PL/R User’s Guide - R Procedural Language
2. Installation
3. Functions and Arguments
4. Passing Data Values
5. Using Global Data
6. Database Access and Support Functions
6.1. Normal Support
6.2. RPostgreSQL Compatibility Support
7. PostgreSQL Support Functions
8. Aggregate Functions
9. Window Functions
10. Loading R Modules at Startup
11. R Function Names
12. Trigger Procedures
13. License
Overview
PL/R is a loadable procedural language that enables you to write PostgreSQL functions and triggers in the
R programming language1. PL/R offers most (if not all) of the capabilities a function writer has in the R
language.
Commands are available to access the database via the PostgreSQL Server Programming Interface (SPI)
and to raise messages via elog()
. There is no way to access internals of the database backend. However
the user is able to gain OS-level access under the permissions of the PostgreSQL user ID, as with a C
function. Thus, any unprivileged database user should not be permitted to use this language. It must be
installed as an untrusted procedural language so that only database superusers can create functions in it.
The writer of a PL/R function must take care that the function cannot be used to do anything unwanted, since it will be able to do anything that could be done by a user logged in as the database administrator. An implementation restriction is that PL/R procedures cannot be used to create input/output functions for new data types.
Installation
If you are going to compile PostgreSQL from the source, use the following commands from the untared and unzipped file downloaded from http://www.postgresql.org/ftp/source/:
./configure --enable-R-shlib --prefix=/opt/postgres_plr && make && make install
Place source tar file in the contrib dir in the PostgreSQL source tree and untar it. The shared object for the R call handler is built and installed in the PostgreSQL library directory via the following commands (starting from /path/to/postgresql_source/contrib):
cd plr
make
make install
You may explicitly include the path of pg_config to PATH
, such as
cd plr
PATH=/usr/pgsql-9.4/bin/:$PATH; USE_PGXS=1 make
echo ’PATH=/usr/pgsql-9.4/bin/:$PATH; USE_PGXS=1 make install’ | sudo sh
If you want to use git to pull the repository, run the following command before the make command:
git clone https://github.com/postgres-plr/plr
As of PostgreSQL 8.0.0, PL/R can also be built without the PostgreSQL source tree. Untar PL/R where ever you prefer. The shared object for the R call handler is built and installed in the PostgreSQL library directory via the following commands (starting from/path/to/plr):
cd plr
USE_PGXS=1 make
USE_PGXS=1 make install
Win32 - adjust paths according to your own setup, and be sure to restart the PostgreSQL service after changing:
In Windows environment:
R_HOME=C:\Progra~1\R\R-3.5.1
Path=%PATH%;%R_HOME\x64\bin
In MSYS:
export R_HOME=/c/progra~1/R/R-2.5.
export PATH=$PATH:/c/progra~1/PostgreSQL/8.2/bin
USE_PGXS=1 make
USE_PGXS=1 make install
You can use plr.sql
(which is created in contrib/plr) to create the language and support functions
in your database of choice:
psql mydatabase < plr.sql
Alternatively you can create the language manually using SQL commands:
CREATE FUNCTION plr_call_handler()
RETURNS LANGUAGE_HANDLER
AS ’$libdir/plr’ LANGUAGE C;
CREATE LANGUAGE plr HANDLER plr_call_handler;
As of PostgreSQL 9.1 you can use the new CREATE EXTENSION
command:
CREATE EXTENSION plr;
This is not only simple, it has the added advantage of tracking all PL/R installed objects as dependent on the extension, and therefore they can be removed just as easily if desired:
DROP EXTENSION plr;
Tip If a language is installed into template1
, all subsequently created databases will have the
language installed automatically.
Tip In addition to the documentation, the plr.out.* files in plr/expected are a good source of usage examples.
Tip R headers are required. Download and install R prior to building PL/R. R must have been built
with the --enable-R-shlib
option when it was configured, in order for the libR shared object library
to be available.
Tip: Additionally, libR must be findable by your runtime linker. On Linux, this involves adding an entry in /etc/ld.so.conf for the location of libR (typically $R_HOME/bin or $R_HOME/lib), and then running ldconfig. Refer toman ldconfigor its equivalent for your system.
