I wrote:
> Michael Paquier <michael@paquier.xyz> writes:
>> One part that I have found a bit strange lately about guc.c is that we
>> have mix the core machinery with the SQL-callable parts. What do you
>> think about the addition of a gucfuncs.c in src/backend/utils/adt/ to
>> split things a bit more?
> I might be wrong, but I think the SQL-callable stuff makes use
> of some APIs that are currently private in guc.c. So we'd have
> to expose more API to make that possible. Maybe that wouldn't
> be a bad thing, but it seems to be getting beyond the original
> idea here.
I tried this just to see, and it worked out better than I thought.
The key extra idea is to also pull out the functions implementing
the SET and SHOW commands, because (unsurprisingly) those are just
about in the same place dependency-wise as the SQL functions, and
they have some common subroutines.
I had to export get_config_unit_name(), config_enum_get_options(),
and _ShowOption() (here renamed to ShowGUCOption()) to make this
work. That doesn't seem too awful.
v2 attached does this, without any further relocation of hook
functions as yet. I now see these file sizes:
$ wc guc*c
2629 9372 69467 guc-file.c
6425 22282 176816 guc.c
1048 3005 26962 guc_funcs.c
939 2693 22915 guc_hooks.c
4877 13163 126769 guc_tables.c
15918 50515 422929 total
$ size guc*o
text data bss dec hex filename
13653 4 112 13769 35c9 guc-file.o
46589 0 564 47153 b831 guc.o
8509 0 0 8509 213d guc_funcs.o
6951 0 112 7063 1b97 guc_hooks.o
43570 62998 216 106784 1a120 guc_tables.o
So this removes just about a thousand more lines from guc.c,
which seems worth doing.
regards, tom lane
