How Massive Can Python Projects Really Get? A Deep Dive into Code Size and Maintainability

People often claim that dynamic languages are fun for quick development but become a refactor nightmare, yet many high‑traffic sites and famous open‑source projects (e.g., GitHub, Instagram) are built with them, prompting curiosity about how large such projects can actually grow.

The largest known dynamic‑language project is OpenStack, whose codebase has reached several million lines of Python and continues to grow, serving as a prime example of the language’s scalability.

To analyze this, the author selected several well‑known Python projects from GitHub (and a few from other repositories), using the cloc tool (v1.72) to count only Python files as of 3 January 2018, excluding other file types.

The resulting table, sorted by lines of code, shows that aside from CPython, the top three largest projects are operations‑focused, contrary to the expectation that feature‑rich applications like Odoo would rank higher.

Sentry tops the list with nearly 700 k lines of pure Python code, while three projects fall in the 300‑500 k range (including CPython). Several projects have 200 k, 100 k, or fewer lines, demonstrating that dynamic languages comfortably handle projects up to hundreds of thousands of lines, though projects reaching millions of lines inevitably face splitting challenges.

When breaking down code, blanks, and comments, Sentry again stands out with a disproportionately low comment count, suggesting limited documentation effort by its authors.

Additional metrics were added:

Average lines per file: values range from 100 to 600 lines, with no sharp concentration. The top two projects (Pandas, NumPy) are tightly related to mathematics, which often leads to more modular code.

Comment‑to‑code ratio: while excessive comments are unnecessary, a lack of comments can hinder maintenance. Projects like Ansible, NumPy, Fabric, and Salt show higher comment ratios, indicating greater author investment and potentially higher trustworthiness. CPython’s lower comment count is offset by its extensive external documentation.

File‑type distribution: besides Python files, C, HTML, JavaScript, and .PO language resource files appear. Notably, Django and Django‑CMS contain more .PO lines than Python lines, reflecting their extensive internationalization support.

The analysis underscores that code is only part of a project; ancillary work such as documentation, localization, and other non‑code tasks can consume a substantial portion of effort and should not be overlooked.