Adjust string comparison in jsonpath
authorAlexander Korotkov <akorotkov@postgresql.org>
Sun, 11 Aug 2019 19:54:53 +0000 (22:54 +0300)
committerAlexander Korotkov <akorotkov@postgresql.org>
Sun, 11 Aug 2019 19:54:53 +0000 (22:54 +0300)
commitd54ceb9e176152f930e60709e07c636e8e5414f5
tree96178702c53c066b11f4a56733a0b4c3b452778e
parentcabe0f298ea7efade11d8171c617e668934d0d09
Adjust string comparison in jsonpath

We have implemented jsonpath string comparison using default database locale.
However, standard requires us to compare Unicode codepoints.  This commit
implements that, but for performance reasons we still use per-byte comparison
for "==" operator.  Thus, for consistency other comparison operators do per-byte
comparison if Unicode codepoints appear to be equal.

In some edge cases, when same Unicode codepoints have different binary
representations in database encoding, we diverge standard to achieve better
performance of "==" operator.  In future to implement strict standard
conformance, we can do normalization of input JSON strings.

Original patch was written by Nikita Glukhov, rewritten by me.

Reported-by: Markus Winand
Discussion: https://postgr.es/m/8B7FA3B4-328D-43D7-95A8-37B8891B8C78%40winand.at
Author: Nikita Glukhov, Alexander Korotkov
Backpatch-through: 12
src/backend/utils/adt/jsonpath_exec.c
src/test/regress/expected/jsonb_jsonpath.out
src/test/regress/sql/jsonb_jsonpath.sql