Tip: R_HOME must be defined in the environment of the user under which PostgreSQL is started, beforethe postmaster is started. Otherwise PL/R will refuse to load. See plr_environ(), which allows examination of the environment available to the PostgreSQL postmaster process.
Tip: On Win32 platform, R will consider a registry entry created by R installer if it fails to find R_HOME environment variable. If you accepted installer defaults, there is no need to set R_HOME on this platform. Be careful removing older version of R as it may take away InstallPath entry away from HKLM/SOFTWARE/R-core/R.
Functions and Arguments
To create a function in the PL/R language, use standard R syntax, but without the enclosing braces or
function assignment. Instead of myfunc <- function(arguments) { function body }
, the body
of your PL/R function is just sqlfunction body
CREATE OR REPLACE FUNCTIONfuncname(argument-types)
RETURNSreturn-typeAS ’
function body
’ LANGUAGE ’plr’;
The body of the function is simply a piece of R script. When the function is called, the argument values
are passed as variables arg1...argN
to the R script. The result is returned from the R code in the usual
way. For example, a function returning the greater of two integer values could be defined as:
CREATE OR REPLACE FUNCTION r_max (integer, integer) RETURNS integer AS ’
if (arg1 > arg2)
return(arg1)
else
return(arg2)
’ LANGUAGE ’plr’ STRICT;
Starting with PostgreSQL 8.0, arguments may be explicitly named when creating a function. If an argument
is explicitly named at function creation time, that name will be available to your R script in place of
the usual argNvariable
. For example:
CREATE OR REPLACE FUNCTION sd(vals float8[]) RETURNS float AS ’
sd(vals)
’ LANGUAGE ’plr’ STRICT;
Starting with PostgreSQL 8.4, a PL/R function may be declared to be a WINDOW
. In this case, in addition to
the usual argN(or named)
variables, PL/R automatically creates several other arguments to your function.
For each explicit argument, a corresponding variable called farg1...fargN
is passed to the R script.
These contain an R vector of all the values of the related argument for the moving WINDOW
frame within
the current PARTITION
. For example:
CREATE OR REPLACE
FUNCTION r_regr_slope(float8, float8)
RETURNS float8 AS
$BODY$
slope <- NA
y <- farg
x <- farg
if (fnumrows==9) try (slope <- lm(y ~ x)$coefficients[2])
return(slope)
LANGUAGE plr WINDOW;
In the preceding example,farg1
and farg2
are R vectors containing the current row’s data plus that of
related rows. The determination as to which rows qualify as related is determined by the frame specification
of the query at run time. The example also illustrates one of two additional autogenerated arguments.
fnumrows
is the number of rows in the current WINDOW
frame. The other (not shown) auto-argument
is called prownum
. This argument provides the 1-based row offset of the current row in the current
PARTITION
. See Window Functions for more information and a more complete example.
In some of the the definitions above, note the clause STRICT
, which saves us from having to think about
NULL input values: if a NULL is passed, the function will not be called at all, but will just return a
NULL result automatically. In a non-strict function, if the actual value of an argument is NULL, the
corresponding argN
variable will be set to a NULLR
object. For example, suppose that we wanted r_max
with one null and one non-null argument to return the non-null argument, rather than NULL:
CREATE OR REPLACE FUNCTION r_max (integer, integer) RETURNS integer AS ’
if (is.null(arg1) && is.null(arg2))
return(NULL)
if (is.null(arg1))
return(arg2)
if (is.null(arg2))
return(arg1)
if (arg1 > arg2)
return(arg1)
arg
’ LANGUAGE ’plr’;
As shown above, to return a NULL value from a PL/R function, return NULL
. This can be done whether
the function is strict or not.
