Re: WIP: a way forward on bootstrap data

> On May 6, 2018, at 12:08 PM, John Naylor <jcnaylor@gmail.com> wrote:
> 
> On 5/7/18, Mark Dilger <hornschnorter@gmail.com> wrote:
>> Hackers,
>> 
>> Have you already considered and rejected the idea of having
>> genbki.pl/Catalog.pm define constants that can be used in
>> the catalog .dat files?  I'm mostly curious if people think
>> the resulting .dat files are better or worse using constants
>> of this sort.  For example:
> ...
>> +               # pg_cast constants for castcontext
>> +               use constant IMPLICIT => 'i';
>> +               use constant ASSIGNMENT => 'a';
>> +               use constant EXPLICIT => 'e';
> 
> The comment refers to pg_cast, but these constants apply globally.
> It's also not the right place from a maintainability perspective, and
> if it was, now these values have different macros defined in two
> places. This is not good.
> 
>> -  castcontext => 'a', castmethod => 'f' },
>> +  castcontext => ASSIGNMENT, castmethod => FUNCTION },
> 
> For one, this breaks convention that the values are always
> single-quoted. If you had a use case for something like this, I would
> instead use the existing lookup infrastructure and teach genbki.pl to
> parse the enums (or #defines as the case might be) in the relevant
> header file. You'd need some improvement in readability to justify
> that additional code, though. I don't think this example quite passes
> (it's pretty obvious locally what the letters refer to), but others
> may feel differently.


John, Tom, thanks for the feedback.  I share your concerns about my
straw-man proposal. But....

In the catalogs, 'f' usually means 'false', not 'function'.  A person
reading pg_cast.dat could see:

  castmethod => 'f'

and think that meant a binary conversion, since castmethod is false.
That's almost exactly wrong.  Hence my desire to write

  castmethod => FUNCTION

I don't have any better proposal, though.

mark