Cache get decoder class from string (#117)

* Moved the logic from obtaining a decoder class from a string to its own method. This allows caching, which greatly improves performance.

* Added default maxsize arg to lru_cache for Python <3.8 compatibility

* Revert the caching changes of get_decoder_class, instead cache convert_type_string and compile the regex for faster execution.

* Revert the caching changes of get_decoder_class, instead cache convert_type_string and compile the regex for faster execution.

* Okay, actually reverted the initial changes this time.

Author: thewhaleking

scalecodec/base.py

@@ -17,6 +17,7 @@
 import re
 import warnings
 from abc import ABC, abstractmethod
+from functools import lru_cache
 from typing import Optional, TYPE_CHECKING, Union
 
 from scalecodec.constants import TYPE_DECOMP_MAX_RECURSIVE
@@ -62,8 +63,11 @@ def __init__(self, config_id=None, ss58_format=None, only_primitives_on_init=Fal
         self.only_primitives_on_init = only_primitives_on_init
         self.ss58_format = ss58_format
         self.implements_scale_info = implements_scale_info
+        self.arrow_match_re = re.compile(r'^([^<]*)<(.+)>$')
+        self.bracket_match_re = re.compile(r'^\[([A-Za-z0-9]+); ([0-9]+)\]$')
 
     @classmethod
+    @lru_cache(maxsize=128)
     def convert_type_string(cls, name):
 
         name = re.sub(r'T::', "", name)
@@ -130,7 +134,7 @@ def get_decoder_class(self, type_string: Union[str, dict]):
             if type_string[-1:] == '>':
 
                 # Extract sub types
-                type_parts = re.match(r'^([^<]*)<(.+)>$', type_string)
+                type_parts = self.arrow_match_re.match(type_string)
 
                 if type_parts:
                     type_parts = type_parts.groups()
@@ -151,7 +155,7 @@ def get_decoder_class(self, type_string: Union[str, dict]):
                 decoder_class.build_type_mapping()
 
             elif type_string[0] == '[' and type_string[-1] == ']':
-                type_parts = re.match(r'^\[([A-Za-z0-9]+); ([0-9]+)\]$', type_string)
+                type_parts = self.bracket_match_re.match(type_string)
 
                 if type_parts:
                     type_parts = type_parts.groups()
@@ -1076,4 +1080,3 @@ def generate_type_decomposition(cls, _recursion_level: int = 0, max_recursion: i
         return cls.__name__.lower()
 
 
-