Context and scope

Desktop and graphical user interfaces (GUIs) are often wrapper to command-line tools. Especially Linux developments, the ability to effectively interface with other programs through the shell.

In December 2022, I began gaining interest in the video encoding and editing tool ffmpeg. At the same time, algorithmic content (split-screen subway surfers, am i the asshole (AITA), etc.) was taking over TikTok.

I create a basic script to turn C-Span testimony into a many split-screen subway surfers-style content (I am a big C-Span fan). After rendering hundreds of videos, uploading videos everyday was the most painstaking. To circumvent this, I built a proof-of-concept command line application. Taking Rich Nguyen’s Software Development Essentials at the time. I decided to take the opportunity to try and develop the script into a higher-quality module.

No real public modules existed for this purpose, so I decided to build one. Most of all, I think publishing this and going through three iterations was mostly an exercise of the challenges to writing good documentation.

Goals and non-goals

Learning about Selenium and how to publish and maintain open-source software is the goal. While using a Selenium-controlled browser could be argued to be overkill, experience with the technology-particularly how it applies to end-to-end integration testing — is something with which I wanted to gain personal experience.

APIs

The python module is designed with two entry points:

• upload_video for uploading single videos (also called by the command line interface)
• upload_videos is called and has multiple uploads

To manage browser configurations, browsers.py allows a selenium.webdriver object ot be created with custom:

• driver
• options
• services

Because running multiple accounts with a custom login flow was planned, auth.AuthBackend was built to support either: cookie or username and password authenticator.

Uploading a video consists of 5 sequential steps:

• _go_to_upload
• _set_video
• _set_interactivity
• _set_description
• _post_video

Both logging and configuration are singletons in __init__.py and config.toml contains all XPath selectors.

Data Storage

No persistent data storage is necessary for this project.

Code / Pseudo Code

go_to_homepage()
set_authentication_cookies()
upload_video()
set_video_metadata()
click_upload_button()

Degree of Constraints

A direct-posting REST API would eliminate teh need for tiktok-uploader all-together. Furthermore, the now-available schedule post feature would have probably never even lead me to create a library at all.

Alternatives Considered

A system built off-of the requests python module was considered. My desire to learn about Selenium, module design and module communication lead to this not being the right personal decision.

Cross-cutting concerns

Platform changes to TikTok are extremely likely to break the automation. Furthermore, language and localization difference (in the European Union TikTok will block the page before accepting cookies).