fix and rename real_len

[tgolsson]

Fixes #606 -- see added test cases. This logic matches what happens draw_target.rs as well. It'll now also correctly handle lines with only ANSI codes.

indicatif/src/draw_target.rs

Lines 505 to 518 in 81aa4c6

	let diff = if line.is_empty() {
	// Empty line are new line
	1
	} else {
	// Calculate real length based on terminal width
	// This take in account linewrap from terminal
	let terminal_len = (line_width as f64 / term_width as f64).ceil() as usize;
	// If the line is effectively empty (for example when it consists
	// solely of ANSI color code sequences, count it the same as a
	// new line. If the line is measured to be len = 0, we will
	// subtract with overflow later.
	usize::max(terminal_len, 1)
	};

[djc]

Thanks for working on this!

Can we share the code with draw_target somehow, so that it is more obviously using the same algorithm?

If we're going to move this I'd like to see two separate commits, one that moves/renames the code and one that changes the algorithm. Ideally tests that work with the old algorithm are introduced before this, and then the fixing commit adds tests that didn't work previously -- this will make it easy for me to review. (Please squash linting changes etc.)

[tgolsson]

Can we share the code with draw_target somehow, so that it is more obviously using the same algorithm?

Doesn't seem like a worthwhile change seeing as that is counting lines while drawing, while this does it ahead of time.

If we're going to move this I'd like to see two separate commits, one that moves/renames the code and one that changes the algorithm. Ideally tests that work with the old algorithm are introduced before this, and then the fixing commit adds tests that didn't work previously -- this will make it easy for me to review. (Please squash linting changes etc.)

Please see new commits -- first commit moves + add working tests, second commit adds the failing tests, third commit fixes the failing tests.

[djc]

Thanks, that is better. A few nits to address...

[tgolsson]

CI seems like a flake, can't repro at all locally.

[djc]

Yup, the CI failures seemed to go away on rerunning. This is looking great, thanks!