Overview

According to Python docs, newlines can be one of:

• None
• string (if line endings are consistent)
• Tuple of strings (if multiple line endings are used)

Meanwhile, open builtin takes None or a string representing the new line according to Python docs. Hence, there is a mismatch between the two for the case of multiple line endings. cookiecutter will get something like newlines=("\n", "\r\n") and try to pass that to open function. In turn, this will throw an error that looks like this:

TypeError: open() argument 'newline' must be str or None, not tuple

This PR proposes that cookiecutter takes the first detected new line when it encounters template files with mixed line endings. Whilst this means that original line endings aren't completed preserved, this seems like a better user experience than throwing a TypeError. It should also make cookiecutter more resilient against (arguably) malformed inputs.

Fixes #1737.

Testing

• Added a test test_generate_file_handles_mixed_line_endings to reproduce the issue, it only passes with the proposed patch ✅
• Existing tests pass locally too ✅
• Confirmed that proposed changes are covered via the coverage report ✅

Notes

• New test_generate_file_handles_mixed_line_endings test asserts that either CRLF or LF newlines are returned, as Python docs seem to make no guarantees on what order line endings are returned in case of file having mixed newlines.