Contents

• PostgreSQL Views and Functions
• Repositories
• Building the extension
• Installation
  • Install the package from source
  • Install the package with PGXN
• Table of content
• List of functions
  • Functions to check data types
    • FUNCTION is_date
    • FUNCTION is_time
    • FUNCTION is_timestamp
    • FUNCTION is_real
    • FUNCTION is_double_precision
    • FUNCTION is_numeric
    • FUNCTION is_bigint
    • FUNCTION is_integer
    • FUNCTION is_smallint
    • FUNCTION is_boolean
    • FUNCTION is_json
    • FUNCTION is_jsonb
    • FUNCTION is_empty
    • FUNCTION is_hex
  • Functions about encryption
    • FUNCTION sha256
  • Functions and views to get extended system information
    • FUNCTION pg_schema_size
    • VIEW pg_db_views
    • VIEW pg_foreign_keys
    • VIEW pg_functions
    • VIEW pg_active_locks
    • VIEW pg_table_matview_infos
    • VIEW pg_object_ownership
    • VIEW pg_partitioned_tables_infos
  • Functions about encodings
    • FUNCTION is_encoding
    • FUNCTION is_latin1
    • FUNCTION return_not_part_of_latin1
    • FUNCTION replace_latin1
    • FUNCTION return_not_part_of_encoding
    • FUNCTION replace_encoding
  • User defined aggregates
    • AGGREGATE gap_fill
    • AGGREGATE array_min
    • AGGREGATE array_max
    • AGGREGATE array_avg
    • AGGREGATE array_sum
  • Format functions
    • German formats
  • Conversion functions
    • FUNCTION to_unix_timestamp
    • FUNCTION hex2bigint
  • Other functions
    • FUNCTION array_trim

The functions and views published in this repository are some of those, which I regularly need in my daily job.

As these procedures are sort of essential, at least to me, I install them in the public schema. That way they are available for everyone without having to know a schema name.

All functions and views are covered by tests. The tests are done in simple SQL statements. Each test does raise division by zero if it fails.

The tests have been done with PostgreSQL 11 up to 16 beta 1.

All examples have been tested with PostgreSQL 16 beta 1, differences in the behavior of previous versions are noted.

The main repository is now on GitLab. A mirror will stay on GitHub.

If you discover any issue, please file them on https://gitlab.com/sjstoelting/pgsql-tweaks/issues.

All functions and tests are located in single files.

The files for the extension are build by the shell script "create-sql.sh". To be able to run the script, you need to have a configuration with the connection information. Please copy "build.cfg.example" to "build.cfg" and change the configuration to fit your environment.
The script assumes, that you have a .pgpass file with login information matching the configuration.

You may either, install all functions as a package, or install single functions of your choice.

Get the source by either, download the code as ZIP file, or by git clone.

cd pgsql_tweaks
make install

Afterwards you are able to create the extension in a database:

CREATE EXTENSION pgsql_tweaks;

pgsql_tweaks is now available over the PostgreSQL extension management, PGXN.

The installation is done with the PGXN installer.

pgxn install pgsql_tweaks

Afterwards you are able to create the extension in a database:

1 List of functions

1.1 Functions to check data types
1.1.1 FUNCTION is_date
1.1.2 FUNCTION is_time
1.1.3 FUNCTION is_timestamp
1.1.4 FUNCTION is_real
1.1.5 FUNCTION is_double_precision
1.1.6 FUNCTION is_numeric
1.1.7 FUNCTION is_bigint
1.1.8 FUNCTION is_integer
1.1.9 FUNCTION is_smallint
1.1.10 FUNCTION is_boolean
1.1.11 FUNCTION is_json
1.1.12 FUNCTION is_jsonb
1.1.13 FUNCTION is_empty
1.1.14 FUNCTION is_hex

1.2 Functions about encryption
1.2.1 FUNCTION sha256

1.3 Functions and views to get extended system information
1.3.1 FUNCTION pg_schema_size
1.3.2 VIEW pg_db_views
1.3.3 VIEW pg_foreign_keys
1.3.4 VIEW pg_functions
1.3.4 VIEW pg_active_locks
1.3.5 VIEW pg_table_matview_infos
1.3.6 VIEW pg_object_ownership
1.3.7 VIEW pg_partitioned_tables_infos

