PostgreSQL 10 was released. We have been waiting for new features to appear. My colleagues tested the new version in the spring and sent bug reports: one , two . And I was very pleased with the update and prepared the translation of the Release Notes into Russian. Let's discuss.

Content

E.1.1. Review (TLDR)
E.1.2. Migration to version 10
E.1.3. Changes
E.1.3.1. Server
E.1.3.1.1. Parallel requests
E.1.3.1.2. Indices
E.1.3.1.3. Locks
E.1.3.1.4. Optimizer
E.1.3.1.5. Overall performance
E.1.3.1.6. Monitoring
E.1.3.1.6.1. Logging
E.1.3.1.6.2. pg_stat_activity
E.1.3.1.7. Authentication
E.1.3.1.8. Server configuration
E.1.3.1.9. Reliability
E.1.3.1.9.1. Proactive Write Log (WAL)
E.1.3.2. Replication and recovery
E.1.3.3. Requests
E.1.3.4. Auxiliary commands
E.1.3.5. Data types
E.1.3.6. Functions
E.1.3.7. Server languages
E.1.3.7.1. PL / Tcl
E.1.3.8. Client Interfaces
E.1.3.9. Client applications
E.1.3.9.1. psql
E.1.3.9.2. pgbench
E.1.3.10. Server applications
E.1.3.10.1. pg_dump, pg_dumpall, pg_restore
E.1.3.10.2. pg_basebackup
E.1.3.10.3. pg_ctl
E.1.3.11. Source
E.1.3.12. Additional modules

E.1.1. Review (TLDR)

Major improvements in PostgreSQL 10:

• Logical replication using publish / subscribe
• Declarative table partitioning
• Improved parallel query processing
• Significant overall performance improvements
• Enhanced password checker based on SCRAM-SHA-256
• Improved monitoring and administration

All of the above is described in detail below.

E.1.2. Migration to version 10

To upgrade from any previous versions, you must dump / restore using pg_dumpall , or use pg_upgrade .
Version 10 contains a number of changes that may affect compatibility with previous releases. Here are the changes:

