Import and initialization optimizations (#2368)

* Lazy imports of graph object hierarchy for Python 3.7+

This involved splitting validators/graph object classes back into separate files

* lazy submodule imports for plotly and plotly.io modules

* Don't force import pandas and numpy if we're only checking whether an object belongs to these packages

If pandas isn't loaded, we don't need to check whether a value is a DataFrame, and this way we
don't pay the pandas import time.

* Don't auto-import all trace type classes

* Don't construct validators up front for every object when object is created.

Create them lazily, and cache then for use across graph objects of the same type

* Test fixes

* Add optional validation using go.validate object as callable of context manager

* Delay additional imports

* dynamic to static docstrings for Figure io methods

Saves 200ms of startup time!

* Update packages/python/plotly/_plotly_utils/importers.py

Co-Authored-By: Emmanuelle Gouillart <[email protected]>

* Use module-level __dir__ function to restore IPython tab completion with Python 3.7+

* Don't fail hierarchy test when ipywidgets not installed

* cut commented print statements

* Test fix, FigureWidget is not expected to be importable when ipywidgets is not installed

Co-authored-by: Emmanuelle Gouillart <[email protected]>

Author: jonmmease

Diff for: packages/python/plotly/_plotly_utils/basevalidators.py

@@ -35,8 +35,8 @@ def to_scalar_or_list(v):
     # Python native scalar type ('float' in the example above).
     # We explicitly check if is has the 'item' method, which conventionally
     # converts these types to native scalars.
-    np = get_module("numpy")
-    pd = get_module("pandas")
+    np = get_module("numpy", should_load=False)
+    pd = get_module("pandas", should_load=False)
     if np and np.isscalar(v) and hasattr(v, "item"):
         return v.item()
     if isinstance(v, (list, tuple)):
@@ -74,7 +74,9 @@ def copy_to_readonly_numpy_array(v, kind=None, force_numeric=False):
         Numpy array with the 'WRITEABLE' flag set to False
     """
     np = get_module("numpy")
-    pd = get_module("pandas")
+
+    # Don't force pandas to be loaded, we only want to know if it's already loaded
+    pd = get_module("pandas", should_load=False)
     assert np is not None
 
     # ### Process kind ###
@@ -166,8 +168,8 @@ def is_homogeneous_array(v):
     """
     Return whether a value is considered to be a homogeneous array
     """
-    np = get_module("numpy")
-    pd = get_module("pandas")
+    np = get_module("numpy", should_load=False)
+    pd = get_module("pandas", should_load=False)
     if (
         np
         and isinstance(v, np.ndarray)
@@ -2455,6 +2457,10 @@ def validate_coerce(self, v, skip_invalid=False):
         v._plotly_name = self.plotly_name
         return v
 
+    def present(self, v):
+        # Return compound object as-is
+        return v
+
 
 class TitleValidator(CompoundValidator):
     """
@@ -2549,6 +2555,10 @@ def validate_coerce(self, v, skip_invalid=False):
 
         return v
 
+    def present(self, v):
+        # Return compound object as tuple
+        return tuple(v)
+
 
 class BaseDataValidator(BaseValidator):
     def __init__(
@@ -2559,7 +2569,7 @@ def __init__(
         )
 
         self.class_strs_map = class_strs_map
-        self._class_map = None
+        self._class_map = {}
         self.set_uid = set_uid
 
     def description(self):
@@ -2595,21 +2605,17 @@ def description(self):
 
         return desc
 
-    @property
-    def class_map(self):
-        if self._class_map is None:
-
-            # Initialize class map
-            self._class_map = {}
-
-            # Import trace classes
+    def get_trace_class(self, trace_name):
+        # Import trace classes
+        if trace_name not in self._class_map:
             trace_module = import_module("plotly.graph_objs")
-            for k, class_str in self.class_strs_map.items():
-                self._class_map[k] = getattr(trace_module, class_str)
+            trace_class_name = self.class_strs_map[trace_name]
+            self._class_map[trace_name] = getattr(trace_module, trace_class_name)
 
-        return self._class_map
+        return self._class_map[trace_name]
 
     def validate_coerce(self, v, skip_invalid=False):
+        from plotly.basedatatypes import BaseTraceType
 