Composite-type (tuple) arguments are passed to the procedure as R data.frames. The element names of
the frame are the attribute names of the composite type. If an attribute in the passed row has the NULL
value, it will appear as an “NA” in the frame. Here is an example:
CREATE TABLE emp (name text, age int, salary numeric(10,2));
INSERT INTO emp VALUES (’Joe’, 41, 250000.00);
INSERT INTO emp VALUES (’Jim’, 25, 120000.00);
INSERT INTO emp VALUES (’Jon’, 35, 50000.00);
CREATE OR REPLACE FUNCTION overpaid (emp) RETURNS bool AS ’
if (200000 < arg1$salary) {
return(TRUE)
}
if (arg1$age < 30 && 100000 < arg1$salary) {
return(TRUE)
}
return(FALSE)
’ LANGUAGE ’plr’;
SELECT name, overpaid(emp) FROM emp;
name | overpaid
------+----------
Joe | t
Jim | t
Jon | f
(3 rows)
There is also support for returning a composite-type result value:
CREATE OR REPLACE FUNCTION get_emps() RETURNS SETOF emp AS ’
names <- c("Joe","Jim","Jon")
ages <- c(41,25,35)
salaries <- c(250000,120000,50000)
df <- data.frame(name = names, age = ages, salary = salaries)
return(df)
’ LANGUAGE ’plr’;
select* from get_emps();
name | age | salary
------+-----+-----------
Jim | 41 | 250000.
Joe | 25 | 120000.
Jon | 35 | 50000.
(3 rows)
An alternative method may be used to create a function in PL/R, if certain criteria are met. First, the function must be a simple call to an existing R function. Second, the function name used for the PL/R function must match that of the R function exactly. If these two criteria are met, the PL/R function may be defined with no body, and the arguments will be passed directly to the R function of the same name.
For example:
create or replace function sd(_float8) returns float as '' language ’plr’;
select round(sd(’{1.23,1.31,1.42,1.27}’::_float8)::numeric,8);
round
------------
0.
(1 row)
Tip Because the function body is passed as an SQL string literal to CREATE FUNCTION
, you have to
escape single quotes and backslashes within your R source, typically by doubling them.
Passing Data Values
The argument values supplied to a PL/R function’s script are the input arguments converted to a corresponding R form. See Table 4-1. Scalar PostgreSQL values become single element R vectors. One exception to this are scalar bytea values. These are first converted to R raw type, and then processed by the R unserialize command. One-dimensional PostgreSQL arrays are converted to multi-element R vectors, two-dimensional PostgreSQL arrays are mapped to R matrixes, and three-dimensional PostgreSQL arrays are converted to three-dimensional R arrays. Greater than three-dimensional arrays are not supported. Composite-types are transformed into R data.frames.
Table 4-1. Function Arguments
PostgreSQL type | R type |
---|---|
boolean | logical |
int2,int4 | integer |
int8,float4,float8,cash,numeric | numeric |
bytea | object |
everything else | character |
Conversely, the return values are first coerced to R character, and therefore anything that resolves to a string that is acceptable input format for the function’s declared return type will produce a result. Again, there is an exception for scalar bytea return values. In this case, the R object being returned is first processed by the R serialize command, and then the binary result is directly mapped into a PostgreSQL bytea datum. Similar to argument conversion, there is also a mapping between the dimensionality of the declared PostgreSQL return type and the type of R object. That mapping is shown in Table 4-2
Table 4-2. Function Result Dimensionality
PgSQL return type | R type | Result | Example |
---|---|---|---|
scalar | array,matrix,vector | first column of first row | c(1,2,3) in R returns 1 in PostgreSQL |
setof scalar | 1D array,greater than 2D array, vector | multi-row, 1 column set | array(1:10) in R returns 10 rows in PostgreSQL |
scalar | data.frame | textual representation of the first column’s vector | data.frame(c(1,2,3)) in R returns ’c(1, 2, 3)’ |
setof scalar | 2D array,matrix,data.frame | #columns > 1, error; #columns == 1,multi-row, 1 column set | (as.data.frame(array(1:10,c(2,5))))[,1] in R returns 2 rows of scalar |
array | 1D array,greater than 3D array,vector | 1D array | array(1:8,c(2,2,2,2)) in R returns {1,2,3,4,5,6,7,8,1,2,3,4,5,6,7,8} |
array | 2D array,matrix,data.frame | 2D array | array(1:4,c(2,2)) in R returns {{1,3},{2,4}} |
array | 3D array | 3D array | array(1:8,c(2,2,2)) in R returns {{{1,5},{3,7}},{{2,6},{4,8}}} |
composite | 1D array,greater than 2D array,vector | first row, 1 column | array(1:8,c(2,2,2)) in R returns 1 row of scalar |
setof composite | 1D array,greater than 2D array,vector | multi-row, 1 column set | array(1:8,c(2,2,2)) in R returns 8 rows of scalar |
composite | 2D array,matrix,data.frame | first row, multi-column | array(1:4,c(2,2)) in R returns 1 row of 2 columns |
setof composite | 2D array,matrix,data.frame | multi-row, multi-column set | array(1:4,c(2,2)) in R returns 2 rows of 2 columns |
Using Global Data
Sometimes it is useful to have some global status data that is held between two calls to a procedure or
is shared between different procedures. Equally useful is the ability to create functions that your PL/R
functions can share. This is easily done since all PL/R procedures executed in one backend share the same
R interpreter. So, any global R variable is accessible to all PL/R procedure calls, and will persist for the
duration of the SQL client connection. An example of using a global object appears in the pg.spi.execp
example, in Database Access and Support Functions.
