Unexpected line length check behaviour when line has tabs.

[Nuno-Mota]

Hi.

It seems like the counting of line length with tabs has an unexpected behaviour.

From what I can infer, the length of each line is first counted considering that tabs have a size of 1.
If the resulting length is equal to or greater than the line-length setting, then the size is recomputed taking into account the tab-size setting.
Furthermore, the line needs to have at least 1 space (as discussed in #6424).

Is this the intended behaviour? If so, is there some place that explains the reasoning behind it?

This was tested for ruff 0.0.282.

MWE:

pyproject.toml

[tool.ruff]
line-length = 6
tab-size = 4
show-source = true

test file (note that github uses a default tab size of 8)

# aaaa
# aaaaa
    # a
	# a
	# aa
	# aaa
	# aaaa
		# a
		# aa
		# aaa

if True: # noqa: E501
	[12]
	[12 ]
	[1,2]
	[1, 2]

ruff's output

ruff test3.py 
test3.py:2:7: E501 Line too long (7 > 6 characters)
  |
1 | # aaaa
2 | # aaaaa
  |       ^ E501
3 |     # a
4 |     # a
  |

test3.py:3:7: E501 Line too long (7 > 6 characters)
  |
1 | # aaaa
2 | # aaaaa
3 |     # a
  |       ^ E501
4 |     # a
5 |     # aa
  |

test3.py:6:4: E501 Line too long (9 > 6 characters)
  |
4 |     # a
5 |     # aa
6 |     # aaa
  |       ^^^ E501
7 |     # aaaa
8 |         # a
  |

test3.py:7:4: E501 Line too long (10 > 6 characters)
  |
5 |     # aa
6 |     # aaa
7 |     # aaaa
  |       ^^^^ E501
8 |         # a
9 |         # aa
  |

test3.py:9:3: E501 Line too long (12 > 6 characters)
   |
 7 |     # aaaa
 8 |         # a
 9 |         # aa
   |         ^^^^ E501
10 |         # aaa
   |

test3.py:10:3: E501 Line too long (13 > 6 characters)
   |
 8 |         # a
 9 |         # aa
10 |         # aaa
   |         ^^^^^ E501
11 | 
12 | if True: # noqa: E501
   |

test3.py:14:4: E501 Line too long (9 > 6 characters)
   |
12 | if True: # noqa: E501
13 |     [12]
14 |     [12 ]
   |       ^^^ E501
15 |     [1,2]
16 |     [1, 2]
   |

test3.py:16:4: E501 Line too long (10 > 6 characters)
   |
14 |     [12 ]
15 |     [1,2]
16 |     [1, 2]
   |       ^^^^ E501
   |

Found 8 errors.
✗ - status code 1

[charliermarsh]

Do you mind clarifying what behavior here is unexpected? "...then the size is recomputed taking into account..." sounds like an implementation detail as-described, and not a behavior per se.

[Nuno-Mota]

Do you mind clarifying what behavior here is unexpected?

Of course.

The issue is that tab-indented lines that are longer than the line-length setting (taking into consideration the tab-size setting), might not be flagged for E501.

"...then the size is recomputed taking into account..." sounds like an implementation detail as-described, and not a behavior per se.

That might be. This is simply the behaviour I was able to infer.
In any case it seems to lead to an erroneous line length count in certain cases.

[charliermarsh]

Ah, I think that's a result of the performance optimization I proposed in https://github.com/astral-sh/ruff/pull/5811/files, which doesn't hold true for tabs (for which a single-byte character can be wider than one byte).