Implement a Batch Image Processor
Company: Anthropic
Role: Software Engineer
Category: Coding & Algorithms
Difficulty: medium
Interview Round: Technical Screen
Implement a program that applies basic image transformations to a batch of images.
You are given a list of input image paths and, for each image, a sequence of operations to apply in order. Support the following operations:
1. **grayscale**: convert the image to grayscale.
2. **scale(factor)**: multiply both width and height by a positive floating-point scale factor.
3. **resize(width, height)**: resize the image to the exact target dimensions.
### Part 1
Write a function that processes a small batch of images correctly and saves the transformed results.
Example function shape:
`process_images(tasks: list[ImageTask], output_dir: str) -> list[str]`
Where each `ImageTask` contains:
- `input_path`: path to the source image
- `operations`: ordered list of operations such as `grayscale`, `scale(0.5)`, or `resize(200, 300)`
The function should return the output paths in the same order as the input tasks.
### Part 2
Now assume the images are very large and the batch size is also large. The interviewer asks you to improve performance on a multi-core machine.
Update your approach so that:
- multiple images can be processed in parallel,
- output ordering is preserved,
- a failure in one task does not corrupt other results,
- the design is appropriate for CPU-heavy image processing.
You may use a standard Python imaging library such as Pillow.
Quick Answer: This question evaluates a candidate's competency in image processing operations and parallel/concurrent programming, covering basic transformations (grayscale, scale, resize) and considerations for CPU-bound batch processing.
Part 1: Process a Small Batch of Image Tasks
You are given a small batch of image-processing tasks. Because an online judge cannot read or write real image files, each image is represented by metadata: its path, width, height, color mode, and an ordered list of operations. Apply the operations in order and return the metadata that would be saved for each output image. Supported operations are: ('grayscale',) to change the mode to 'L'; ('scale', factor) to multiply width and height by factor, using floor and then clamping each dimension to at least 1; and ('resize', new_width, new_height) to set exact dimensions. The output filename is built from the input filename by inserting '_out' before the extension. If there is no extension, just append '_out'. Return results in the same order as the input tasks.
Constraints
- 0 <= len(tasks) <= 10^4
- 1 <= width, height <= 10^6
- Each task has 0 to 50 operations
- In this part, all operations are valid: scale factors are positive and resize dimensions are positive integers