• Hash indexes must be rebuilt after running pg_upgrade from any previous major PostgreSQL versions (Mithun Cy, Robert Haas, Amit Kapila). Important changes in hash indexes make rebuilding necessary. pg_upgrade will create a rebuild help script.
• Rename the pg_xlog log pg_xlog to pg_wal and rename the pg_clog transaction pg_clog to pg_xact (Michael Paquier).
  Users sometimes thought that these directories contained only non-essential log files, deleted proactive write log files or files with transactional status manually, which led to irreparable data loss. Such renames should prevent similar user actions in the future.
• Rename SQL functions, utilities and options related to “xlog” in “wal” (Robert Haas).
  For example, pg_switch_xlog() will become pg_switch_wal() , pg_receivexlog will become pg_receivewal ; --xlogdir will become --waldir . This is done for consistency with the renaming of pg_xlog ; in general, the term “xlog” will not be used by the user in the future.
• Rename WAL functions and views using lsn instead of location (David Rowley). Previously, there were contradictions when using two terms.
• The implementation of the SRF (set-returning functions) involved in the SELECT list (Andres Freund) has been changed. Set-returning functions (SRFs) are now computed before calculating scalar expressions in the SELECT list, as if they were placed in a LATERAL FROM expression. This gives more robust semantics when using multiple SRFs. If they return a different number of rows, the shorter results will be extended to the longest by adding NULL values. Previously, the smallest common factor of the sizes of the SRF results was found, and they were repeated cyclically. Also, the SRF call is now denied with CASE and COALESCE constructs. For more information, see Section 37.4.8 .
• When ALTER TABLE ... ADD PRIMARY KEY sets the attribute of the column NOT NULL , this change also applies to inherited child tables (Michael Paquier).
• The statement trigger statements are prevented from being called more than once per request (statement) (Tom Lane). In cases of using CTE writers, when updating the same table, or with another CTE writer, BEFORE STATEMENT - or AFTER STATEMENT triggers were called more than once. Also, if there were statement-triggers on a table affected by a foreign key constraint (such as ON DELETE CASCADE ), they could be called more than once for an external SQL statement. This was against the standard, so it was changed.
• Sequence metadata has been moved to the new pg_sequence (Peter Eisentraut) system catalog. The new table contains fields that can be modified using the nextval() function, that is, last_value , log_cnt and is_called . The remaining properties of the sequences, such as the initial value and increment, are stored in the corresponding row of the pg_sequence directory. ALTER SEQUENCE now fully transactional, implying locking a sequence to commit. The functions nextval() and setval() remain non-transactional.
  The main incompatibility introduced by this change is that a sample from the table with sequences will return only the three fields listed above. To obtain the remaining properties of the sequence, the application must find them in pg_sequence . The new system representation pg_sequences can also be used for this purpose; it provides column names that are more compatible with existing code.
• pg_basebackup WAL streaming required to restore the default backup (Magnus Hagander). This changes the default value of the -X/--xlog-method pg_basebackup in stream . The value of none is added to reproduce the old behavior. The -x option of the pg_basebackup command pg_basebackup been removed (use -X fetch instead).
• Changed the way pg_hba.conf uses logical replication (Peter Eisentraut). In previous releases, logical replication required the use of the replication keyword in a database column. From this release, logical replication uses the usual occurrence of the name of the database or the keyword all . Physical replication continues to use the replication keyword. Since logical replication is a novelty of this release, the change affects only users of third-party plug-ins for replication.
• All pg_ctl actions are pending default (Peter Eisentraut). Up to this point, some pg_ctl actions did not wait for completion and required the -w option to be used for this.
• The default value of the server parameter log_directory from pg_log to log (Andreas Karlsson).
• The ssl_dh_params_file configuration option has been ssl_dh_params_file to specify a file with modified Diffie-Hellman parameters (Heikki Linnakangas). This removes the hard-case undocumented file name dh1024.pem . Note that dh1024.pem no longer checked by default; You must set the option value to use your own Diffie-Hellman parameters.
• The default Diffie-Hellman parameter size is increased to 2048 for the OpenSSL ephemeral Diffie-Hellman algorithm (Heikki Linnakangas). The size of the supplied Diffie-Hellman parameters has been increased from 1024 to 2048 bits, making Diffie-Hellman key exchange more resistant to brute force attacks. However, some older SSL implementations, in particular, some editions of Java Runtime Environment version 6, will not accept Diffie-Hellman parameters longer than 1024 bits, and therefore will not be able to connect via SSL. If you need to support such old clients, you can use the usual 1024-bit Diffie-Hellman parameters instead of the default ones. See ssl_dh_params_file .
• Removed the ability to save unencrypted passwords on the server (Heikki Linnakangas). The server parameter password_encryption no longer supports off or plain values. The UNENCRYPTED option UNENCRYPTED no longer supported in the CREATE/ALTER USER ... PASSWORD command. Similarly, the option --unencrypted was removed from the --unencrypted command. Unencrypted passwords when migrating from older versions will be kept encrypted in this release. The default value for the password_encryption parameter is still md5 .
• Added server parameters min_parallel_table_scan_size and min_parallel_index_scan_size to manage parallel requests (Amit Kapila, Robert Haas). This replaces the min_parallel_relation_size , which turned out to be too general.
• The value of the shared_preload_libraries parameter and others like it is not reduced to lower case, unless it is in quotes (QL Zhuo). These settings are a list of file names, but up to this point were treated as SQL identifiers, which are processed according to other rules.
• Removed server parameter sql_inheritance (Robert Haas). Changing the default value for this parameter meant that queries to the parent tables did not include data from the child tables. The SQL standard requires that they be included; however, this was the default with PostgreSQL 7.1.
• It is allowed to transfer multidimensional arrays to PL / Python functions and return as nested Python lists (Alexey Grishchenko, Dave Cramer, Heikki Linnakangas). This capability required backward incompatible changes in the processing of arrays of composite PL / Python types. Previously, it was possible to return an array with composite values ​​by writing, for example, [[col1, col2], [col1, col2]] ; but now it is interpreted as a two-dimensional array. Compound types in arrays should now be written as Python tuples, not lists, to eliminate ambiguity; This means that you should write like this: [(col1, col2), (col1, col2)] .
• Removed autoload feature from PL / Tcl's (Tom Lane) module. This functionality has been replaced by two new pltcl.start_proc and pltclu.start_proc server options, which are easier to use and more similar to those found in other procedural languages.
• Removed support for versions up to 8.0 in pg_dump / pg_dumpall (Tom Lane). Users who need to remove a dump from versions before 8.0 need to use PostgreSQL version 9.6 or earlier. The resulting output should successfully boot into newer versions.
• Removed support for timestamps and floating point intervals (Tom Lane). Removed configuration parameter --disable-integer-datetimes . Floating point timestamps have some advantages, but were not used by default with PostgreSQL 8.3.
• Removed server support for client-server protocol version 1.0 (Tom Lane). This protocol was not supported with PostgreSQL 6.3.
• Removed contrib/tsearch2 (Robert Haas) module. This module provided compatibility with the full-text version of the pre-8.3 version.
• Removed the createlang and droplang command line droplang (Peter Eisentraut). They are deprecated in PostgreSQL 9.1. Use the CREATE EXTENSION and DROP EXTENSION commands instead.
• Removed support for feature call convention version 0 (Andres Freund). Extensions that use C functions must comply with version 1 of the function calling convention. Version 0 is outdated in 2001.