1.4 Functions about encodings
1.4.1 FUNCTION is_encoding
1.4.2 FUNCTION is_latin1
1.4.3 FUNCTION return_not_part_of_latin1
1.4.4 FUNCTION replace_latin1
1.4.4.1 replace_latin1(s text)
1.4.4.2 replace_latin1(s text, replacement text)
1.4.4.3 replace_latin1(s text, s_search text[], s_replace text[])
1.4.5 FUNCTION return_not_part_of_encoding
1.4.6 FUNCTION replace_encoding
1.4.6.1 replace_encoding(s text, e text)
1.4.6.2 replace_encoding(s text, e text, replacement text)
1.4.6.3 replace_encoding(s text, s_search text[], s_replace text[])

1.5 User defined aggregates
1.5.1 AGGREGATE gap_fill
1.5.2 AGGREGATE array_min
1.5.3 AGGREGATE array_max
1.5.4 AGGREGATE array_avg
1.5.5 AGGREGATE array_sum

1.6 Format functions
1.6.1 FUNCTION date_de
1.6.2 FUNCTION datetime_de

1.7 Conversion functions
1.7.1 FUNCTION to_unix_timestamp
1.7.2 FUNCTION hex2bigint

1.8 Other functions
1.8.1 FUNCTION array_trim

The function checks strings for being a date.
You might pass a second parameter to use a format string. Without the format, the default format of PostgreSQL is used.

There has been a behaviour change in PostgreSQL 10. A conversion is now handled strict, as in previous versions the conversion tried to calculate dates.

SELECT is_date('31.02.2018', 'DD.MM.YYYY') AS res;
-- Result is true because the conversion would return a valid result for
SELECT to_date('31.02.2018', 'DD.MM.YYYY');

Result PostgreSQL 9.6 and previous versions:

/**
 * Parameter is in PostgreSQL default format
 */
SELECT is_date('2018-01-01') AS res;

Result:

SELECT is_date('2018-02-31') AS res;

Result:

/**
 * Parameter is in PostgreSQL German format
 */
SELECT is_date('01.01.2018', 'DD.MM.YYYY') AS res;

Result:

SELECT is_date('31.02.2018', 'DD.MM.YYYY') AS res;

Result:

The function checks strings for being a time.
You might pass a second parameter to use a format string. Without the format, the default format of PostgreSQL is used.

There has been a behaviour change in PostgreSQL 10. A conversion is now handled strict, as in previous versions the conversion tried to calculate time.

SELECT is_time('25.33.55,456574', 'HH24.MI.SS,US') AS res;
-- Result is true because the conversion would return a valid result for
SELECT to_timestamp('25.33.55,456574', 'HH24.MI.SS,US')::TIME;

Result PostgreSQL 9.6 and previous versions:

/**
 * Parameter is in PostgreSQL default format
 */
SELECT is_time('14:33:55.456574') AS res;

Result:

SELECT is_time('25:33:55.456574') AS res;

Result:

/**
 * Parameter is some time format
 */
SELECT is_time('14.33.55,456574', 'HH24.MI.SS,US') AS res;

Result:

SELECT is_time('25.33.55,456574', 'HH24.MI.SS,US') AS res;

Result:

The function checks strings for being a timestamp.
You might pass a second parameter to use a format string. Without the format, the default format of PostgreSQL is used.

There has been a behaviour change in PostgreSQL 10. A conversion is now handled strict, as in previous versions the conversion tried to calculate a date.

SELECT is_timestamp('2018-01-01 25:00:00') AS res;
-- Result is true because the conversion would return a valid result for
SELECT to_timestamp('01.01.2018 25:00:00', 'DD.MM.YYYY HH24.MI.SS')::TIMESTAMP;

Result PostgreSQL 9.6 and previous versions:

/**
 * Parameter is in PostgreSQL default format
 */
SELECT is_timestamp('2018-01-01 00:00:00') AS res;

Result:

SELECT is_timestamp('2018-01-01 25:00:00') AS res;
-- Result is false in PostgreSQL >= 10

Result:

/**
 * Parameter is in PostgreSQL German format
 */
SELECT is_timestamp('01.01.2018 00:00:00', 'DD.MM.YYYY HH24.MI.SS') AS res;

Result:

SELECT is_timestamp('01.01.2018 25:00:00', 'DD.MM.YYYY HH24.MI.SS') AS res;

Result:

The function checks strings for being of data type REAL.

