Bugfix cmd-line overrides and add instance method (#50)

* fixed silent fail with cmd line overrides due to calling parse_known_args. switched command line overrides to be only at the class level to prevent verbosity and clashing. Updated tests to reflect the new command line override syntax. Fixed missing help for iterables of spock class types.

* added in isinstance analog for checking if an obj is a Spock decorated class. Updated docs for Tuple len restrictions.

* WIP. Now catching the List of spock class (for repeated classes) in the cmd line override correctly. Need to figure out how to map to the correct portion of the payload. Fixed Enum overrides not getting caught correctly.

* WIP: Override of list of spock classes working in debug script but failing test.

* finally figured out the right logic for overrides. added tests to catch this in the future.

* finally figured out the right logic for overrides. added tests to catch this in the future.

* updated docstrings

* updated docs

* added missing info from the docstring

docs/advanced_features/Command-Line-Overrides.md

@@ -32,7 +32,7 @@ class ModelConfig:
     save_path: SavePath
     n_features: int
     dropout: Optional[List[float]]
-    hidden_sizes: Tuple[int] = (32, 32, 32)
+    hidden_sizes: Tuple[int, int, int] = (32, 32, 32)
     activation: Activation = 'relu'
     optimizer: Optimizer
     cache_path: Optional[str]
@@ -70,14 +70,52 @@ But with command line overrides we can also pass parameter arguments to override
 file:
 
 ```bash
-$ python tutorial.py --config tutorial.yaml --cache_path /tmp/trash
+$ python tutorial.py --config tutorial.yaml --DataConfig.cache_path /tmp/trash
 ```
 
-Each parameter can be overridden at the global level or the class specific level with the syntax `--name.parameter`. For
-instance, our previous example would override any parameters named `cache_path` regardless of what class they are 
-defined in. In this case `cache_path` in both `ModelConfig` and `DataConfig`. To override just a class specific value 
-we would use the class specific override:
+Each parameter can be overridden **ONLY** at the class specific level with the syntax `--classname.parameter`. For
+instance, our previous example would only override the `DataConfig.cache_path` and not the `ModelConfig.cache_path` even
+though they have the same parameter name (due to the different class names).
 
 ```bash
 $ python tutorial.py --config tutorial.yaml --DataConfig.cache_path /tmp/trash
+```
+
+### Overriding List/Tuple of Repeated `@spock` Classes 
+
+For `List` of Repeated `@spock` Classes the syntax is slightly different to allow for the repeated nature of the type.
+Given the below example code:
+
+```python
+from spock.config import spock
+from typing import List
+
+
+@spock
+class NestedListStuff:
+    one: int
+    two: str
+
+@spock
+class TypeConfig:
+    nested_list: List[NestedListStuff] # To Set Default Value append '= NestedListStuff'
+```
+
+With YAML definitions:
+
+```yaml
+# Nested List configuration
+nested_list: NestedListStuff
+NestedListStuff:
+    - one: 10
+      two: hello
+    - one: 20
+      two: bye
+```
+
+We could override the parameters like so (note that the len must match the defined length from the YAML):
+
+```bash
+$ python tutorial.py --config tutorial.yaml --TypeConfig.nested_list.NestedListStuff.one [1,2] \
+--TypeConfig.nested_list.NestedListStuff.two [ciao,ciao]
 ```

docs/advanced_features/Defaults.md

@@ -34,7 +34,7 @@ class ModelConfig:
     lr: float = 0.01
     n_features: int
     dropout: List[float]
-    hidden_sizes: Tuple[int] = (32, 32, 32)
+    hidden_sizes: Tuple[int, int, int] = (32, 32, 32)
     activation: Activation = 'relu'
 ```
 

docs/advanced_features/Inheritance.md

@@ -38,7 +38,7 @@ class ModelConfig:
     save_path: SavePath
     n_features: int
     dropout: Optional[List[float]]
-    hidden_sizes: Tuple[int] = (32, 32, 32)
+    hidden_sizes: Tuple[int, int, int] = (32, 32, 32)
     activation: Activation = 'relu'
     optimizer: Optimizer
 

docs/advanced_features/Local-Definitions.md

@@ -35,7 +35,7 @@ class ModelConfig:
     save_path: SavePath
     n_features: int
     dropout: Optional[List[float]]
-    hidden_sizes: Tuple[int] = (32, 32, 32)
+    hidden_sizes: Tuple[int, int, int] = (32, 32, 32)
     activation: Activation = 'relu'
     optimizer: Optimizer
     cache_path: Optional[str]

docs/advanced_features/Optional-Parameters.md

@@ -34,7 +34,7 @@ class ModelConfig:
     lr: float = 0.01
     n_features: int
     dropout: Optional[List[float]]
-    hidden_sizes: Tuple[int] = (32, 32, 32)
+    hidden_sizes: Tuple[int, int, int] = (32, 32, 32)
     activation: Activation = 'relu'
 ```
 

