On Fri, Aug 18, 2017 at 1:56 AM, Peter Eisentraut
<peter.eisentraut@2ndquadrant.com> wrote:
> Here are a few assorted patches I made while working on the stdbool set,
> cleaning up various pieces of dead code and weird styles.
>
> - Drop excessive dereferencing of function pointers
- (*next_ProcessUtility_hook) (pstmt, queryString,
+ next_ProcessUtility_hook(pstmt, queryString, context, params,
queryEnv, dest, completionTag);
But this... Personally I like the current grammar which allows one to
make the difference between a function call with something declared
locally and something that may be going to a custom code path. So I
think that you had better not update the system hooks that external
modules can use via shared_preload_libraries.
> - Remove endof macro
Its last use is 1aa58d3a from 2009.
> - Remove unnecessary casts
Those are also quite old things:
src/include/access/attnum.h: ((bool) ((attributeNumber) !=
InvalidAttrNumber))
src/include/access/attnum.h: ((bool) ((attributeNumber) > 0))
src/backend/utils/hash/dynahash.c: *foundPtr = (bool) (currBucket != NULL);
[... etc ...]
> - Remove unnecessary parentheses in return statements
So you would still keep parenthesis like here for simple expressions:
contrib/bloom/blutils.c: return (x - 1);
No objections.
Here are some more:
contrib/intarray/_int_bool.c: return (calcnot) ?
contrib/ltree/ltxtquery_op.c: return (calcnot) ?
And there are many "(0)" "S_ANYTHING" in src/interfaces/ecpg/test/ and
src/interfaces/ecpg/preproc/.
src/port/ stuff is better left off, good you did not touch it.
> - Remove our own definition of NULL
Fine. c.h uses once NULL before enforcing its definition.
> - fuzzystrmatch: Remove dead code
Those are remnants of a323ede, which missed to removed everything.
Looks good to me.
--
Michael