schema.py: Allow passing as Python object

Allow the calling application to load the schema rather than flatten-tool
Partial work around for: #447

CHANGELOG.md

@@ -12,6 +12,7 @@ and this project adheres to [Semantic Versioning](http://semver.org/spec/v2.0.0.
 - Put back some translations that were removed
 - Minor python tidyups
 - Avoid deprecation warning from the ijson package.  https://github.com/OpenDataServices/flatten-tool/pull/458
+- Allow passing of schema as Python object, and update docs
 
 ### Removed
 

docs/getting-started.rst

@@ -38,6 +38,27 @@ will print general help information.
 
 will print help information specific to that sub-command.
 
+Library
+-------
+
+The package can also be used as a Python library, e.g. to flatten
+data:
+
+.. code-block:: python
+
+    flattentool.flatten(input_filename, ...)
+
+or unflatten data:
+
+.. code-block:: python
+
+    flattentool.unflatten(input_filename, ...)
+
+In addition to the functionality available from the command-line,
+Python calls to the library functions also accept a Python object
+as a schema rather than a filename. This allows schema objects to
+be loaded/constructed by the calling application.
+
 Python Version Support
 ----------------------
 

flattentool/schema.py

@@ -152,7 +152,9 @@ def __init__(
                 _("Only one of schema_filename or root_schema_dict should be supplied")
             )
         if schema_filename:
-            if schema_filename.startswith("http"):
+            if isinstance(schema_filename, dict):
+                self.root_schema_dict = schema_filename
+            elif schema_filename.startswith("http"):
                 import requests
 
                 r = requests.get(schema_filename)