SELECT is_real('123.456') AS res;

Result:

SELECT is_real('123,456') AS res;
-- Result is false

Result:

The function checks strings for being of data type DOUBLE PRECISION.

SELECT is_double_precision('123.456') AS res;

Result:

SELECT is_double_precision('123,456') AS res;
-- Result is false

Result:

The function checks strings for being of data type NUMERIC.

SELECT is_numeric('123') AS res;

Result:

SELECT is_numeric('1 2') AS res;
-- Result is false

Result:

The function checks strings for being of data type BIGINT.

SELECT is_bigint('3243546343') AS res;
-- Result is true

Result:

SELECT is_bigint('123.456') AS res;
-- Result is false

The function checks strings for being of data type INTEGER.

SELECT is_integer('123') AS res;
-- Result is true

Result:

SELECT is_integer('123.456') AS res;
-- Result is false

The function checks strings for being of data type SMALLINT.

SELECT is_smallint('123') AS res;
-- Result is true

Result:

SELECT is_smallint('123.456') AS res;
-- Result is false

Result:

The function checks a string variable for containing valid BOOLEAN values.

SELECT is_boolean('t') AS res;
-- Result is true

Result:

SELECT is_boolean('F') AS res;
-- Result is true

Result:

SELECT is_boolean('True') AS res;
-- Result is true

Result:

SELECT is_boolean('False');
-- Result is true

Result:

SELECT is_boolean('0') AS res;
-- Result is true

Result:

SELECT is_boolean('1') AS res;
-- Result is true

Result:

SELECT is_boolean('-1') AS res;
-- Result is false

Result:

The function checks a string variable for containing a valid JSON.

SELECT is_json('{"review": {"date": "1970-12-30", "votes": 10, "rating": 5, "helpful_votes": 0}, "product": {"id": "1551803542", "group": "Book", "title": "Start and Run a Coffee Bar (Start & Run a)", "category": "Business & Investing", "sales_rank": 11611, "similar_ids": ["0471136174", "0910627312", "047112138X", "0786883561", "0201570483"], "subcategory": "General"}, "customer_id": "AE22YDHSBFYIP"}') AS res;

Result:

SELECT is_json('Not a JSON') AS res;

Result:

The function checks a string variable for containing a valid JSONB.

SELECT is_jsonb('{"review": {"date": "1970-12-30", "votes": 10, "rating": 5, "helpful_votes": 0}, "product": {"id": "1551803542", "group": "Book", "title": "Start and Run a Coffee Bar (Start & Run a)", "category": "Business & Investing", "sales_rank": 11611, "similar_ids": ["0471136174", "0910627312", "047112138X", "0786883561", "0201570483"], "subcategory": "General"}, "customer_id": "AE22YDHSBFYIP"}') AS res;

Result:

SELECT is_jsonb('Not a JSONB') AS res;

Result:

The function checks a string variable for being either, NULL or ''.

SELECT is_empty('abc') AS res;
-- Result is false

Result:

SELECT is_empty('') AS res;
-- Result is true

Result:

SELECT is_empty(NULL) AS res;
-- Result is true

Result:

The function checks a string variable for being a hexadecimal number being a bigint.

❗This function needs to have hex2bigint being installed!❗

If you use the package, both functions are installed in the correct sort order.

SELECT is_hex('a1b0') AS res;
-- Result is true

Result:

SELECT is_hex('a1b0w') AS res;
-- Result is false

Result:

SELECT is_hex('a1b0c3c3c3c4b5d3') AS res;
-- Result is false (does not fit into a bigint)

Result:

Creates a function which returns a SHA256 hash for the given string.
The parameter has to be converted into a binary string of bytea. :heavy_exclamation_mark:The function needs the pgcrypto package❗

❗This function has an external dependency and is only installed, if the package pgcrypto is installed❗

SELECT sha256('test-string'::bytea) AS res;

Result:

The function returns the size for schema given as parameter in bytes.

