Re: range-diff: slight usability problem with mistyped ref

[Johannes Schindelin <Johannes.Schindelin@gmx.de>]

[-- Attachment #1: Type: text/plain, Size: 1851 bytes --]

Hi Kristoffer,

On Mon, 3 Apr 2023, Kristoffer Haugsbakk wrote:

> On Tue, Mar 28, 2023, at 13:37, Johannes Schindelin wrote:
> > So even making the exact error message depend on the number of
> > arguments could result in misleading error message.
>
> Yeah, I see. It seems that the variadic nature of the command makes it
> difficult to guess what the user might have meant in all cases.
>
> > Do you have any splendid idea how to phrase the error message (or
> > adapt it to the concrete invocation)?
>
> No. I was going to look more closely at that if-else chain, but given
> the fact that I don’t know C and all the cases that would need to be
> covered (“what if the third argument happens to match both a ref and a
> file?”) I wouldn’t get anywhere.

Knowledge of C is not required to come up with an error message ;-)

The forms in which `git range-diff` can be called are:

	git range-diff A...B [[--] <path>...]
	git range-diff A..B C..D [[--] <path>...]
	git range-diff A B C [[--] <path>...]

In general, I find it hard to guess which form was meant when a typo
causes an error message. For example, `git range-diff A..B C` might have
been intended to be the three-rev form `A B C`, but it could also have
been intended to pass two commit ranges `A..B HEAD..C`.

As such, I wonder whether we can just cleverly _avoid_ saying _which_ form
we suspect the user to have intended.

In other words, instead of

	need two commit ranges

we could say:

	`range-diff` needs three revs, two ranges or a symmetric range

Since this does not require any C expertise, it would probably be a good
idea to let C experts work on other patches, patches that do require their
expertise: The patch to improve the error message does not require any
knowledge of the C language at all.

Ciao,
Johannes