Disable new help display of bool values?

[johnsonjh]

Will there be a way to disable the true/false display for Booleans in help?

This was added unconditionally in 666b603

This breaks my formatting (which isn’t too big of a deal) but it makes the useful area for text much less because I ensure that my text output does not exceed 80 columns.

Sorry for my screenshots here - the gray line is column 80 in both screenshots.

Release

Master

I have quite a few Go projects and this change negatively impacts the way the default help display looks on almost all of them. I feel it is a readability downgrade too as it makes quickly visually scanning the option list more difficult.

[johnsonjh]

I have this workaround now which involves intercepting the output and modifying it, so something like this added in every project:

 var buf bytes.Buffer
 pflag.CommandLine.SetOutput(&buf)
 pflag.PrintDefaults()
 pflag.CommandLine.SetOutput(os.Stdout)
 re := regexp.MustCompile(`\[=true\|false\]`)
 output := re.ReplaceAllString(buf.String(), "             ")
 fmt.Fprint(os.Stdout, output)

… which is gross but it works. This also doesn’t solve the problem of adding the extra whitespace which means there is less actually relevant text on each help line.

---

Please don’t see this as too big of a complaint - I just feel that’s it’s an annoyance because it potentially breaks the help formatting of existing and working code and such things should not be done in small point releases (if we consider the way the output looks as part of the semantic versioning ‘promise’), but I do understand the maintainers here may not share this belief. My workaround is also straightforward but still needs to be added everywhere pflag will be updated.

Because I’ve been majorly burned before (not pflag) by “safe” dependency updates, I have tests that I run every time I update any dependency and I periodically test all my projects with master dependencies, so I don’t get surprised by breakage.

[tomasaschan]

This change was introduced to address one of our more long-standing (and upvoted) issues around how syntax for passing values is different for bool-valued flags versus other types, which has lead to real issues and bugs (reported as comments on the issue). See #410 for more details and discussion history.

I would argue that a) we're in a better place now than before this change, and b) the help output is not part of what we promise to not change as part of semver. It follows that the issue you raise here is not something we should address with further changes.

I'm going to leave this open for a while to allow space for others to disagree with me, but unless I hear good arguments to the contrary, I'll eventually close this issue without further action.

[johnsonjh]

I can respect that, and while I do wish it was configurable, my workaround is acceptable.