public inbox for [email protected]help / color / mirror / Atom feed
[PATCH v27 5/5] Doc: Add Collation Versions section. 3+ messages / 2 participants [nested] [flat]
* [PATCH v27 5/5] Doc: Add Collation Versions section. @ 2020-03-11 02:01 Thomas Munro <[email protected]> 0 siblings, 0 replies; 3+ messages in thread From: Thomas Munro @ 2020-03-11 02:01 UTC (permalink / raw) Supply a brief introduction to collation version concepts. Author: Thomas Munro <[email protected]> Reviewed-by: Julien Rouhaud <[email protected]> Discussion: https://postgr.es/m/CAEepm%3D0uEQCpfq_%2BLYFBdArCe4Ot98t1aR4eYiYTe%3DyavQygiQ%40mail.gmail.com --- doc/src/sgml/charset.sgml | 35 +++++++++++++++++++++++++++++++++++ 1 file changed, 35 insertions(+) diff --git a/doc/src/sgml/charset.sgml b/doc/src/sgml/charset.sgml index 4b4563c5b9..12a82ff18c 100644 --- a/doc/src/sgml/charset.sgml +++ b/doc/src/sgml/charset.sgml @@ -948,6 +948,41 @@ CREATE COLLATION ignore_accents (provider = icu, locale = 'und-u-ks-level1-kc-tr </tip> </sect3> </sect2> + + <sect2 id="collation-versions"> + <title>Collation Versions</title> + + <para> + The ordering defined by a collation is not necessarily fixed over time. + If a collation changes for any reason, persistent data structures such as + b-trees that depend on a stable ordering of text might be corrupted. + <productname>PostgreSQL</productname> defends against this by recording + the current version of each referenced collation for any index that + depends on it in the + <link linkend="catalog-pg-depend"><structname>pg_depend</structname></link> + catalog, if the collation provider makes it available. If the provider + later begins to report a different version, a warning will be issued + when the index is accessed, until either the <xref linkend="sql-reindex"/> + or the <xref linkend="sql-alterindex"/> command is used to update the + version. + </para> + <para> + Version information is available for collations from the + <literal>icu</literal> provider on all operating systems. For the + <literal>libc</literal> provider, versions are currently only available + on systems using the GNU C library (most Linux systems). + </para> + + <note> + <para> + When using the GNU C library for collations, the C library's version + is used as a proxy for the collation version. Many Linux distributions + change collation definitions only when upgrading the C library, but this + approach is imperfect as maintainers are free to back-port newer + collation definitions to older C library releases. + </para> + </note> + </sect2> </sect1> <sect1 id="multibyte"> -- 2.20.1 --o56keubw4jpehgac-- ^ permalink raw reply [nested|flat] 3+ messages in thread
* Using pg_bitutils.h in tidbitmap.c. @ 2025-04-14 12:41 Thomas Munro <[email protected]> 2025-04-24 01:47 ` Re: Using pg_bitutils.h in tidbitmap.c. John Naylor <[email protected]> 0 siblings, 1 reply; 3+ messages in thread From: Thomas Munro @ 2025-04-14 12:41 UTC (permalink / raw) To: pgsql-hackers tidbitmap.c's operations loop over all the bits, but could leap over zeros with bitmap instructions like bitmapset.c. Hard to imagine that matters, but I think it also comes out easier to read? Attachments: [text/x-patch] 0001-Use-pg_bitutils.h-in-tidbitmap.c.patch (4.8K, ../../CA+hUKGLRN4_8sPHrwr4HDRV-_+_MddOyJ_Wc=2C2j6Hc0kkNig@mail.gmail.com/2-0001-Use-pg_bitutils.h-in-tidbitmap.c.patch) download | inline diff: From 7eec670224270c098c2fb3ad4b7748b0769a1e95 Mon Sep 17 00:00:00 2001 From: Thomas Munro <[email protected]> Date: Wed, 19 Mar 2025 02:38:26 +1300 Subject: [PATCH] Use pg_bitutils.h in tidbitmap.c. Instead of looping over every bit, skip the zeros. --- src/backend/nodes/tidbitmap.c | 108 ++++++++++++++++++---------------- 1 file changed, 57 insertions(+), 51 deletions(-) diff --git a/src/backend/nodes/tidbitmap.c b/src/backend/nodes/tidbitmap.c index 41031aa8f2f..6ca01bdb772 100644 --- a/src/backend/nodes/tidbitmap.c +++ b/src/backend/nodes/tidbitmap.c @@ -44,6 +44,7 @@ #include "common/int.h" #include "nodes/bitmapset.h" #include "nodes/tidbitmap.h" +#include "port/pg_bitutils.h" #include "storage/lwlock.h" #include "utils/dsa.h" @@ -242,6 +243,20 @@ static int tbm_shared_comparator(const void *left, const void *right, #include "lib/simplehash.h" +/* + * Return the position of the rightmost bit of *word, and also clear that bit. + * Useful for looping over the ones in a word. *word must not be zero. + */ +static inline int +bmw_pop_rightmost_one(bitmapword *word) +{ + int position = bmw_rightmost_one_pos(*word); + + *word &= ~((bitmapword) 1 << position); + + return position; +} + /* * tbm_create - create an initially-empty bitmap * @@ -474,24 +489,20 @@ tbm_union_page(TIDBitmap *a, const PagetableEntry *bpage) if (bpage->ischunk) { + BlockNumber blockno = bpage->blockno; + /* Scan b's chunk, mark each indicated page lossy in a */ for (wordnum = 0; wordnum < WORDS_PER_CHUNK; wordnum++) { bitmapword w = bpage->words[wordnum]; - if (w != 0) + while (w != 0) { - BlockNumber pg; + int bitnum = bmw_pop_rightmost_one(&w); - pg = bpage->blockno + (wordnum * BITS_PER_BITMAPWORD); - while (w != 0) - { - if (w & 1) - tbm_mark_page_lossy(a, pg); - pg++; - w >>= 1; - } + tbm_mark_page_lossy(a, blockno + bitnum); } + blockno += BITS_PER_BITMAPWORD; } } else if (tbm_page_is_lossy(a, bpage->blockno)) @@ -584,38 +595,29 @@ tbm_intersect_page(TIDBitmap *a, PagetableEntry *apage, const TIDBitmap *b) { /* Scan each bit in chunk, try to clear */ bool candelete = true; + BlockNumber blockno = apage->blockno; for (wordnum = 0; wordnum < WORDS_PER_CHUNK; wordnum++) { bitmapword w = apage->words[wordnum]; + bitmapword neww = w; - if (w != 0) + while (w != 0) { - bitmapword neww = w; - BlockNumber pg; - int bitnum; + int bitnum = bmw_pop_rightmost_one(&w); + BlockNumber pg = blockno + bitnum; - pg = apage->blockno + (wordnum * BITS_PER_BITMAPWORD); - bitnum = 0; - while (w != 0) + if (!tbm_page_is_lossy(b, pg) && + tbm_find_pageentry(b, pg) == NULL) { - if (w & 1) - { - if (!tbm_page_is_lossy(b, pg) && - tbm_find_pageentry(b, pg) == NULL) - { - /* Page is not in b at all, lose lossy bit */ - neww &= ~((bitmapword) 1 << bitnum); - } - } - pg++; - bitnum++; - w >>= 1; + /* Page is not in b at all, lose lossy bit */ + neww &= ~((bitmapword) 1 << bitnum); } - apage->words[wordnum] = neww; - if (neww != 0) - candelete = false; } + apage->words[wordnum] = neww; + if (neww != 0) + candelete = false; + blockno += BITS_PER_BITMAPWORD; } return candelete; } @@ -910,21 +912,14 @@ tbm_extract_page_tuple(TBMIterateResult *iteritem, { bitmapword w = page->words[wordnum]; - if (w != 0) + while (w != 0) { - int off = wordnum * BITS_PER_BITMAPWORD + 1; + int bitnum = bmw_pop_rightmost_one(&w); + int off = wordnum * BITS_PER_BITMAPWORD + bitnum + 1; - while (w != 0) - { - if (w & 1) - { - if (ntuples < max_offsets) - offsets[ntuples] = (OffsetNumber) off; - ntuples++; - } - off++; - w >>= 1; - } + if (ntuples < max_offsets) + offsets[ntuples] = off; + ntuples++; } } @@ -938,18 +933,29 @@ static inline void tbm_advance_schunkbit(PagetableEntry *chunk, int *schunkbitp) { int schunkbit = *schunkbitp; + int wordnum = WORDNUM(schunkbit); + int bitnum = BITNUM(schunkbit); + bitmapword w; + + /* Chop off bits to the right of bitnum in starting word. */ + w = chunk->words[wordnum] & ~((bitmapword) 0) << bitnum; - while (schunkbit < PAGES_PER_CHUNK) + for (;;) { - int wordnum = WORDNUM(schunkbit); - int bitnum = BITNUM(schunkbit); + if (w != 0) + { + *schunkbitp = + wordnum * BITS_PER_BITMAPWORD + bmw_rightmost_one_pos(w); + return; + } - if ((chunk->words[wordnum] & ((bitmapword) 1 << bitnum)) != 0) + if (++wordnum == WORDS_PER_CHUNK) break; - schunkbit++; + w = chunk->words[wordnum]; } - *schunkbitp = schunkbit; + /* No bits found. Point past end. */ + *schunkbitp = WORDS_PER_CHUNK * BITS_PER_BITMAPWORD; } /* -- 2.39.5 ^ permalink raw reply [nested|flat] 3+ messages in thread
* Re: Using pg_bitutils.h in tidbitmap.c. 2025-04-14 12:41 Using pg_bitutils.h in tidbitmap.c. Thomas Munro <[email protected]> @ 2025-04-24 01:47 ` John Naylor <[email protected]> 0 siblings, 0 replies; 3+ messages in thread From: John Naylor @ 2025-04-24 01:47 UTC (permalink / raw) To: Thomas Munro <[email protected]>; +Cc: pgsql-hackers On Mon, Apr 14, 2025 at 7:42 PM Thomas Munro <[email protected]> wrote: > > tidbitmap.c's operations loop over all the bits, but could leap over > zeros with bitmap instructions like bitmapset.c. Hard to imagine that > matters, but I think it also comes out easier to read? Interesting idea -- I toyed with something like this when TID store was being developed. Most of it looks more readable at a glance, although tbm_advance_schunkbit() may be a wash. +bmw_pop_rightmost_one(bitmapword *word) +{ + int position = bmw_rightmost_one_pos(*word); + + *word &= ~((bitmapword) 1 << position); + + return position; +} There's a more succinct way to write the second statement, since here we only care about the rightmost bit: *word &= *word - 1; That has a bonus in that we don't need to know "position" beforehand, so the CPU can schedule that independently of the bitscan, which is 3 or 4 cycles on modern hardware, but like you said, I'm not sure if that matters. -- John Naylor Amazon Web Services ^ permalink raw reply [nested|flat] 3+ messages in thread
end of thread, other threads:[~2025-04-24 01:47 UTC | newest] Thread overview: 3+ messages (download: mbox mbox.gz follow: Atom feed) -- links below jump to the message on this page -- 2020-03-11 02:01 [PATCH v27 5/5] Doc: Add Collation Versions section. Thomas Munro <[email protected]> 2025-04-14 12:41 Using pg_bitutils.h in tidbitmap.c. Thomas Munro <[email protected]> 2025-04-24 01:47 ` Re: Using pg_bitutils.h in tidbitmap.c. John Naylor <[email protected]>
This inbox is served by agora; see mirroring instructions for how to clone and mirror all data and code used for this inbox