-- Returns the size of the schema public in bytes
SELECT pg_schema_size('public');

Result:

-- Returns the size of the schema public formatted
SELECT pg_size_pretty(pg_schema_size('public'));

Result:

Creates a view to get all views of the current database but excluding system views and all views which do start with "pg" or "_pg".

SELECT *
FROM pg_db_views;

Creates a view to get a list of foreign keys in the database. That includes the check for an existing single index, see boolean result of column "is_indexed".

Below PostgreSQL 11 the column "enforced" is not available and therefore not part of the result.

SELECT *
FROM pg_foreign_keys;

Creates a view to get all functions of the current database, excluding those in the schema pg_catalog and information_schema.

As there have been changes to the system tables used in this view, there are now two scripts dependend on the PostgreSQL version on which it has to be used, one for PostgreSQL 11 or newer and one for PostgreSQL 10 or older. This is handled in the script that creates the view.

SELECT *
FROM pg_functions;

Creates a view to view all live locks with all necessary information about the connections and the query.
The view needs PostgreSQL 9.2 as minimum version. The column application_name was added in 9.2.

SELECT *
FROM pg_active_locks;

Result:

Creates a view with information about the size of the table/materialized view and sizes of indexes on that table/materialized view. It does also list all indexes on that table in an array.

SELECT *
FROM pg_table_matview_infos;

Result:

Creates a view with information about the ownership of objects. Since PostgreSQL 11 supports procedures, therefore there is one version vor PostgreSQL 10. This view is supported in PostgreSQL 10 or newer. Older versions are not supported.

SELECT *
FROM pg_object_ownership
WHERE owner = 'stefanie';

Result:

Creates a view to get information about partitioned tables. Since PostgreSQL 10 supports partitions, but they became usable only in later versions. A system table to identify partitions has been added in PostgreSQL 11. Therefore this view is only available on systems with PostgreSQL 11 or newer.

SELECT *
FROM pg_partitioned_tables_infos;

Result:

The function checks if all characters are in included in a given encoding. That is especially useful, if you have to deal with exports into other encodings than the database encoding.

The function with two parameters uses UTF-8 as source encoding,
The one with three parameters uses the third parameter as source encoding.

SELECT is_encoding('Some characters', 'LATIN1') AS res;

Result:

SELECT is_encoding('Some characters, ğ is Turkish and not latin1', 'LATIN1') AS res;
-- Returns false: The Turkish character ğ is not part of latin1

Result:

SELECT is_encoding('Some characters', 'LATIN1', 'UTF8') AS res;

Result:

SELECT is_encoding('Some characters, ğ is Turkish and not latin1', 'UTF8', 'LATIN1') AS res;
-- Returns false: The Turkish character ğ is not part of latin1

Result:

The function is a shortcut for is_encoding('Some characters', 'LATIN1'), you don't have to give the target encoding.

SELECT is_latin1('Some characters') AS res;

Result:

SELECT is_latin1('Some characters, ğ is Turkish and not latin1') AS res;
-- Returns false: The Turkish character ğ is not part of latin1

Result:

The function returns a distinct array containing all characters, which are not defined in latin1.
The function depends on is_latin1 which is part of this repository.

-- Returns an array containing the characters ğ and Ƶ each one time
SELECT return_not_part_of_latin1('ağbƵcğeƵ') AS res;

Result:

The function has three implementations. All implementations depend on the function is_latin1, the function is included in this repository.

The function takes one parameter with characters to be checked and replaced with an empty string, if they are not part of latin1.

SELECT replace_latin1('Some characters, ğ is Turkish and not latin1') AS res;

Result:

The function takes a second parameter which is used to replace all characters, which are not part of latin1.

SELECT replace_latin1(
  'Some characters, ğ is Turkish and not latin1 and replaced with a g',
  'g'
) AS res;

Result:

The function takes as first parameter a string which may or may not have none latin1 characters. The second parameter is an arrays containing all characters, that should be replaced. The third parameter is an array, too. The characters defined in s_search are replaced with the characters in s_replace, it takes the position in the array to identify which character should be replaced by which one.

