NBL-48 Make correct Python identifiers out of GridGain binary type names

This closes #297

Author: isapego

docs/datatypes/parsers.rst

@@ -131,7 +131,7 @@ with your data, in to some API function as a *type conversion hint*.
 |0x1d       |`Enum array`_       |iterable/list                  |:class:`~pygridgain.datatypes.standard.EnumArrayObject`           |
 +-----------+--------------------+-------------------------------+------------------------------------------------------------------+
 |0x67       |`Complex object`_   |object                         |:class:`~pygridgain.datatypes.complex.BinaryObject`               |
-+-----------+--------------------+-------------------------------+---------------------------------------------------------------- -+
++-----------+--------------------+-------------------------------+------------------------------------------------------------------+
 |0x1b       |`Wrapped data`_     |tuple                          |:class:`~pygridgain.datatypes.complex.WrappedDataObject`          |
 +-----------+--------------------+-------------------------------+------------------------------------------------------------------+
 

docs/examples.rst

@@ -240,7 +240,23 @@ Here you can see how :class:`~pygridgain.binary.GenericObjectMeta` uses
 `attrs`_ package internally for creating nice `__init__()` and `__repr__()`
 methods.
 
-You can reuse the autogenerated class for subsequent writes:
+In this case the autogenerated dataclass's name `Person` is exactly matches
+the type name of the Complex object it represents (the content of the
+:py:attr:`~pygridgain.datatypes.base.GridGainDataTypeProps.type_name`
+property). But when Complex object's class name contains characters, that
+can not be used in a Python identifier, for example:
+
+- `.`, when fully qualified Java class names are used,
+- `$`, a common case for Scala classes,
+- `+`, internal class name separator in C#,
+
+then `pygridgain` can not maintain this match. In such cases `pyignite` tries
+to sanitize a type name to derive a “good” dataclass name from it.
+
+If your code needs consistent naming between the server and the client, make
+sure that your GridGain cluster is configured to use `simple class names`_.
+
+Anyway, you can reuse the autogenerated dataclass for subsequent writes:
 
 .. literalinclude:: ../examples/binary_basics.py
   :language: python
@@ -595,3 +611,4 @@ with the following message:
 .. _openssl: https://www.openssl.org/docs/manmaster/man1/openssl.html
 .. _Authentication: https://apacheignite.readme.io/docs/advanced-security#section-authentication
 .. _attrs: https://pypi.org/project/attrs/
+.. _simple class names: https://apacheignite.readme.io/docs/binary-marshaller#binary-name-mapper-and-binary-id-mapper

pygridgain/client.py

@@ -42,6 +42,7 @@
 
 from collections import defaultdict, OrderedDict
 import random
+import re
 from typing import Dict, Iterable, List, Optional, Tuple, Type, Union
 
 from .api.binary import get_binary_type, put_binary_type
@@ -56,7 +57,8 @@
     BinaryTypeError, CacheError, ReconnectError, SQLError, connection_errors,
 )
 from .utils import (
-    entity_id, schema_id, select_version, status_to_exception, is_iterable
+    capitalize, entity_id, schema_id, process_delimiter, select_version,
+    status_to_exception, is_iterable,
 )
 from .binary import GenericObjectMeta
 
@@ -82,6 +84,10 @@ class Client:
     _current_node: int = None
     _nodes: List[Connection] = None
 
+    # used for Complex object data class names sanitizing
+    _identifier = re.compile(r'[^0-9a-zA-Z_.+$]', re.UNICODE)
+    _ident_start = re.compile(r'^[^a-zA-Z_]+', re.UNICODE)
+
     affinity_version: Optional[Tuple] = None
     protocol_version: Optional[Tuple] = None
 
@@ -383,11 +389,40 @@ def _sync_binary_registry(self, type_id: int):
             for schema in type_info['schemas']:
                 if not self._registry[type_id].get(schema_id(schema), None):
                     data_class = self._create_dataclass(
-                        type_info['type_name'],
+                        self._create_type_name(type_info['type_name']),
                         schema,
                     )
                     self._registry[type_id][schema_id(schema)] = data_class
 
+    @classmethod
+    def _create_type_name(cls, type_name: str) -> str:
+        """
+        Creates Python data class name from GridGain binary type name.
+
+        Handles all the special cases found in
+        `java.org.apache.ignite.binary.BinaryBasicNameMapper.simpleName()`.
+        Tries to adhere to PEP8 along the way.
+        """
+
+        # general sanitizing
+        type_name = cls._identifier.sub('', type_name)
+
+        # - name ending with '$' (Scala)
+        # - name + '$' + some digits (anonymous class)
+        # - '$$Lambda$' in the middle
+        type_name = process_delimiter(type_name, '$')
+
+        # .NET outer/inner class delimiter
+        type_name = process_delimiter(type_name, '+')
+
+        # Java fully qualified class name
+        type_name = process_delimiter(type_name, '.')
+
+        # start chars sanitizing
+        type_name = capitalize(cls._ident_start.sub('', type_name))
+
+        return type_name
+
     def register_binary_type(
         self, data_class: Type, affinity_key_field: str = None,
     ):

pygridgain/utils.py

@@ -304,3 +304,18 @@ def run(self):
         if not self.finished.is_set():
             self.function(*self.args, **self.kwargs)
         self.finished.set()
+
+
+def capitalize(string: str) -> str:
+    """
+    Capitalizing the string, assuming the first character is a letter.
+    Does not touch any other character, unlike the `string.capitalize()`.
+    """
+    return string[:1].upper() + string[1:]
+
+
+def process_delimiter(name: str, delimiter: str) -> str:
+    """
+    Splits the name by delimiter, capitalize each part, merge.
+    """
+    return ''.join([capitalize(x) for x in name.split(delimiter)])

tests/test_binary.py

@@ -278,3 +278,29 @@ class MyBinaryTypeV2(
     assert not hasattr(result, 'test_bool')
 
     migrate_cache.destroy()
+
+
+def test_complex_object_names(client):
+    """
+    Test the ability to work with Complex types, which names contains symbols
+    not suitable for use in Python identifiers.
+    """
+    type_name = 'Non.Pythonic#type-name$'
+    key = 'key'
+    data = 'test'
+
+    class NonPythonicallyNamedType(
+        metaclass=GenericObjectMeta,
+        type_name=type_name,
+        schema=OrderedDict([
+            ('field', String),
+        ])
+    ):
+        pass
+
+    cache = client.get_or_create_cache('test_name_cache')
+    cache.put(key, NonPythonicallyNamedType(field=data))
+
+    obj = cache.get(key)
+    assert obj.type_name == type_name, 'Complex type name mismatch'
+    assert obj.field == data, 'Complex object data failure'