E.1.3. Changes

Below is detailed information about the changes between PostgreSQL 10 and the previous major release.

E.1.3.1. Server

E.1.3.1.1. Parallel requests

• Parallel index scans for B-tree (Rahila Syed, Amit Kapila, Robert Haas, Rafia Sabih) are supported. This change allows B-tree index pages to be viewed by various parallel workers.
• Parallel bitmap heap scans (Dilip Kumar) are supported. This allows a single index scan to dispatch a heap of various parallel workers.
• Parallel execution of merge joins (Dilip Kumar) is allowed.
• Parallel execution of unrelated subqueries (Amit Kapila) is allowed.
• Improved the ability of parallel workers to return pre-sorted data (Rushabh Lathia).
• Increased use of parallel queries in procedural functions (Robert Haas, Rafia Sabih).
• Added server parameter max_parallel_workers to limit the number of workers that can be used for parallel processing of the request (Julien Rouhaud). This parameter can be set to a smaller value than max_worker_processes for reserving workers who are different from parallel processing of target requests.

E.1.3.1.2. Indices

• Added WAL write support for hash indexes (Amit Kapila). This makes them resilient to failure and add them to replication. The hazard warning message has been deleted.
• Improved hash index performance (Amit Kapila, Mithun Cy, Ashutosh Sharma).
• Added support for INET and CIDR SP-GiST indexes (Emre Hasegeli).
• Added parameter for more aggressive summation of BRIN-indices (Álvaro Herrera). The new CREATE INDEX option enables the automatic summarization of the previous BRIN page range when a new page range is created.
• Added functions for deleting and re-adding BRIN-summaries for BRIN index ranges (Álvaro Herrera). The new SQL function brin_summarize_range() updates the BRIN-index sumrization for the specified range and brin_desummarize_range() to remove it. This is useful for updating the range summarization, which is now smaller due to UPDATEs and DELETEs.
• Improved accuracy of determining the profitability of scanning the BRIN-index (David Rowley, Emre Hasegeli).
• Faster inserts and updates are allowed through more efficient reuse of GiST index pages (Andrey Borodin).
• Reduced page blocking level during GIN-index update (Andrey Borodin).