docs/advanced_features/Parameter-Groups.md

@@ -36,7 +36,7 @@ class ModelConfig:
     save_path: SavePath
     n_features: int
     dropout: Optional[List[float]]
-    hidden_sizes: Tuple[int] = (32, 32, 32)
+    hidden_sizes: Tuple[int, int, int] = (32, 32, 32)
     activation: Activation = 'relu'
     optimizer: Optimizer
 

docs/basic_tutorial/Building.md

@@ -25,7 +25,7 @@ class Activation(Enum):
 class ModelConfig:
     n_features: int
     dropout: List[float]
-    hidden_sizes: Tuple[int]
+    hidden_sizes: Tuple[int, int, int]
     activation: Activation
 ```
 

docs/basic_tutorial/Configuration-Files.md

@@ -31,7 +31,7 @@ class Activation(Enum):
 class ModelConfig:
     n_features: int
     dropout: List[float]
-    hidden_sizes: Tuple[int]
+    hidden_sizes: Tuple[int, int, int]
     activation: Activation
 ```
 

docs/basic_tutorial/Define.md

@@ -20,9 +20,11 @@ standard library while `Enum` is within the `enum` standard library):
 | int | Optional[int] | Basic integer type parameter (e.g. 2) |
 | str | Optional[str] | Basic string type parameter (e.g. 'foo') |
 | List[type] | Optional[List[type]] | Basic list type parameter of base types such as int, float, etc. (e.g. [10.0, 2.0]) |
-| Tuple[type] | Optional[Tuple[type]] | Basic tuple type parameter of base types such as int, float, etc. (e.g. (10, 2)) |
+| Tuple[type] | Optional[Tuple[type]] | Basic tuple type parameter of base types such as int, float, etc. Length enforced unlike List. (e.g. (10, 2)) |
 | Enum | Optional[Enum] | Parameter that must be from a defined set of values of base types such as int, float, etc. |
 
+Use `List` types when the length of the `Iterable` is not fixed and `Tuple` when length needs to be strictly enforced.
+
 Parameters that are specified without the `Optional[]` type will be considered **REQUIRED** and therefore will raise an
 Exception if not value is specified. 
 
@@ -58,7 +60,7 @@ class Activation(Enum):
 class ModelConfig:
     n_features: int
     dropout: List[float]
-    hidden_sizes: Tuple[int]
+    hidden_sizes: Tuple[int, int, int]
     activation: Activation
 ```
 
@@ -103,7 +105,7 @@ class ModelConfig:
     save_path: SavePath
     n_features: int
     dropout: List[float]
-    hidden_sizes: Tuple[int]
+    hidden_sizes: Tuple[int, int, int]
     activation: Activation
 ```
 
@@ -123,11 +125,11 @@ spock Basic Tutorial
 configuration(s):
 
   ModelConfig (Main model configuration for a basic neural net)
-    save_path       Optional[SavePath]    spock special keyword -- path to write out spock config state (default: None)
-    n_features      int                   number of data features 
-    dropout         List[float]           dropout rate for each layer 
-    hidden_sizes    Tuple[int]            hidden size for each layer 
-    activation      Activation            choice from the Activation enum of the activation function to use 
+    save_path       Optional[SavePath]      spock special keyword -- path to write out spock config state (default: None)
+    n_features      int                     number of data features 
+    dropout         List[float]             dropout rate for each layer 
+    hidden_sizes    Tuple[int, int, int]    hidden size for each layer 
+    activation      Activation              choice from the Activation enum of the activation function to use 
 
   Activation (Options for activation functions)
     relu    str    relu activation 
@@ -141,8 +143,9 @@ In another file let's write our simple neural network code: `basic_nn.py`
 
 Notice that even before we've built and linked all of the related `spock` components together we are referencing the 
 parameters we have defined in our `spock` class. Below we are passing in the `ModelConfig` class as a parameter 
-`model_config` to the `__init__` function where we can then access the parameters with `.` notation. We could have 
-also passed in individual parameters instead if that is the preferred syntax.
+`model_config` to the `__init__` function where we can then access the parameters with `.` notation (if we import 
+the `ModelConfig` class here and add it as a type hint to `model_config` most IDE auto-complete will work out of the 
+box). We could have also passed in individual parameters instead if that is the preferred syntax.
 
 ```python
 import torch.nn as nn

docs/basic_tutorial/Saving.md

@@ -24,7 +24,7 @@ class ModelConfig:
     save_path: SavePath
     n_features: int
     dropout: List[float]
-    hidden_sizes: Tuple[int]
+    hidden_sizes: Tuple[int, int, int]
     activation: Activation
 ```
 