-- First identify the characters which should be replaced, which are {ğ,Ƶ}
SELECT return_not_part_of_latin1('ağbƵcğeƵ') AS res;

-- The ğ will be replaced whit a g and the Ƶ with a Z}
SELECT 'ağbƵcğeƵ' AS original
  , replace_latin1(
      'ağbƵcğeƵ',
      string_to_array('ğ,Ƶ', ','),
      string_to_array('g,Z', ',')
    ) AS res;

Result:

The function returns a distinct array containing all characters, which are not defined in the second parameter as encoding.
The function depends on is_encoding which is part of this repository.

-- Returns an array containing the characters ğ and Ƶ each one time
SELECT return_not_part_of_encoding('ağbƵcğeƵ', 'latin1') AS res;

Result:

The function has three implementations. All implementations depend on the function is_encoding, the function is included in this repository.

The function takes one parameter with characters to be checked and replaced with an empty string, if they are not part of the encoding given in the second parameter.

SELECT replace_encoding(
  'Some characters, ğ is Turkish and not latin1',
  'latin1'
) AS res;

Result:

The function takes a third parameter which is used to replace all characters which are not part of the encoding given in parameter 2.

SELECT replace_encoding(
  'Some characters, ğ is Turkish and not latin1 and replaced with a g',
  'latin1',
  'g'
) AS res;

Result:

The function takes as first parameter a string which may or may not have none latin1 characters. The second parameter is an arrays containing all characters, that should be replaced. The third parameter is an array, too. The characters defined in s_search are replaced with the characters in s_replace, it takes the position in the array to identify which character should be replaced by which one.

-- First identify the characters which should be replaced, which are {ğ,Ƶ}
SELECT return_not_part_of_latin1('ağbƵcğeƵ') AS res;

-- The ğ will be replaced whit a g and the Ƶ with a Z}
SELECT 'ağbƵcğeƵ' AS original
  , replace_encoding(
      'ağbƵcğeƵ',
      string_to_array('ğ,Ƶ', ','),
      string_to_array('g,Z', ',')
  ) AS res;

Result:

The aggregate is used in Window Functions to show the last value in case the current value is null.

BEGIN;

CREATE TABLE test_gap_fill(id INTEGER, some_value text);

INSERT INTO test_gap_fill(id, some_value) VALUES
  (1, 'value 1'),
  (1, NULL),
  (2, 'value 2'),
  (2, NULL),
  (2, NULL),
  (3, 'value 3')
;

SELECT id
  , some_value
FROM test_gap_fill
;

ROLLBACK;

Result:

BEGIN;

CREATE TABLE test_gap_fill(id INTEGER, some_value text);

INSERT INTO test_gap_fill(id, some_value) VALUES
  (1, 'value 1'),
  (1, NULL),
  (2, 'value 2'),
  (2, NULL),
  (2, NULL),
  (3, 'value 3')
;

-- Fill the empty rows with values
SELECT id
  , gap_fill(some_value) OVER (ORDER BY id) AS some_value
FROM test_gap_fill
;

ROLLBACK;

Result:

Calculate minimum values from arrays.

Supported data types are SMALLINT, INTEGER, BIGINT, REAL, DOUBLE PRECISION, NUMERIC, and TEXT;

SELECT array_min(ARRAY[45, 60, 43, 99]::SMALLINT[]);

Result:

SELECT array_min(ARRAY[45, 60, 43, 99]::INTEGER[]);

Result:

SELECT array_min(ARRAY[45, 60, 43, 99]::BIGINT[]);

Result:

SELECT array_min(ARRAY[45.6, 60.8, 43.7, 99.3]::REAL[]);

Result:

SELECT array_min(ARRAY[45.6, 60.8, 43.7, 99.3]::DOUBLE PRECISION[]);

Result:

SELECT array_min(ARRAY[45.6, 60.8, 43.7, 99.3]::NUMERIC[]);

Result:

SELECT array_min(ARRAY['def', 'abc', 'ghi']::TEXT[]);

Result:

Calculate minimum values from arrays.

