RFC: change default for logging to +nowrap

[hvr]

I'm curious what others think about this, hence I'm putting this up for discussion. And being biased, I'm also specifically interested in hearing about reasons why +nowrap is undesirable.

By default, cabal line-wraps its logging output to hardcoded line-length of 79 characters via wrapText:

cabal/Cabal/Distribution/Utils/Generic.hs

Lines 108 to 131 in 3b39059

	-- | Wraps text to the default line width. Existing newlines are preserved.
	wrapText :: String -> String
	wrapText = unlines
	. map (intercalate "\n"
	. map unwords
	. wrapLine 79
	. words)
	. lines
	-- | Wraps a list of words to a list of lines of words of a particular width.
	wrapLine :: Int -> [String] -> [[String]]
	wrapLine width = wrap 0 []
	where wrap :: Int -> [String] -> [String] -> [[String]]
	wrap 0 [] (w:ws)
	| length w + 1 > width
	= wrap (length w) [w] ws
	wrap col line (w:ws)
	| col + length w + 1 > width
	= reverse line : wrap 0 [] (w:ws)
	wrap col line (w:ws)
	= let col' = col + length w + 1
	in wrap col' (w:line) ws
	wrap _ [] [] = []
	wrap _ line [] = [reverse line]

In #4091, a verbosity level modified +nowrap was introduced which suppresses this line-wrapping;

IMO, having cabal pre-linewrap its output interacts badly with terminals or editors if these re-wrap dynamically when resizing the window, readability is IMHO not improved by line-wrapping, and it makes it harder to copy and paste as the terminal can't easily reconstruct the original log message's line endings, and thus the pre-wrapped text is copy-n-pasted into the target including the line-wrapping (which may again not be adequate for the target text-body width).

As an example, this here is a solver failure pre-wrapped via wrapText

$ cabal new-repl -w ghc-8.6.1 --build-depends 'lens'
Resolving dependencies...
cabal: Could not resolve dependencies:
[__0] trying: fake-package-0 (user goal)
[__1] trying: lens-4.17 (dependency of fake-package)
[__2] trying: transformers-0.5.5.0/installed-0.5... (dependency of lens)
[__3] next goal: hashable (dependency of lens)
[__3] rejecting: hashable-1.2.7.0 (conflict: transformers =>
base==4.12.0.0/installed-4.1..., hashable => base>=4.4 && <4.12)
[__3] rejecting: hashable-1.2.6.1 (conflict: transformers =>
base==4.12.0.0/installed-4.1..., hashable => base>=4.4 && <4.11)
[__3] rejecting: hashable-1.2.6.0 (conflict: transformers =>
base==4.12.0.0/installed-4.1..., hashable => base>=4.4 && <4.10)
[__3] rejecting: hashable-1.2.5.0, hashable-1.2.4.0 (conflict: transformers =>
base==4.12.0.0/installed-4.1..., hashable => base>=4.0 && <4.10)
[__3] rejecting: hashable-1.2.3.3, hashable-1.2.3.2 (conflict: transformers =>
base==4.12.0.0/installed-4.1..., hashable => base>=4.0 && <4.9)
[__3] rejecting: hashable-1.2.3.1, hashable-1.2.3.0, hashable-1.2.2.0,
hashable-1.2.1.0, hashable-1.2.0.10, hashable-1.2.0.9, hashable-1.2.0.8,
hashable-1.2.0.7, hashable-1.2.0.6, hashable-1.2.0.5, hashable-1.2.0.4,
hashable-1.2.0.3, hashable-1.2.0.2, hashable-1.2.0.1, hashable-1.2.0.0,
hashable-1.1.2.5, hashable-1.1.2.4, hashable-1.1.2.3 (conflict: transformers
=> base==4.12.0.0/installed-4.1..., hashable => base>=4.0 && <4.8)
[__3] rejecting: hashable-1.1.2.2, hashable-1.1.2.1, hashable-1.1.2.0,
hashable-1.1.1.0, hashable-1.1.0.0, hashable-1.0.1.1, hashable-1.0.1.0,
hashable-1.0.0 (conflict: lens => hashable>=1.1.2.3 && <1.3)
[__3] fail (backjumping, conflict set: hashable, lens, transformers)
After searching the rest of the dependency tree exhaustively, these were the
goals I've had most trouble fulfilling: hashable, lens, base, transformers,
fake-package