A globally available, user named, R function (the R function name of PL/R functions is not the same as its
PostgreSQL function name; see: R Function Names) can be created dynamically using the provided PostgreSQL
function install_rcmd(text)
. Here is an example:
select install_rcmd(’pg.test.install <-function(msg) {print(msg)}’);
install_rcmd
--------------
OK
(1 row)
create or replace function pg_test_install(text) returns text as ’
pg.test.install(arg1)
’ language ’plr’;
select pg_test_install(’hello world’);
pg_test_install
-----------------
hello world
(1 row)
A globally available, user named, R function can also be automatically created and installed in the R
interpreter. See: Loading R Modules at Startup PL/R also provides a global variable called
pg.state.firstpass.
This variable is reset to TRUE the first time each PL/R
function is called,
for a particular query.
On subsequent calls the value is left unchanged. This allows one or more PL/R functions to perform a
possibly expensive initialization on the first call, and reuse the results for the remaining rows in
the query.
For example:
create table t (f1 int);
insert into t values (1);
insert into t values (2);
insert into t values (3);
create or replace function f1() returns int as ’
msg <- paste("enter f1, pg.state.firstpass is", pg.state.firstpass)
pg.thrownotice(msg)
if (pg.state.firstpass == TRUE)
pg.state.firstpass<<- FALSE
msg <- paste("exit f1, pg.state.firstpass is", pg.state.firstpass)
pg.thrownotice(msg)
return(0)
’ language plr;
create or replace function f2() returns int as ’
msg <- paste("enter f2, pg.state.firstpass is", pg.state.firstpass)
pg.thrownotice(msg)
if (pg.state.firstpass == TRUE)
pg.state.firstpass<<- FALSE
msg <- paste("exit f2, pg.state.firstpass is", pg.state.firstpass)
pg.thrownotice(msg)
return(0)
’ language plr;
select f1(), f2(), f1 from t;
NOTICE: enter f1, pg.state.firstpass is TRUE
NOTICE: exit f1, pg.state.firstpass is FALSE
NOTICE: enter f2, pg.state.firstpass is TRUE
NOTICE: exit f2, pg.state.firstpass is FALSE
NOTICE: enter f1, pg.state.firstpass is FALSE
NOTICE: exit f1, pg.state.firstpass is FALSE
NOTICE: enter f2, pg.state.firstpass is FALSE
NOTICE: exit f2, pg.state.firstpass is FALSE
NOTICE: enter f1, pg.state.firstpass is FALSE
NOTICE: exit f1, pg.state.firstpass is FALSE
NOTICE: enter f2, pg.state.firstpass is FALSE
NOTICE: exit f2, pg.state.firstpass is FALSE
f1 | f2 | f
----+----+----
0 | 0 | 1
0 | 0 | 2
0 | 0 | 3
(3 rows)
create or replace function row_number() returns int as ’
if (pg.state.firstpass)
{
assign("pg.state.firstpass", FALSE, env=.GlobalEnv)
lclcntr<- 1
}
else
lclcntr<- plrcounter + 1
assign("plrcounter", lclcntr, env=.GlobalEnv)
return(lclcntr)
’ language ’plr’;
SELECT row_number(), f1 from t;
row_number | f
------------+----
1 | 1
2 | 2
3 | 3
(3 rows)
Database Access and Support
Functions
The following commands are available to access the database from the body of a PL/R procedure, or in support thereof:
Normal Support
pg.spi.exec(character query)
Execute an SQL query given as a string. An error in the query causes an error to be raised. Otherwise,
the command’s return value is the number of rows processed for INSERT
,UPDATE
, or DELETE
statements, or zero if the query is a utility statement. If the query is a SELECT
statement, the values of
the selected columns are placed in an R data.frame with the target column names used as the frame
column names. However, non-numeric columns are not converted to factors. If you want all non-
numeric columns converted to factors, a convenience function pg.spi.factor
(described below)
is provided.