         # Import Histogram2dcontour, this is the deprecated name of the
         # Histogram2dContour trace.
@@ -2621,13 +2627,11 @@ def validate_coerce(self, v, skip_invalid=False):
             if not isinstance(v, (list, tuple)):
                 v = [v]
 
-            trace_classes = tuple(self.class_map.values())
-
             res = []
             invalid_els = []
             for v_el in v:
 
-                if isinstance(v_el, trace_classes):
+                if isinstance(v_el, BaseTraceType):
                     # Clone input traces
                     v_el = v_el.to_plotly_json()
 
@@ -2641,25 +2645,25 @@ def validate_coerce(self, v, skip_invalid=False):
                     else:
                         trace_type = "scatter"
 
-                    if trace_type not in self.class_map:
+                    if trace_type not in self.class_strs_map:
                         if skip_invalid:
                             # Treat as scatter trace
-                            trace = self.class_map["scatter"](
+                            trace = self.get_trace_class("scatter")(
                                 skip_invalid=skip_invalid, **v_copy
                             )
                             res.append(trace)
                         else:
                             res.append(None)
                             invalid_els.append(v_el)
                     else:
-                        trace = self.class_map[trace_type](
+                        trace = self.get_trace_class(trace_type)(
                             skip_invalid=skip_invalid, **v_copy
                         )
                         res.append(trace)
                 else:
                     if skip_invalid:
                         # Add empty scatter trace
-                        trace = self.class_map["scatter"]()
+                        trace = self.get_trace_class("scatter")()
                         res.append(trace)
                     else:
                         res.append(None)

Diff for: packages/python/plotly/_plotly_utils/importers.py

@@ -0,0 +1,50 @@
+import importlib
+
+
+def relative_import(parent_name, rel_modules=(), rel_classes=()):
+    """
+    Helper function to import submodules lazily in Python 3.7+
+
+    Parameters
+    ----------
+    rel_modules: list of str
+        list of submodules to import, of the form .submodule
+    rel_classes: list of str
+        list of submodule classes/variables to import, of the form ._submodule.Foo
+
+    Returns
+    -------
+    tuple
+        Tuple that should be assigned to __all__, __getattr__ in the caller
+    """
+    module_names = {rel_module.split(".")[-1]: rel_module for rel_module in rel_modules}
+    class_names = {rel_path.split(".")[-1]: rel_path for rel_path in rel_classes}
+
+    def __getattr__(import_name):
+        # In Python 3.7+, lazy import submodules
+
+        # Check for submodule
+        if import_name in module_names:
+            rel_import = module_names[import_name]
+            return importlib.import_module(rel_import, parent_name)
+
+        # Check for submodule class
+        if import_name in class_names:
+            rel_path_parts = class_names[import_name].split(".")
+            rel_module = ".".join(rel_path_parts[:-1])
+            class_name = import_name
+            class_module = importlib.import_module(rel_module, parent_name)
+            return getattr(class_module, class_name)
+
+        raise AttributeError(
+            "module {__name__!r} has no attribute {name!r}".format(
+                name=import_name, __name__=parent_name
+            )
+        )
+
+    __all__ = list(module_names) + list(class_names)
+
+    def __dir__():
+        return __all__
+
+    return __all__, __getattr__, __dir__

Diff for: packages/python/plotly/_plotly_utils/optional_imports.py

@@ -12,7 +12,7 @@
 _not_importable = set()
 
 
-def get_module(name):
+def get_module(name, should_load=True):
     """
     Return module or None. Absolute import is required.
 
@@ -23,6 +23,8 @@ def get_module(name):
     """
     if name in sys.modules:
         return sys.modules[name]
+    if not should_load:
+        return None
     if name not in _not_importable:
         try:
             return import_module(name)

Diff for: packages/python/plotly/_plotly_utils/utils.py

@@ -147,7 +147,7 @@ def encode_as_sage(obj):
     @staticmethod
     def encode_as_pandas(obj):
         """Attempt to convert pandas.NaT"""
-        pandas = get_module("pandas")
+        pandas = get_module("pandas", should_load=False)
         if not pandas:
             raise NotEncodable
 
@@ -159,7 +159,7 @@ def encode_as_pandas(obj):
     @staticmethod
     def encode_as_numpy(obj):
         """Attempt to convert numpy.ma.core.masked"""
-        numpy = get_module("numpy")
+        numpy = get_module("numpy", should_load=False)
         if not numpy:
             raise NotEncodable