Abstract

This project is an example of creation of service that accept user file (MS Excel format) and validate is using rules implemented with pytest.

The challenge is that pytest due to its internal architecture is optimized for one time command line launch. But if you want to call it programmatically - it may require from your patience and experimenting. Expecially, if you want to run it several times from a services. And especially, if you want some of your tests being executed for each of the line of the file.

Key findings and takeaways

1. Programmatic call to pytest for collecting the results of execution is implemented with custom plugin intercepting moments of reporting individual test and making test summary.

2. Pass of the "file-to-be-tested" as command line parameter for pytest call may not well handle cases for "row-base" testing due to the "late visibility" of the file name (as part of the fixture). By that time test collection phase will be already completed. So, another custom plugin will be required, which will intercept process of "configuring" of the test run.

3. pytest should be run in isolation (separate process) to handle dynamically passed data-to-test. While passing results of the test execution one should be careful to ensure that all results are serializable ("pickle-able").

Detail description of the approach, discovered issues and other reference information is described in the habr.com article (RU)

Project structure and dependencies

Folders

Folder	Description
/app	The code of t he service and Docker files is kept under folder.
/app/<type>	Tests for 2 sample file types are stored under subfolders named by file types.
/examples	Samples MS Excel files which are used for corresponding tests.

Application structure

File	Description
common_types.py	Declaration of the types, used in service and validation process.
main.py	Main file of the service, implemented with FastAPI
validator.py	Module handling pytest execution for given file and type

Dependencies

Project relies on several well-known libraries.

Web service implementation:

• fastapi
• python-multipart
• pydantic
• uvicorn

Testing

• pytest
• pytest-dependency

Processing MS Excel files

• openpyxl
• pandas