database/postgres
1
0
Fork 0
mirror of https://github.com/postgres/postgres.git synced 2025-06-14 18:42:34 +03:00
Code Activity

Fix issues around strictness of SIMILAR TO.

Browse Source 
As a result of some long-ago quick hacks, the SIMILAR TO operator
and the corresponding flavor of substring() interpreted "ESCAPE NULL"
as selecting the default escape character '\'.  This is both
surprising and not per spec: the standard is clear that these
functions should return NULL for NULL input.

Additionally, because of inconsistency of the strictness markings
of 3-argument substring() and similar_escape(), the planner could not
inline the SQL definition of substring(), resulting in a substantial
performance penalty compared to the underlying POSIX substring()
function.

The simplest fix for this would be to change the strictness marking
of similar_escape(), but if we do that we risk breaking existing views
that depend on that function.  Hence, leave similar_escape() as-is
as a compatibility function, and instead invent a new function
similar_to_escape() that comes in two strict variants.

There are a couple of other behaviors in this area that are also
not per spec, but they are documented and seem generally at least
as sane as the spec's definition, so leave them alone.  But improve
the documentation to describe them fully.

Patch by me; thanks to Álvaro Herrera and Andrew Gierth for review
and discussion.

Discussion: https://postgr.es/m/14047.1557708214@sss.pgh.pa.us
This commit is contained in:
Tom Lane
2019-09-07 14:21:59 -04:00
parent c5bc7050af
commit ca70bdaefe
7 changed files with 193 additions and 41 deletions
Download Patch File Download Diff File Expand all files Collapse all files

85
src/backend/utils/adt/regexp.c
View File

@ -654,15 +654,18 @@ textregexreplace(PG_FUNCTION_ARGS)
}
/*
* similar_escape()
* Convert a SQL:2008 regexp pattern to POSIX style, so it can be used by
* our regexp engine.
* similar_to_escape(), similar_escape()
*
* Convert a SQL "SIMILAR TO" regexp pattern to POSIX style, so it can be
* used by our regexp engine.
*
* similar_escape_internal() is the common workhorse for three SQL-exposed
* functions. esc_text can be passed as NULL to select the default escape
* (which is '\'), or as an empty string to select no escape character.
*/
Datum
similar_escape(PG_FUNCTION_ARGS)
static text *
similar_escape_internal(text *pat_text, text *esc_text)
{
text *pat_text;
text *esc_text;
text *result;
char *p,
*e,
@ -673,13 +676,9 @@ similar_escape(PG_FUNCTION_ARGS)
bool incharclass = false;
int nquotes = 0;
/* This function is not strict, so must test explicitly */
if (PG_ARGISNULL(0))
PG_RETURN_NULL();
pat_text = PG_GETARG_TEXT_PP(0);
p = VARDATA_ANY(pat_text);
plen = VARSIZE_ANY_EXHDR(pat_text);
if (PG_ARGISNULL(1))
if (esc_text == NULL)
{
/* No ESCAPE clause provided; default to backslash as escape */
e = "\\";
@ -687,12 +686,11 @@ similar_escape(PG_FUNCTION_ARGS)
}
else
{
esc_text = PG_GETARG_TEXT_PP(1);
e = VARDATA_ANY(esc_text);
elen = VARSIZE_ANY_EXHDR(esc_text);
if (elen == 0)
e = NULL; /* no escape character */
else
else if (elen > 1)
{
int escape_mblen = pg_mbstrlen_with_len(e, elen);
@ -898,6 +896,65 @@ similar_escape(PG_FUNCTION_ARGS)
SET_VARSIZE(result, r - ((char *) result));
return result;
}
/*
* similar_to_escape(pattern, escape)
*/
Datum
similar_to_escape_2(PG_FUNCTION_ARGS)
{
text *pat_text = PG_GETARG_TEXT_PP(0);
text *esc_text = PG_GETARG_TEXT_PP(1);
text *result;
result = similar_escape_internal(pat_text, esc_text);
PG_RETURN_TEXT_P(result);
}
/*
* similar_to_escape(pattern)
* Inserts a default escape character.
*/
Datum
similar_to_escape_1(PG_FUNCTION_ARGS)
{
text *pat_text = PG_GETARG_TEXT_PP(0);
text *result;
result = similar_escape_internal(pat_text, NULL);
PG_RETURN_TEXT_P(result);
}
/*
* similar_escape(pattern, escape)
*
* Legacy function for compatibility with views stored using the
* pre-v13 expansion of SIMILAR TO. Unlike the above functions, this
* is non-strict, which leads to not-per-spec handling of "ESCAPE NULL".
*/
Datum
similar_escape(PG_FUNCTION_ARGS)
{
text *pat_text;
text *esc_text;
text *result;
/* This function is not strict, so must test explicitly */
if (PG_ARGISNULL(0))
PG_RETURN_NULL();
pat_text = PG_GETARG_TEXT_PP(0);
if (PG_ARGISNULL(1))
esc_text = NULL; /* use default escape character */
else
esc_text = PG_GETARG_TEXT_PP(1);
result = similar_escape_internal(pat_text, esc_text);
PG_RETURN_TEXT_P(result);
}