`tokenize.untokenize()` does not preserve the whitespace of the original source

[kelly-eclipse]

Bug report

Bug description:

tokenize.untokenize() replaces the whitespace between tokens and after indents with spaces:

>>> import tokenize
>>> code = b"if False:\n\tprint(\n\t\t'hello')"
>>> readline = iter(code.splitlines(keepends=True)).__next__
>>> tokenize.untokenize(tokenize.tokenize(readline))
b"if False:\n\tprint(\n\t 'hello')"

Note the second tab before 'hello' becomes a space.

The culprit is Untokenizer.add_whitespace which is called from the loop inside Untokenizer.untokenize whenever a 5-tuple iterator is used. The fifth item of the tuple is the original line; therefore, the loop could pass it to add_whitespace to reconstruct the original whitespace.

CPython versions tested on:

3.10, CPython main branch

Operating systems tested on:

Windows

Linked PRs

• gh-156072: Fix tokenize.untokenize to preserve the original line's whitespace #156073