Is it better to collect all test files into one folder, or just let them distribute everywhere?

Question

Now I have a python project, I write my unit-testing code in many of its folders, the tree looks like:

Project
├── module1
│   ├── submodule1
│   │   ├── base.py
│   │   ├── model.py
│   │   └── tests.py
│   ├── conn.py
│   ├── __init__.py
│   └── tests.py
├── errors.py
├── __init__.py
├── router.py
└── tests.py

You can see there are many tests.py files in the projects, at root folder in modules. I use nose to help me do testing, When I want to test all of them, just run nosetests Project at upper folder, everything works fine.

Then comes to my question: I have seen some projects like tornado, they put all test files in one folder so that it is clear to manage them (I guess), is that a better way than what I do currently, if yes, then why ? And according to the philosophy of Python, is there anything wrong in my way of doing test ?

Answer 1

Personnally, I prefer to store my test files into a specific subdirectory test, yet I like to keep test and code very near, so I have test subdirectories every where in my project.

 Project
    ├── module1
    │   ├── submodule1
    │   │   ├── base.py
    │   │   ├── model.py
    │   │   └── test
    │   │       └── submodule1_test.py
    │   ├── conn.py
    │   ├── __init__.py
    │   └── test
    │       └── module1_test.py
    ├── errors.py
    ├── __init__.py
    ├── router.py
    └── test
        └── errors_test.py
        └── router_test.py

Very easy to make a packaging with avoiding all tests directories.