If a field of a SELECT result is NULL, the target variable for it is set to “NA”. For example:
create or replace function test_spi_tup(text) returns setof record as ’
pg.spi.exec(arg1)
’ language ’plr’;
select * from test_spi_tup(’select oid, NULL::text as nullcol,
typname from pg_type where typname = ”oid” or typname = ”text”’)
as t(typeid oid, nullcol text, typename name);
typeid | nullcol | typename
--------+---------+----------
25 | | text
26 | | oid
(2 rows)
The NULL values were passed to R as “NA”, and on return to PostgreSQL they were converted back to NULL.
pg.spi.prepare(character query,integer vector type_vector)
Prepares and saves a query plan for later execution. The saved plan will be retained for the life of
the current backend.
The query may use arguments, which are placeholders for values to be supplied whenever the plan is
actually executed. In the query string, refer to arguments by the symbols $1...$n
. If the query uses
arguments, the values of the argument types must be given as a vector. Pass NA
for type_vector
if the query has no arguments. The argument types must be identified by the type Oids, shown in
pg_type. Global variables are provided for this use. They are named according to the convention
TYPENAMEOID, where the actual name of the type, in all capitals, is substituted for TYPENAME.
A support function, load_r_typenames()
must be used to make the predefined global variables
available for use:
select load_r_typenames();
load_r_typenames
------------------
OK
(1 row)
Another support function,r_typenames()
may be used to list the predefined Global variables:
select * from r_typenames();
typename | typeoid
-----------------+---------
ABSTIMEOID | 702
ACLITEMOID | 1033
ANYARRAYOID | 2277
ANYOID | 2276
BITOID | 1560
BOOLOID | 16
[...]
TRIGGEROID | 2279
UNKNOWNOID | 705
VARBITOID | 1562
VARCHAROID | 1043
VOIDOID | 2278
XIDOID | 28
(59 rows)
The return value from pg.spi.prepare
is a query ID to be used in subsequent calls to
pg.spi.execp. See spi_execp
for an example.
pg.spi.execp(external pointer saved_plan,variable listvalue_list)
Execute a query previously prepared with pg.spi.prepare.saved_planis the external pointer
returned by pg.spi.prepare
. If the query references arguments, a value_list
must be supplied:
this is an R list of actual values for the plan arguments. It must be the same length as the argument
type_vector previously given to pg.spi.prepare. Pass NA
for value_list
if the query has
no arguments. The following illustrates the use of pg.spi.prepare
and pg.spi.execp
with and
without query arguments:
create or replace function test_spi_prep(text) returns text as ’
sp <<- pg.spi.prepare(arg1, c(NAMEOID, NAMEOID));
print("OK")
’ language ’plr’;
select test_spi_prep(’select oid, typname from pg_type
where typname = $1 or typname = $2’);
test_spi_prep
---------------
OK
(1 row)
create or replace function test_spi_execp(text, text, text) returns setof record as ’
pg.spi.execp(pg.reval(arg1), list(arg2,arg3))
’ language ’plr’;
select * from test_spi_execp(’sp’,’oid’,’text’) as t(typeid oid, typename name);
typeid | typename
--------+----------
25 | text
26 | oid
(2 rows)
create or replace function test_spi_prep(text) returns text as ’
sp <<- pg.spi.prepare(arg1, NA);
print("OK")
’ language ’plr’;
select test_spi_prep(’select oid, typname from pg_type
where typname = ”bytea” or typname = ”text”’);
test_spi_prep
---------------
OK
(1 row)
create or replace function test_spi_execp(text) returns setof record as ’
pg.spi.execp(pg.reval(arg1), NA)
’ language ’plr’;
select * from test_spi_execp(’sp’) as t(typeid oid, typename name);
typeid | typename
--------+----------
17 | bytea
25 | text
(2 rows)
create or replace function test_spi_prep(text) returns text as ’
sp <<- pg.spi.prepare(arg1);
print("OK")
’ language ’plr’;
select test_spi_prep(’select oid, typname from pg_type
where typname = ”bytea” or typname = ”text”’);
test_spi_prep
---------------
OK
(1 row)
create or replace function test_spi_execp(text) returns setof record as ’
pg.spi.execp(pg.reval(arg1))
’ language ’plr’;
select * from test_spi_execp(’sp’) as t(typeid oid, typename name);
typeid | typename
--------+----------
17 | bytea
25 | text
(2 rows)
NULL arguments should be passed as individual NA
values in value_list.
