Software Engineering Project 1

This project takes in pairs of xml and png files. It parses the xml files and draws bounding boxes around the leaf-level elements on the corresponding UI screenshots.

Steps for Running Project

1. Clone this repository with git clone [repo-link]
2. Navigate into the repository with cd software_engineering_project_1
3. Optionally, create a new virtual environment with pyenv virtualenv [python-version] [venv-name]. If you don't have pyenv installed, refer to this guide
4. Run pip install -r requirements.txt to install appropriate python packages. If you don't have pip installed, refer to the documentation for installation instructions.
5. With python packages installed, run python bounding_boxes.py
6. Respond to the input prompts accordingly -- If you leave the inputs blank, the program will draw the bounding boxes for all xml-png pairs in Programming-Assignment-Data.
7. A new folder will be created, called outlined_images, which contains the UI screenshots with bounding boxes. Check the images in this folder for correctness.

Explaination of Design

The program in bounding_boxes.py takes in user input for specifying the directory containing xml-png pairs, and also allows the user to specify a single file to parse instead of running on the entire directory. This allows the user a bit more control over the program. I included guardrails anywhere I thought they were necessary, such as checking that user-inputted files actually exist, and try-excepting possible points of failure like parsing XML files or saving the annotated image in local storage.

The real functionality of parsing the files is included in the function dig. Using functionality provided by Python's built-in xml parser, the function iterates through all XML elements and finds those with no children, or leaf nodes. I possibly could have implemented dig recursively, but the iter function provided by Python's xml parser seemed more intuitive.

I also defined a function for parsing the XML bounds attribute into an element that is decipherable to the Python Pillow package. If this code was used in a more extensive code base, I would probably define a class Rectangle to make the return type of parse_out_points less complicated, but considering this code is only being used in one file and for a single project, I think it's fine to leave as is. The choice to return a list element that held two tuples was purely driven by the fact that the Pillow package takes points in that format when drawing rectangles on images.