E.1.3.1.3. Locks

• Reduced the level of blocking required to add values ​​to enums (Andrew Dunstan, Tom Lane). Previously, it was not possible to run the ALTER TYPE ... ADD VALUE command in a transaction block if this enumeration was not created in it. Now only links to uncommitted values ​​of transfers from other transactions are prohibited.
• Reduced the level of locks required to change the parameters of the tables (Simon Riggs, Fabrízio Mello). For example, changing the setting of the effective_io_concurrency table can be done with a more lightweight lock.
• The setting of the predicate lock thresholds (Dagfinn Ilmari Mannsåker) is allowed. The promotion of locks can be controlled by two new server parameters: max_pred_locks_per_relation and max_pred_locks_per_page .

E.1.3.1.4. Optimizer

• Added multicolumn optimizer statistics for calculating the correlation coefficient and the number of unique values ​​(Tomas Vondra, David Rowley, Álvaro Herrera). Added new CREATE STATISTICS , ALTER STATISTICS and DROP STATISTICS commands. This feature is useful when evaluating the memory usage of a query and when combining statistics from different columns.
• The performance of queries affected by row-level security (Tom Lane) has been increased. The optimizer has a lot of knowledge about where it can place RLS filters, allowing you to build better plans and securely provide RLS conditions.

E.1.3.1.5. Overall performance

• Aggregate functions are computed that calculate the current amount when using numeric type arithmetic, including some variants of SUM() , AVG() and STDDEV() (Heikki Linnakangas).
• Improved performance of character encoding transformations using radix trees (Kyotaro Horiguchi, Heikki Linnakangas).
• The overhead of calculating expressions during the execution of the query, as well as the overhead of calling plan nodes (Andres Freund), has been reduced. This is particularly useful for queries that process a large number of rows.
• Added ability to use hash aggregation with grouping sets (Andrew Gierth).
• Uniqueness guarantees are used to optimize certain types of associations (David Rowley).
• The sorting performance of the macaddr data macaddr (Brandur Leach) has been increased.
• The overhead of tracking statistics in sessions that use many thousands of relationships (Aleksander Alekseev) has been reduced.

E.1.3.1.6. Monitoring

• Added explicit control of displaying the time of planning and executing the EXPLAIN command (Ashutosh Bapat). By default, the planning and execution time was displayed by the EXPLAIN ANALYZE command and was not displayed in other cases. The new option command EXPLAIN SUMMARY allows you to explicitly control the display.
• Added default monitoring rules (Dave Page). The new roles pg_monitor , pg_read_all_settings , pg_read_all_stats and pg_stat_scan_tables make it easier to configure privileges.
• Correct update of the statistics collector at runtime REFRESH MATERIALIZED VIEW (Jim Mlodgenski).

E.1.3.1.6.1. Logging

• Changed the default value of the log_line_prefix parameter to include a timestamp with milliseconds and PID in each postmaster output line (Christoph Berg). In the past, the prefix was empty.
• Added functions to return the contents of the log directory and WAL (Dave Page)
  New functions pg_ls_logdir () and pg_ls_waldir () can be run by non-super users with the correct privilege settings.
• Added pg_current_logfile () function to display the name of the file to which the error collector writes the current stderr or csvlog (Gilles Darold).
• The address and port number of each listening socket is reported in the server log during postmaster startup (Tom Lane). In addition, if there is an error connecting to the listening socket, the specific address to which you tried to connect is indicated.
• Reduced the number of messages in the log about starting and stopping launcher subprocesses (Tom Lane). Now they use the message level DEBUG1 .
• Reduced the verbosity of low debug levels of messages managed by log_min_messages (Robert Haas). It also changes the verbosity of the debug levels of client_min_messages .

E.1.3.1.6.2. pg_stat_activity

• Added reporting of low-level wait states in pg_stat_activity (Michael Paquier, Robert Haas, Rushabh Lathia). This change allows you to report multiple low-level locks, including latch waits, read / write fsync files, client read / write, and synchronous replication.
• Added display of background processes, background workers and walsender processes in pg_stat_activity (Kuntal Ghosh, Michael Paquier). This simplifies monitoring. New column backend_type identifies the type of process.
• SQL mapping of running parallel workers (Rafia Sabih) has been added to pg_stat_activity.
• Renamed pg_stat_activity.wait_event_type LWLockTranche and LWLockNamed to LWLock (Robert Haas). This makes the output more consistent.