Except for the way in which the query and its arguments are specified,pg.spi.execp
works just
like pg.spi.exec
.
pg.spi.cursor_open( character cursor_name,external pointer saved_plan,variable list value_list)
Opens a cursor identified by cursor_name. The cursor can then be used to scroll through the results
of a query plan previously prepared by pg.spi.prepare. Any arguments to the plan should be specified
in arg values similar to pg.spi.execp
. Only read-only cursors are supported at the moment.
plan <- pg.spi.prepare(’SELECT * FROM pg_class’);
cursor_obj <- pg.spi.cursor_open(’my_cursor’,plan);
Returns a cursor object that be be passed to pg.spi.cursor_fetch
pg.spi.cursor_fetch(external pointer cursor,boolean forward,integer rows)
Fetches rows from the cursor object previously returned by pg.spi.cursor_open
. If forward is
TRUE then the cursor is moved forward to fetch at most the number of rows required by the rows
parameter. If forward is FALSE then the cursor is moved backwards at most the number of rows
specified. rows indicates the maximum number of rows that should be returned.
plan <- pg.spi.prepare(’SELECT * FROM pg_class’);
cursor_obj <- pg.spi.cursor_open(’my_cursor’,plan);
data <- pg.spi.cursor_fetch(cursor_obj,TRUE,as.integer(10));
Returns a data frame containing the results.
pg.spi.cursor_close(external pointer cursor)
Closes a cursor previously opened by pg.spi.cursor_open
plan <- pg.spi.prepare(’SELECT * FROM pg_class’);
cursor_obj <- pg.spi.cursor_open(’my_cursor’,plan);
pg.spi.cursor_close(cursor_obj);
pg.spi.lastoid()
Returns the OID of the row inserted by the last query executed via pg.spi.exec
or pg.spi.execp
,
if that query was a single-row INSERT. (If not, you get zero.)
pg.quoteliteral(character SQL_string)
Duplicates all occurrences of single quote and backslash characters in the given string. This may
be used to safely quote strings that are to be inserted into SQL queries given to pg.spi.exec
or
pg.spi.prepare
.
pg.quoteident(character SQL_string)
Return the given string suitably quoted to be used as an identifier in an SQL query string. Quotes
are added only if necessary (i.e., if the string contains non-identifier characters or would be case folded). Embedded quotes are properly doubled. This may be used to safely quote strings that are to
be inserted into SQL queries given to pg.spi.exec
or pg.spi.prepare
.
pg.thrownotice(character message)
pg.throwerror(character message)
Emit a PostgreSQL NOTICE
or ERROR
message.ERROR
also raises an error condition:
further execution of the function is abandoned, and the current transaction is aborted.
pg.spi.factor(data.frame data)
Accepts an R data.frame
as input, and converts all non-numeric columns to factors
. This may
be useful for data.frames
produced by pg.spi.exec
or pg.spi.prepare
, because the PL/R conversion
mechanism does not do that for you.
