None):
         value = _todict(value, validate=False, context={})
         props = cls.resolve_references(schema or cls._schema).get("properties", {})
         resolver = jsonschema.RefResolver.from_schema(cls._rootschema or cls._schema)
-        return jsonschema.validate(value, props.get(name, {}), resolver=resolver)
+        return validate_jsonschema(value, props.get(name, {}), resolver=resolver)
 
     def __dir__(self):
         return list(self._kwds.keys())
@@ -555,7 +572,7 @@ def from_dict(
             for possible_schema in schemas:
                 resolver = jsonschema.RefResolver.from_schema(rootschema)
                 try:
-                    jsonschema.validate(dct, possible_schema, resolver=resolver)
+                    validate_jsonschema(dct, possible_schema, resolver=resolver)
                 except jsonschema.ValidationError:
                     continue
                 else:

altair/utils/tests/test_plugin_registry.py

@@ -26,7 +26,7 @@ def test_plugin_registry():
     assert plugins.get() is None
     assert repr(plugins) == "TypedCallableRegistry(active='', registered=[])"
 
-    plugins.register("new_plugin", lambda x: x ** 2)
+    plugins.register("new_plugin", lambda x: x**2)
     assert plugins.names() == ["new_plugin"]
     assert plugins.active == ""
     assert plugins.get() is None
@@ -46,7 +46,7 @@ def test_plugin_registry():
 def test_plugin_registry_extra_options():
     plugins = GeneralCallableRegistry()
 
-    plugins.register("metadata_plugin", lambda x, p=2: x ** p)
+    plugins.register("metadata_plugin", lambda x, p=2: x**p)
     plugins.enable("metadata_plugin")
     assert plugins.get()(3) == 9
 
@@ -86,7 +86,7 @@ def test_plugin_registry_global_settings():
 def test_plugin_registry_context():
     plugins = GeneralCallableRegistry()
 
-    plugins.register("default", lambda x, p=2: x ** p)
+    plugins.register("default", lambda x, p=2: x**p)
 
     # At first there is no plugin enabled
     assert plugins.active == ""

altair/utils/tests/test_utils.py

@@ -129,9 +129,9 @@ def test_sanitize_nullable_integers():
     )
 
     df_clean = sanitize_dataframe(df)
-    assert {col.dtype.name for _, col in df_clean.iteritems()} == {"object"}
+    assert {col.dtype.name for _, col in df_clean.items()} == {"object"}
 
-    result_python = {col_name: list(col) for col_name, col in df_clean.iteritems()}
+    result_python = {col_name: list(col) for col_name, col in df_clean.items()}
     assert result_python == {
         "int_np": [1, 2, 3, 4, 5],
         "int64": [1, 2, 3, None, 5],
@@ -157,9 +157,9 @@ def test_sanitize_string_dtype():
     )
 
     df_clean = sanitize_dataframe(df)
-    assert {col.dtype.name for _, col in df_clean.iteritems()} == {"object"}
+    assert {col.dtype.name for _, col in df_clean.items()} == {"object"}
 
-    result_python = {col_name: list(col) for col_name, col in df_clean.iteritems()}
+    result_python = {col_name: list(col) for col_name, col in df_clean.items()}
     assert result_python == {
         "string_object": ["a", "b", "c", "d"],
         "string_string": ["a", "b", "c", "d"],
@@ -182,9 +182,9 @@ def test_sanitize_boolean_dtype():
     )
 
     df_clean = sanitize_dataframe(df)
-    assert {col.dtype.name for _, col in df_clean.iteritems()} == {"object"}
+    assert {col.dtype.name for _, col in df_clean.items()} == {"object"}
 
-    result_python = {col_name: list(col) for col_name, col in df_clean.iteritems()}
+    result_python = {col_name: list(col) for col_name, col in df_clean.items()}
     assert result_python == {
         "bool_none": [True, False, None],
         "none": [None, None, None],

doc/releases/changes.rst

@@ -3,6 +3,15 @@
 Altair Change Log
 =================
 
+Version 4.2.1 (released XXX XX, 2023)
+-------------------------------------
+
+Bug Fixes
+~~~~~~~~~
+- Disable uri-reference format check in jsonsschema (#2771)
+- Replace ``iteritems`` with ```items``` due to pandas deprecation (#2683)
+
+
 Version 4.2.0 (released Dec 29, 2021)
 -------------------------------------
 - Update Vega-Lite from version 4.8.1 to version 4.17.0;