E.1.3.1.7. Authentication

• Added support for SCRAM-SHA-256 for transferring and storing passwords (Michael Paquier, Heikki Linnakangas). This gives more security than the current ( md5 ) transfer and storage method.
• Changed server password type of the password_encryption from boolean to enum (Michael Paquier). This was necessary to support additional password hashing options.
• Added pg_hba_file_rules to display pg_hba.conf content (Haribabu Kommi). The contents of the file are shown, but not the current active settings.
• Support for multiple RADIUS servers (Magnus Hagander). All parameters related to RADIUS are now recorded in plural and support a comma-separated list of servers.

E.1.3.1.8. Server configuration

• The SSL configuration has become possible to update when the configuration is reloaded (Andreas Karlsson, Tom Lane). This allows you to reconfigure SSL without rebooting the server using pg_ctl reload , SELECT pg_reload_conf() or sending a SIGHUP signal. However, reloading the SSL configuration does not work if the SSL server key requires a password, since it is not possible to prompt for a password. In this case, the initial configuration will be applied.
• The maximum value of bgwriter_lru_maxpages is made really unlimited (Jim Nasby).

E.1.3.1.9. Reliability

• After creating or deleting files, the fsync call is made to their parent directory (Michael Paquier). This reduces the risk of data loss in the event of a power failure.

E.1.3.1.9.1. Proactive Write Log (WAL)

• Removed unnecessary checkpoints and WAL archiving with system downtime (Michael Paquier)
• The server parameter wal_consistency_checking has been added to add WAL information that needs to be checked on the standby (Kuntal Ghosh, Robert Haas). Any verification error generates a fatal error on standbay.
• Increased maximum configurable WAL segment to gigabytes (Beena Emerson)
  The larger size of the WAL segment allows you to call archive_command less often and manage fewer WAL files.

E.1.3.2. Replication and recovery

• Added the ability to logically replicate tables on standby (Petr Jelinek). Logical replication provides more flexibility than physical replication, including replication between different major versions of PostgreSQL and selective replication.
• It is allowed to wait for confirmation of a commit from standbays regardless of the order in which they are listed in synchronous_standby_names (Masahiko Sawada). , synchronous_standby_names . synchronous_standby_names ANY , . .
• (Magnus Hagander, Dang Minh Huong). , wal_level , max_wal_senders , max_replication_slots hot_standby , .
• pg_hba.conf (Michael Paquier). pg_hba.conf . , pg_basebackup .
• pg_stat_replication (Thomas Munro). : write_lag , flush_lag replay_lag .
• Log Sequence Number (LSN) recovery.conf (Michael Paquier). XID'.
• pg_stop_backup() , WAL' (David Steele). pg_stop_backup() .
• (Petr Jelinek). .
• Access Exclusive (Simon Riggs, David Rowley).
• (Stas Kelvich, Nikhil Sontakke, Michael Paquier).

E.1.3.3. Requests

• XMLTABLE , XML - (Pavel Stehule, Álvaro Herrera).
• UPDATE ... SET (column_list) = row_constructor (Tom Lane). row_constructor ROW ; . table_name.* row_constructor , row_constructors .
• , Unicode U+7FF (Tom Lane). , , [[:alpha:]] .

E.1.3.4.

• , (Amit Langote). .
• AFTER - , (Kevin Grittner, Thomas Munro). , .
• (Stephen Frost). , , . , . .
• REFERENCES - (Tom Lane). REFERENCES - . , - SQL. ( ) , REFERENCES - .
• (Matheus Oliveira). ALTER DEFAULT PRIVILEGES .
• CREATE SEQUENCE AS , (Peter Eisentraut). , .
• COPY view FROM source INSTEAD INSERT - (Haribabu Kommi). COPY .
• DDL, (Peter Eisentraut). , DROP FUNCTION , . SQL.
• , DROP (Peter Eisentraut).
• IF NOT EXISTS CREATE SERVER , CREATE USER MAPPING CREATE COLLATION (Anastasia Lubennikova, Peter Eisentraut).
• VACUUM VERBOSE xmin (Masahiko Sawada, Simon Riggs). log_autovacuum_min_duration .
• VACUUM - (Claudio Freire, Álvaro Herrera).