Supported data types are SMALLINT, INTEGER, BIGINT, REAL, DOUBLE PRECISION, NUMERIC, and TEXT;

SELECT array_max(ARRAY[45, 60, 43, 99]::SMALLINT[]);

Result:

SELECT array_max(ARRAY[45, 60, 43, 99]::INTEGER[]);

Result:

SELECT array_max(ARRAY[45, 60, 43, 99]::BIGINT[]);

Result:

SELECT array_max(ARRAY[45.6, 60.8, 43, 99.3]::REAL[]);

Result:

SELECT array_max(ARRAY[45.6, 60.8, 43, 99.3]::DOUBLE PRECISION[]);

Result:

SELECT array_max(ARRAY[45.6, 60.8, 43, 99.3]::NUMERIC[]);

Result:

SELECT array_max(ARRAY['def', 'abc', 'ghi']::TEXT[]);

Result:

Calculate average values from arrays.

Supported data types are SMALLINT, INTEGER, BIGINT, REAL, DOUBLE PRECISION, and NUMERIC;

SELECT array_avg(ARRAY[45, 60, 43, 99]::SMALLINT[]);

Result:

SELECT array_avg(ARRAY[45, 60, 43, 99]::INTEGER[]);

Result:

SELECT array_avg(ARRAY[45, 60, 43, 99]::BIGINT[]);

Result:

SELECT array_avg(ARRAY[45.6, 60.8, 43, 99.3]::REAL[]);

Result:

SELECT array_avg(ARRAY[45.6, 60.8, 43, 99.3]::DOUBLE PRECISION[]);

Result:

SELECT array_avg(ARRAY[45.6, 60.8, 43, 99.3]::NUMERIC[]);

Result:

Calculate sum of values from arrays.

Supported data types are SMALLINT, INTEGER, BIGINT, REAL, DOUBLE PRECISION, and, NUMERIC;

SELECT array_sum(ARRAY[45, 60, 43, 99]::SMALLINT[]);

Result:

SELECT array_sum(ARRAY[45, 60, 43, 99]::INTEGER[]);

Result:

SELECT array_sum(ARRAY[45, 60, 43, 99]::BIGINT[]);

Result:

SELECT array_sum(ARRAY[45.6, 60.8, 43.7, 99.3]::REAL[]);

Result:

SELECT array_sum(ARRAY[45.6, 60.8, 43.7, 99.3]::DOUBLE PRECISION[]);

Result:

SELECT array_sum(ARRAY[45.6, 60.8, 43.7, 99.3]::NUMERIC[]);

Result:

Several countries use different formats for numbers, dates, and timestamps. Therefore I needed some functions, which is easy to remember than the format codes which differ in different programming languages.

Creates a function which returns the given date in German format.

SELECT date_de('2018-01-01') AS d_de;

Result:

Creates a function which returns the given timestamp in German format.

SELECT datetime_de('2018-01-01 13:30:30 GMT') AS ts_de;

Result:

Creates two functions which returns unix timestamp for the a given timestamp or a given timestamp with time zone.

-- Timestamp without time zone, server uses German / Berlin time zone
SELECT to_unix_timestamp('2018-01-01 00:00:00') AS unix_timestamp;

Result:

-- Timestamp with time zone
SELECT to_unix_timestamp('2018-01-01 00:00:00+01') AS unix_timestamp;

Result:

Creates a functions which returns a hexadecimal number given as text as bigint.

SELECT hex2bigint('a1b0') AS hex_as_bigint;

Result:

Removes empty strings and null entries from a given array. In addition the function can remove duplicate entries. The function supports strings, numbers, dates, and timestamps with or without time zone.

-- Untrimmed timestamp array with time zone with duplicates
SELECT array_trim(ARRAY['2018-11-11 11:00:00 MEZ',NULL,'2018-11-11 11:00:00 MEZ']::TIMESTAMP WITH TIME ZONE[]) AS trimmed_array;

Result:

-- Timestamp array with time zone with duplicates
SELECT ARRAY['2018-11-11 11:00:00 MEZ',NULL,'2018-11-11 11:00:00 MEZ']::TIMESTAMP WITH TIME ZONE[] AS untrimmed_array;

Result:

-- Timestamp array with time zone without duplicates
SELECT array_trim(ARRAY['2018-11-11 11:00:00 MEZ',NULL,'2018-11-11 11:00:00 MEZ']::TIMESTAMP WITH TIME ZONE[], TRUE) AS trimmed_array_distinct;

Result: