Align on a style in messages -- to end with '.', or not?

[MichaelChirico]

since "most" error messages dont end with ., but NIT

Originally posted by @ben-schwen in #7562 (comment)

[MichaelChirico]

Not worth spending too much time here. A first step is to quantify how often messages end in . vs. not; in parallel, we can check what the broader community does (in R, and in other languages).

If we do decide to make a pass at aligning here, it should be sure to include a linter that can prevent backsliding as it's sure to keep cropping up.

[ashish-um]

Hi @MichaelChirico, I'd like to help with this.

I wrote a script to scan the R/ directory and quantify the current state of error/warning messages. Here is the breakdown:

Style	Count	Percentage
No Dot (e.g., stop("Column not found"))	303	~64%
With Dot (e.g., stop("Column not found."))	168	~36%

Even though "No Dot" is currently more common, I propose we standardize on keeping the trailing dot.

Reasoning:

• Multi-sentence Consistency: We have many complex error messages (e.g., "RHS has length %d. Expecting length 2."). If we drop the final dot, these messages become inconsistent ("Sentence one. Sentence two"), which looks awkward.

• Grammar: Most messages are complete sentences (Subject-Verb-Object).

• Translation: Standard punctuation usually aids the translation process (Issue Unify messages for translation #7562).

I am willing to put in the extra effort to update the ~300 messages that currently lack a dot to unify the style.

What do you think?

[jangorecki]

Ending dot is not very relevant. If editing error message already, then we could look at it and add, otherwise it is just noice. We are making code to crush data not to shine error messages.

[ashish-um]

Understood. I'll hold off on a mass-cleanup PR to avoid the noise. I'll keep the style guide in mind if I'm touching those lines for other functional reasons. Thanks for the clarification!