Re: [PATCH 0/4] Replace --defines with --header, and introduce --html

[Akim Demaille]

Hi Kaz,

Sorry for the delays on this message, and your others.  Thanks
for your answer.

> Le 15 sept. 2020 à 10:51, Kaz Kylheku <kaz@kylheku.com> a écrit :
> 
> On 2020-09-12 08:38, Akim Demaille wrote:
>> I'd be happy to read comments about this branch.  Too bad --header and
>> --html both start with H, there can only be one.  Ideas for a short
>> option for --html are most welcome.
> 
> "Defines", more often than not, does refer to "preprocessor pound defines",
> but not always. It's an informal word that broadly covers interface
> definitions of all sorts.
> 
> If we search for the phrase "file contains defines", we can find
> citations for that refering to other than preprocessor definitions.\
> 
> For instance:
> 
> https://www.cse.unr.edu/~miles/compiler/docs/html/ast_8h-source.html
> 
> A comment says "file contains defines of the various types of
> astnodes and the fromASTFile function"; these are actually C++
> class declarations.

This sounds really weird to me, "definitions" sound much more appropriate.
I could not find any dictionary supporting "define" as a noun.


> Anyway, the bike-shedding debate between "header" and "defines"
> is already settled by the fact that the option already exists
> under one of those two names.

Well, "defines" is inappropriate, and the fact that the option exists
does not legitimate that we keep it mysterious to newcomers.  I believe
--header/%header to be vastly clearer that --defines/%defines, and with
less possible confusion with --define/%define.

History mandates backward compatibility, which does not prevent fixes.

> "defines" isn't the greatest name, but it's not terrible. Terrible
> is, e.g., calling a function "creat" instead of "create".

:) :) :)

> It pairs with -d, which comes from POSIX. Long and short options
> don't have to share the same initial letter, but it's somewhat nice
> if they do, and from your remark about "header" and "html", it seems
> that you agree.

Unfortunately -d is not --defines, because -d does not accept an
argument, as per POSIX.  That's why -H is not -d.


> Adding a synonym and deprecating the old one creates unnecessary
> churn that doesn't correspond to any change in functionality or
> issue being solved. Which is why we still have "creat".

Well, in the present case, I was also aiming at converging to the
choice made in byacc, -H.

So I'm installing these changes.

Cheers!