Re: [PATCH] Use Boyer-Moore-Horspool for simple LIKE contains patterns - Mailing list pgsql-hackers

Hi Greg,

Thanks for the careful review.  I have attached a v2 patch.

> git grep shows we already use BMH in src/backend/utils/adt/varlena.c
> Worth acknowledging that in a code comment somewhere? I didn't see any
> obvious advantage to refactoring things out at quick glance, but a mention
> might be nice.

Agreed.  I added a comment at the top of like_bmh.c that cross-references the
existing Boyer-Moore-Horspool implementation in varlena.c and explains why I
kept the implementations separate.  The varlena.c code searches one
(haystack, needle) pair with an adaptively sized skip table, whereas the LIKE
path interprets its internal backslash escapes while extracting the literal
and caches the prepared search state in FmgrInfo for use across rows.  I did
not find a clean way to share that machinery without introducing more coupling
than seemed useful.

> + * by '%' wildcards.  Remove backslash escapes while building the search
> + * state.
>
> Slightly off comment. This is for like_bmh_pattern_is_eligible - we are not
> removing here, just skipping things when we count.

Right.  I reworded the comment to say that the eligibility check skips
backslash escapes while counting the literal length.  The escapes are removed
later, when the search state is built.

> if (i + 1 >= plen - 1)
>
> Worth a comment to explain that we are catching the '%foo\%' case here.

Added.  The new comment explains that this rejects patterns such as
'%foo\%', where the backslash escapes the closing '%' rather than a literal
byte.

> pattern_stable = get_fn_expr_arg_stable(flinfo, 1);
>
> /*
>  * ScalarArrayOpExpr invokes the operator once per array element.  The
>  * array expression can be stable while the pattern passed to this function
>  * changes between calls, so it must not use a cached search state.
>  */
> if (flinfo->fn_expr != NULL && IsA(flinfo->fn_expr, ScalarArrayOpExpr))
>     pattern_stable = false;
>
> My first thought was to make this an if/else so we don't reclobber, but
> seeing how later on we check collation every time, I'm wondering if we
> shouldn't just check the pattern as well every time via a memcmp like
> regexp.c does in RE_compile_and_cache (and remove that block above).
> So we store it verbatim in the like_bmh_init() function with memcpy, then
> make the check inside like_bmh_match() that looks like this:
>
> unlikely(collation has changed)
>
> into:
>
> unlikely(
>   collation has changed
>   OR pattern length has changed
>   OR pattern itself has changed (e.g. memcmp true)
> )
>
> Also means you could then roll get_fn_expr_arg_stable into that big old ||
> grouping, and remove pattern_stable entirely.

I implemented the suggested verbatim-pattern cache and benchmarked it directly
against the initial patch's structural-stability design.  The test scanned two
million rows per transaction, with a warmup followed by the median of seven
pgbench runs of 40 transactions each.  The benchmark used an AMD EPYC 7763
host with 8 vCPUs, GCC 11.4.0, and an -O2 -g build, using a UTF-8 database
with C locale.  The results below are median latency per scan:

case                                  initial patch  memcmp    vs. initial
------------------------------------  -------------  --------  -----------
constant, 4-byte literal                    63.7 ms   65.1 ms        +2.3%
constant, 32-byte literal                   48.8 ms   48.4 ms        -0.8%
constant, 4-byte literal, 8-byte input      43.2 ms   44.0 ms        +1.8%
non-constant, fixed value at runtime        92.9 ms   57.0 ms       -38.6%
non-constant, changes on every row          91.6 ms  172.4 ms       +88.2%

The per-row length check and memcmp were therefore not particularly expensive
for stable constant patterns.  The more important tradeoff involved
non-constant patterns.  When the value remained fixed at runtime, the verbatim
cache was faster because it could use BMH.  When the pattern changed on every
row, however, it was substantially slower than the initial patch, which sends
that case to the existing generic matcher.  The verbatim variant had to repeat
the eligibility check and rebuild the 256-entry skip table for every row.

I then tested a hybrid of the two approaches.  Patterns that
get_fn_expr_arg_stable() identifies as a Const or external Param keep the
existing comparison-free search state.  An eligible non-stable pattern stores
its verbatim bytes and is revalidated with a length check and memcmp.  On the
first mismatch, the state is changed permanently to the generic marker.  The
mismatching row and all later rows use the existing matcher; the eligibility
check and skip-table build are never repeated.

ScalarArrayOpExpr still has to be classified as non-stable, since its array
expression can be a Const while the operator receives a different element on
each call.  It now uses the same revalidation path and falls back permanently
if the elements differ.

I reran the comparison on aarch64 using two clean build trees based on the
same source revision and configured with the same options.  Both servers used
the same data directory.  The table contained two million 32-byte strings, a
fixed pattern column, and an alternating pattern column.  Parallel query was
disabled, each server was warmed before measurement, and the server order was
alternated in ABBA order.  The figures below are medians of 16 EXPLAIN
(ANALYZE, TIMING OFF) runs:

case                              initial patch  hybrid    vs. initial
--------------------------------  -------------  --------  -----------
constant pattern                       194.1 ms  185.6 ms        -4.4%
non-constant, fixed at runtime         299.8 ms  199.9 ms       -33.3%
non-constant, changes every row        303.4 ms  304.9 ms        +0.5%
generic fallback control               288.5 ms  289.5 ms        +0.3%

The constant-pattern difference appears to be a compiler-dependent code-layout
effect rather than a benefit of the hybrid design, so I do not interpret it as
a general speedup.  More importantly, the runtime-fixed case captures the
benefit of the verbatim cache, while the row-varying case tracks the generic
fallback control instead of rebuilding the 256-entry skip table for every row.

The attached v2 patch uses this hybrid design.  Thus the common stable path
does not pay a memcmp, runtime-fixed non-constant values can use BMH, and a
pattern that is observed to vary falls back without any rebuild penalty.

> Hm...that collation test and message is already caught and done by
> GenericMatchText, so you could throw !OidIsValid(collation) into that ||
> group as well, and remove the ereport section entirely. It then falls
> through later to GenericMatchText, which complains about the collation
> there.

Done.  The invalid-collation case is now included in the rejection group and
falls through to GenericMatchText.  I removed the duplicate ereport block from
like_bmh.c.

> It did have one test failure:
>
> @@ -151,8 +151,8 @@
>     p    | matched
>  --------+---------
>   %abcd% | t
> - %b%e%  | f
>   %b_d%  | t
> + %b%e%  | f
>   %wxyz% | f
>  (4 rows)
>
> I think it's from the "Row-varying patterns must use the generic matcher."
> test.

Thanks for catching this.  This was a locale-dependent sort-order issue in the
test, not a matcher failure.  The query now uses ORDER BY p COLLATE "C".

I retested the revised patch against PostgreSQL HEAD 0348090: all 246 core
regression tests passed, including like_bmh, and all four contrib/pg_trgm
tests passed.

Thanks,
Atsushi Ogawa