E.1.3.5. Data types

• JSON JSONB (Dmitry Dolgov). ts_headline() to_tsvector() .
• EUI-64 MAC- macaddr8 (Haribabu Kommi). MAC- EUI-48 (type macaddr ).
• (Peter Eisentraut). SERIAL , SQL.
• ENUM (Dagfinn Ilmari Mannsåker). ALTER TYPE… RENAME VALUE .
• ( anyarray ) to_json() to_jsonb() (Andrew Dunstan). , anyarray ( , pg_stats ) JSON -, .
• of money int8 (Peter Eisentraut). int8 - float8 money -- float8 . . , money int8 , , .
• money (Peter Eisentraut).

E.1.3.6. Functions

• regexp_match() (Emre Hasegeli). regexp_matches() , , , .
• jsonb , (Magnus Hagander).
• json_populate_record() JSON (Nikita Glukhov). SQL JSON, JSON. , array_in() record_in() JSON-, , .
• txid_current_ifassigned() NULL , (Craig Ringer). txid_current() , , . .
• txid_status() (Craig Ringer). , , .
• make_date() , . e. (Álvaro Herrera).
• to_timestamp() to_date() (Artur Zakirov). , to_date('2009-06-40','YYYY-MM-DD') 2009-07-10 . .

E.1.3.7.

• cursor() execute() PL/Python (Peter Eisentraut). - .
• GET DIAGNOSTICS PL/pgSQL (Tom Lane). .

E.1.3.7.1. PL/Tcl

• PL/Tcl (Karl Lehenbauer).
• PL/Tcl (Victor Wagner). PL/Tcl- .
• pltcl.start_proc pltclu.start_proc PL/Tcl (Tom Lane).

E.1.3.8.

• URI libpq (Robert Haas, Heikki Linnakangas). libpq .
• URI libpq / (Victor Wagner, Mithun Cy). . libpq target_session_attrs .
• libpq (Julian Markwort). .
• PQencryptPasswordConn() , (Michael Paquier, Heikki Linnakangas). MD5 - PQencryptPassword() . SCRAM-SHA-256- .
• ecpg 4.12 10 (Tom Lane). ecpg PostgreSQL.

E.1.3.9. Client applications

E.1.3.9.1. psql

• psql (Corey Huinker). psql \if , \elif , \else \endif . , .
• psql \gx ( \g ) ( \x ) (Christoph Berg).
• psql , (Tom Lane). , psql.
• psql (Daniel Vérité, Tom Lane). psql . \set , . , \set on ; , . \unset , , . , , psql.
• , psql (Fabien Coelho).
• psql \d ( ) \dD ( ), , null, (Peter Eisentraut)
  “Modifiers”.
• \d (Daniel Gustafsson). stderr, stdout, .
• psql (Jeff Janes, Ian Barwick, Andreas Karlsson, Sehrope Sarkuni, Thomas Munro, Kevin Grittner, Dagfinn Ilmari Mannsåker).

E.1.3.9.2. pgbench

• pgbench --log-prefix (Masahiko Sawada).
• pgbench (Fabien Coelho). .
• -M , (Tom Lane).

E.1.3.10. Server applications

• pg_receivewal -Z/--compress (Michael Paquier).
• pg_recvlogical --endpos (Craig Ringer). --startpos .
• initdb --noclean --nosync --no-clean --no-sync (Vik Fearing, Peter Eisentraut). .

