On 2011-11-14 15:45, Yeb Havinga wrote:
> On 2011-10-15 07:41, Tom Lane wrote:
>> Yeb Havinga<yebhavinga@gmail.com> writes:
>>> Hello Royce,
>>> Thanks again for testing.
>> I looked this patch over but concluded that it's not ready to apply,
>> mainly because there are too many weird behaviors around error
>> reporting.
>
> Thanks again for the review and comments. Attached is v3 of the patch
> that addresses all of the points made by Tom. In the regression test I
> added a section under --- START ADDITIONAL TESTS that might speedup
> testing.
Please disregard the previous patch: besides that it contained an unused
function, it turned out my statement that all of Tom's points were
addressed was not true - the attached patch fixes the remaining issue of
putting two kinds of errors at the correct start of the current argument
location.
I also put some more comments in the regression test section: mainly to
assist providing testcases for review, not for permanent inclusion.
To address a corner case of the form 'p1 := 1 -- comments\n, p2 := 2' it
was necessary to have read_sql_construct not trim trailing whitespace,
since that results in an expression of the form '1 -- comments, 2' which
is wrong.
regards,
Yeb Havinga