@@ -83,7 +83,26 @@ def main():
     # A simple description
     description = 'spock Tutorial'
     # Build out the parser by passing in Spock config objects as *args after description
-    config = ConfigArgBuilder(ModelConfig, desc=description, create_save_path=True).save().generate()
+    config = ConfigArgBuilder(ModelConfig, desc=description).save(create_save_path=True).generate()
+    # One can now access the Spock config object by class name with the returned namespace
+    # For instance...
+    print(config.ModelConfig)
+```
+
+### Override UUID Filename
+
+By default `spock` uses an automatically generated UUID as the filename when saving. This can be overridden with the
+`file_name` keyword argument. The specified filename will be appended with .spock.cfg.file_extension (e.g. .yaml, 
+.toml or. json).
+
+In: `tutorial.py`
+
+```python
+def main():
+    # A simple description
+    description = 'spock Tutorial'
+    # Build out the parser by passing in Spock config objects as *args after description
+    config = ConfigArgBuilder(ModelConfig, desc=description).save(file_name='cool_name_here').generate()
     # One can now access the Spock config object by class name with the returned namespace
     # For instance...
     print(config.ModelConfig)

examples/tutorial/advanced/tutorial.py

@@ -25,7 +25,7 @@ class ModelConfig:
     save_path: SavePath
     n_features: int
     dropout: Optional[List[float]]
-    hidden_sizes: Tuple[int] = (32, 32, 32)
+    hidden_sizes: Tuple[int, int, int] = (32, 32, 32)
     activation: Activation = 'relu'
     optimizer: Optimizer
     cache_path: Optional[str]

examples/tutorial/basic/tutorial.py

@@ -35,7 +35,7 @@ class ModelConfig:
     save_path: SavePath
     n_features: int
     dropout: List[float]
-    hidden_sizes: Tuple[int]
+    hidden_sizes: Tuple[int, int, int]
     activation: Activation
 
 

spock/backend/attr/builder.py

@@ -46,54 +46,7 @@ def print_usage_and_exit(self, msg=None, sys_exit=True, exit_code=1):
             sys.exit(exit_code)
 
     def _handle_help_info(self):
-        # List to catch Enum classes and handle post spock wrapped attr classes
-        enum_list = []
-        for attrs_class in self.input_classes:
-            # Split the docs into class docs and any attribute docs
-            class_doc, attr_docs = self._split_docs(attrs_class)
-            print('  ' + attrs_class.__name__ + f' ({class_doc})')
-            # Keep a running info_dict of all the attribute level info
-            info_dict = {}
-            for val in attrs_class.__attrs_attrs__:
-                # If the type is an enum we need to handle it outside of this attr loop
-                # Match the style of nested enums and return a string of module.name notation
-                if isinstance(val.type, EnumMeta):
-                    enum_list.append(f'{val.type.__module__}.{val.type.__name__}')
-                # if there is a type (implied Iterable) -- check it for nested Enums
-                nested_enums = self._extract_enum_types(val.metadata['type']) if 'type' in val.metadata else []
-                if len(nested_enums) > 0:
-                    enum_list.extend(nested_enums)
-                # Grab the base or type info depending on what is provided
-                type_string = repr(val.metadata['type']) if 'type' in val.metadata else val.metadata['base']
-                # Regex out the typing info if present
-                type_string = re.sub(r'typing.', '', type_string)
-                # Regex out any nested_enums that have module path information
-                for enum_val in nested_enums:
-                    split_enum = f"{'.'.join(enum_val.split('.')[:-1])}."
-                    type_string = re.sub(split_enum, '', type_string)
-                # Regex the string to see if it matches any Enums in the __main__ module space
-                # for val in sys.modules
-                # Construct the type with the metadata
-                if 'optional' in val.metadata:
-                    type_string = f"Optional[{type_string}]"
-                info_dict.update(self._match_attribute_docs(val.name, attr_docs, type_string, val.default))
-            self._handle_attributes_print(info_dict=info_dict)
-        # Convert the enum list to a set to remove dupes and then back to a list so it is iterable
-        enum_list = list(set(enum_list))
-        # Iterate any Enum type classes
-        for enum in enum_list:
-            enum = self._get_enum_from_sys_modules(enum)
-            # Split the docs into class docs and any attribute docs
-            class_doc, attr_docs = self._split_docs(enum)
-            print('  ' + enum.__name__ + f' ({class_doc})')
-            info_dict = {}
-            for val in enum:
-                info_dict.update(self._match_attribute_docs(
-                    attr_name=val.name,
-                    attr_docs=attr_docs,
-                    attr_type_str=type(val.value).__name__
-                ))
-            self._handle_attributes_print(info_dict=info_dict)
+        self._attrs_help(self.input_classes)
 
     def _handle_arguments(self, args, class_obj):
         attr_name = class_obj.__name__