E.1.3.10.1. pg_dump , pg_dumpall , pg_restore

• pg_restore (Michael Banck). -N/--exclude-schema .
• --no-blobs pg_dump (Guillaume Lelarge). .
• pg_dumpall --no-role-passwords , (Robins Tharakan, Simon Riggs). pg_dumpall , .
• (Petr Jelinek).
• fsync() , pg_dump pg_dumpall (Michael Paquier). , .. . --no-sync .

E.1.3.10.2. pg_basebackup

• pg_basebackup WAL tar- (Magnus Hagander). WAL tar- base backup'.
• pg_basebackup (Magnus Hagander). , pg_basebackup WAL .
• fsync pg_basebackup pg_receivewal (Michael Paquier).
• pg_basebackup --no-sync fsync (Michael Paquier).
• pg_basebackup', (David Steele).

E.1.3.10.3. pg_ctl

• pg_ctl (Peter Eisentraut).
• pg_ctl wait ( --wait ) no-wait ( --no-wait ) (Vik Fearing).
• pg_ctl ( --options ) (Peter Eisentraut).
• pg_ctl start --wait postmaster.pid , (Tom Lane). postmaster , postmaster.pid, pg_ctl , , . , , .
• pg_ctl / postmaster' (Tom Lane). pg_ctl postmaster', .
• pg_ctl , (Peter Eisentraut). start promote 1, 0, . stop .

E.1.3.11. Source

• (Peter Eisentraut, Tom Lane). (, 10.1), (eg, 9.6.3). , — . (, 10 9.6). , PostgreSQL .
• pgindent (Piotr Stefaniak, Tom Lane). pg_bsd_indent, , FreeBSD. , C. , (, , ), , .
• ICU (Peter Eisentraut). ICU , . --with-icu . .
• PG_FUNCTION_INFO_V1 DLLEXPORT Windows (Laurenz Albe). extern , DLLEXPORT .
• SPI SPI_push() , SPI_pop() , SPI_push_conditional() , SPI_pop_conditional() SPI_restore_connection() (Tom Lane). . , , .
  , SPI_palloc() SPI-; palloc() , . .
• (Thomas Munro, Robert Haas).
• slab-like (Tomas Vondra).
• POSIX SysV Linux FreeBSD (Tom Lane). SysV.
• 64- - (Andres Freund).
• 64- ARM64 (Roman Shaposhnik).
• clock_gettime() , , (Tom Lane). gettimeofday() , clock_gettime() .
• (Magnus Hagander, Michael Paquier, Heikki Linnakangas). , configure , --disable-strong-random . pgcrypto , .
• WaitLatchOrSocket() Windows (Andres Freund).
• tupconvert.c , OID (Ashutosh Bapat, Tom Lane). OID , Datum, , , OID.
• SCO Unixware (Tom Lane).
• (Alexander Lakhin).
• XSLT PostgreSQL (Peter Eisentraut). Jade, DSSSL, JadeTex.
• HTML- XSLT- (Peter Eisentraut).

E.1.3.12. Additional modules

• file_fdw , (Corey Huinker, Adam Gomaa).
• postgres_fdw , (Jeevan Chalke, Ashutosh Bapat). , , .
• postgres_fdw (David Rowley, Ashutosh Bapat, Etsuro Fujita).
• OID postgres_fdw (Etsuro Fujita). OID .
• btree_gist btree_gin (Andrew Dunstan). exclusion constraints.
• btree_gist UUID (Paul Jungwirth).
• amcheck , B-tree (Peter Geoghegan).
• $N ? pg_stat_statements (Lukas Fittl).
• - ( cube ) (Tom Lane). infinite- NaN -.
• pg_buffercache (Ivan Kartyshov). .
• pgstattuple pgstathashindex() - (Ashutosh Sharma).
• GRANT pgstattuple (Stephen Frost). .
• - pgstattuple (Amit Kapila).
• pageinspect page_checksum() , (Tomas Vondra).
• pageinspect bt_page_items() (Tomas Vondra).
• pageinspect - (Jesper Pedersen, Ashutosh Sharma).

From translator

.

— . . , WAL.

. , , .

.

: .

? ?