Architecture on Ashlynn Antrobus

Unit Testing Difficulty is an Indicator of (Low) Code Quality

Mon, 26 Aug 2024 23:01:24 +0000

When I first read about unit testing, I’ll admit, it sounded like one of those things that only seasoned developers would bother with. It seemed like a maze of confusing terms and intimidating processes that only experts could navigate. I didn’t write my first test for over a year after I started my coding journey. Until then, my version of testing was typing python my\_project.py in the terminal. Or, I will sheepishly admit, just hitting the “Run” button in VS Code (and Replit before that). Then waiting for it to crash and debugging from there. It was my version of compiled languages developers who claim that if the code compiles, it must be free of errors.

Evolution of a codebase

Mon, 15 Jul 2024 04:27:52 +0000

When I wrote the first version of Lorebinders, I hadn’t yet started exploring object oriented programming. v1 has comprised by a series of functions strung in a row, one, calling the next.

Unfortunately (or maybe fortunately!), that code was gone before I discovered version control. It is gone for ever.

For the first refactoring, I organized it a little better, but still depended sole on functions like these:

def search_names(chapters: list, folder_name: str, num_chapters: int, character_lists: list, character_lists_index: int) -> list:

 character_lists_path = os.path.join(folder_name, "character_lists.json")
 role_script = ner_role_script(folder_name)
 model = "gpt_three"
 max_tokens = 1000
 temperature = 0.2

 with tqdm(total = num_chapters, unit = "Chapter", ncols = 40, bar_format = "|{l_bar}{bar}|", position = 0, leave = True) as progress_bar:
 for chapter_index, chapter in enumerate(chapters):
 progress_bar.set_description(f"\033[92mProcessing chapter {chapter_index + 1} of {num_chapters}", refresh = True)
 if chapter_index < character_lists_index:
 progress_bar.update(1)
 continue
 chapter_number = chapter_index + 1
 prompt = f"Text: {chapter}"
 character_list = cf.call_gpt_api(model, prompt, role_script, temperature, max_tokens)
 chapter_tuple = (chapter_number, character_list)
 character_lists.append(chapter_tuple)
 cf.append_json_file(chapter_tuple, character_lists_path)
 progress_bar.update(1)
 cf.clear_screen()
 return character_lists

When I started v2, leaning into object oriented programing principles, I quickly learned that classes don’t make dependency and coupling any simpler. I needed to initialize the API class, get the text from the chapter, and save the response to the chapter. It looked easy to organize when each class was was separate.

Back to the future and import annotations

Sun, 30 Jun 2024 22:56:25 +0000

After last week, I thought I was done with module import errors. Turns out, I had only just started with them.

After months of hard work, I have finally written all of the code for my LoreBinders project. Now to see if it works!

So I wrote tests for everything, typed pytest into the terminal and hit enter.

ImportError while importing test module 'vscode\ProsePal\lorebinders\tests\test_ai_factory.py'.
Hint: make sure your test modules/packages have valid Python names.
Traceback:
 from lorebinders._managers import EmailManager, RateLimitManager
src\lorebinders\_managers.py:4: in 
from _types import AIModelRegistry, APIProvider, Model, ModelFamily
src\lorebinders_types.py:12: in
from _managers import AIProviderManager, EmailManager, ErrorManager
src\lorebinders_managers.py:4: in
from _types import AIModelRegistry, APIProvider, Model, ModelFamily
E ImportError: cannot import name 'AIModelRegistry' from partially initialized module '_types' (most likely due to a circular import)

15 times. Different modules, same problem.

Architecture on Ashlynn Antrobus

Unit Testing Difficulty is an Indicator of (Low) Code Quality

Evolution of a codebase

Back to the __future__ and import annotations

Back to the future and import annotations