...and the same with +nowrap:

$ cabal new-repl -w ghc-8.6.1 --build-depends 'lens' -vnormal+nowrap
Resolving dependencies...
cabal: Could not resolve dependencies:
[__0] trying: fake-package-0 (user goal)
[__1] trying: lens-4.17 (dependency of fake-package)
[__2] trying: transformers-0.5.5.0/installed-0.5... (dependency of lens)
[__3] next goal: hashable (dependency of lens)
[__3] rejecting: hashable-1.2.7.0 (conflict: transformers => base==4.12.0.0/installed-4.1..., hashable => base>=4.4 && <4.12)
[__3] rejecting: hashable-1.2.6.1 (conflict: transformers => base==4.12.0.0/installed-4.1..., hashable => base>=4.4 && <4.11)
[__3] rejecting: hashable-1.2.6.0 (conflict: transformers => base==4.12.0.0/installed-4.1..., hashable => base>=4.4 && <4.10)
[__3] rejecting: hashable-1.2.5.0, hashable-1.2.4.0 (conflict: transformers => base==4.12.0.0/installed-4.1..., hashable => base>=4.0 && <4.10)
[__3] rejecting: hashable-1.2.3.3, hashable-1.2.3.2 (conflict: transformers => base==4.12.0.0/installed-4.1..., hashable => base>=4.0 && <4.9)
[__3] rejecting: hashable-1.2.3.1, hashable-1.2.3.0, hashable-1.2.2.0, hashable-1.2.1.0, hashable-1.2.0.10, hashable-1.2.0.9, hashable-1.2.0.8, hashable-1.2.0.7, hashable-1.2.0.6, hashable-1.2.0.5, hashable-1.2.0.4, hashable-1.2.0.3, hashable-1.2.0.2, hashable-1.2.0.1, hashable-1.2.0.0, hashable-1.1.2.5, hashable-1.1.2.4, hashable-1.1.2.3 (conflict: transformers => base==4.12.0.0/installed-4.1..., hashable => base>=4.0 && <4.8)
[__3] rejecting: hashable-1.1.2.2, hashable-1.1.2.1, hashable-1.1.2.0, hashable-1.1.1.0, hashable-1.1.0.0, hashable-1.0.1.1, hashable-1.0.1.0, hashable-1.0.0 (conflict: lens => hashable>=1.1.2.3 && <1.3)
[__3] fail (backjumping, conflict set: hashable, lens, transformers)
After searching the rest of the dependency tree exhaustively, these were the goals I've had most trouble fulfilling: hashable, lens, base, transformers, fake-package

Personally, I consider the +nowrap version easier to parse visually, as it retains more of its original structure. I don't have to cognitively unwrap long lines into their original form to unpack the information.

...opinions?

[23Skidoo]

I don't care that much about this (especially since the default can be overturned), but I slightly prefer wrapped output.

[phadej]

On my laptop

% echo $COLUMNS
198

So current default looks silly, I'm 👍 for +nowrap, we should let terminal emulators do their job

[chshersh]

@hvr I agree with +nowrap for logging. The last example looks more readable!

I often see that this solver error confuses beginners. Maybe we can go even further and improve UX of this particular error message? Like, use some fancy ascii symbols. Something like this:

fake-package-0
└── ❌ lens-4.17
    ├── ✅ transformers-0.5.5.0
    └── ❌ hashable
        ├── ❌ hashable-1.2.7.0
        │   ├── transformers-0.5.5.0 => base==4.12.0.0
        │   └── hashable-1.2.7.0     => base>=4.4 && <4.12
        ├── ❌ hashable-1.2.6.1
        │   ├── transformers-0.5.5.0 => base==4.12.0.0
        │   └── hashable-1.2.6.1     => base>=4.4 && <4.11
        ...

Probably, this format should be discussed more. It's just an idea.

[hvr]

@chshersh could you add your comment to #2858 (which concerns itself with improving the solver error reporting)?