Michael Fuhr <mike@fuhr.org> writes:
> On Mon, Aug 15, 2005 at 08:21:23PM -0400, Tom Lane wrote:
>> Given that we consider trailing spaces in char(n) to be semantically
>> insignificant, would it make sense to strip them before doing the
>> regex pattern match?
> How standards-compliant would that be? Does the standard specify
> what should happen when using SIMILAR TO with a char(n) value?
Hmm ... suddenly I'm getting a strong sense of deja vu ... think we've
been around this merry-go-round before. SQL99 says
ii) The <predicate>
MC LIKE PC
is true if there exists a partitioning of MCV into substrings such that:
1) A substring of MCV is a sequence of 0 (zero) or more contiguous <character
representation>sof MCV and each <character representation> of MCV is part of exactly one
substring.
2) If the i-th substring specifier of PCV is an arbitrary character specifier, the
i-thsubstring of MCV is any single <character representation>.
3) If the i-th substring specifier of PCV is an arbitrary string specifier, then the
i-thsubstring of MCV is any sequence of 0 (zero) or more <character
representation>s.
4) If the i-th substring specifier of PCV is neither an arbitrary character specifier
noran arbitrary string specifier, then the i-th substring of MCV is equal to that
substringspecifier according to the collating sequence of the <like predicate>, without the appending
of <space> characters to MCV, and has the same length as that substring specifier.
5) The number of substrings of MCV is equal to the number of substring specifiers of
PCV.
Rule ii.4 says that you use the collating sequence associated with the
data values, which is where the SQL spec keeps its space sensitivity
information --- but the restrictions about not adding space characters
and having the same length seem to be intended to prevent use of
pad-space-insensitivity to create a match.
I think we read this text before, came to the same conclusion, and
put in the special operator to make it behave that way. So ...
never mind.
regards, tom lane