RPostgreSQL Compatibility Support
The following functions are intended to provide some level of compatibility between PL/R and RPostgreSQL (PostgreSQL DBI package). This allows, for example, a function to be first prototyped using an R client, and then easily moved to PL/R for production use.
dbDriver(character dvr_name)
dbConnect (DBIDriver drv, character user, character password, character host, character dbname,
character port,character tty,character options)
dbSendQuery(DBIConnection conn,character sql)
fetch(DBIResult rs,integer num_rows)
dbClearResult(DBIResult rs)
dbGetQuery(DBIConnection conn,character sql)
dbReadTable(DBIConnection conn,character name)
dbDisconnect(DBIConnection conn)
dbUnloadDriver(DBIDriver drv)
These functions nominally work like their RPostgreSQL counterparts except that all queries are performed in the current database. Therefore all driver and connection related parameters are ignored, and dbDriver, dbConnect, dbDisconnect, and dbUnloadDriver are no-ops.
PostgreSQL Support Functions
The following commands are available to use in PostgreSQL queries to aid in the use of PL/R functions:
plr_version()
Displays PL/R version as a text string.
install_rcmd(text R_code)
Install R code, given as a string, into the interpreter. See Using Global Data for an example.
reload_plr_modules()
Force re-loading of R code from the plr_modulestable. It is useful after modifying the contents of plr_modules, so that the change will have an immediate effect.
plr_singleton_array(float8 first_element)
Creates a new PostgreSQL array, using element first_element
. This function is predefined to
accept one float8 value and return a float8 array. The C function that implements this PostgreSQL
function is capable of accepting and returning other data types, although the return type must be an
array of the input parameter type. It can also accept multiple input parameters. For example, to define
a plr_array
function to create a text array from two input text values:
CREATE OR REPLACE FUNCTION plr_array (text, text)
RETURNS text[]
AS ’$libdir/plr’,’plr_array’
LANGUAGE ’C’ WITH (isstrict);
select plr_array(’hello’,’world’);
plr_array
---------------
{hello,world}
(1 row)
plr_array_push(float8[] array, float8 next_element)
Pushes a new element onto the end of an existing PostgreSQL array. This function is predefined to
accept one float8 array and a float8 value, and return a float8 array. The C function that implements
this PostgreSQL function is capable of accepting and returning other data types. For example, to
define a plr_array_push
function to add a text value to an existing text array:
CREATE OR REPLACE FUNCTION plr_array_push (_text, text)
RETURNS text[]
AS ’$libdir/plr’,’plr_array_push’
LANGUAGE ’C’ WITH (isstrict);
select plr_array_push(plr_array(’hello’,’world’), ’how are you’);
plr_array_push
-----------------------------
{hello,world,"how are you"}
(1 row)
plr_array_accum(float8[]state_value,float8next_element)
Creates a new array using next_element if state_value is NULL. Otherwise, pushes
next_element onto the end of state_value. This function is predefined to accept one
float8 array and a float8 value, and return a float8 array. The C function that implements this
PostgreSQL function is capable of accepting and returning other data types. For example, to define
a plr_array_accum
function to add an int4 value to an existing int4 array:
CREATE OR REPLACE FUNCTION plr_array_accum (_int4, int4)
RETURNS int4[]
AS ’$libdir/plr’,’plr_array_accum’
LANGUAGE ’C’;
select plr_array_accum(NULL, 42);
plr_array_accum
-------------
{42}
(1 row)
select plr_array_accum(’{23,35}’, 42);
plr_array_accum
-----------------
{23,35,42}
(1 row)
This function may be useful for creating custom aggregates. See Aggregate Functions for an example.
load_r_typenames()
Installs datatype Oid variables into the R interpreter as globals. See also r_typenames
below.
r_typenames()
Displays the datatype Oid variables installed into the R interpreter as globals. See Database Access and Support Functions for an example.
plr_environ()
Displays the environment under which the Postmaster is currently running. This may be useful to debug issues related to R specific environment variables. This function is installed with EXECUTE permission revoked from PUBLIC.
plr_set_display(text display)
Sets the DISPLAY environment variable under which the Postmaster is currently running. This may be useful if using R to plot to a virtual frame buffer. This function is installed with EXECUTE permission revoked from PUBLIC.
plr_get_raw(bytea serialized_object)
By default, when R objects are returned as type bytea
, the R object is serialized using an internal
R function prior to sending to PostgreSQL. This function deserializes the R object using another
internal R function, and returns the pure raw bytes to PostgreSQL. This is useful, for example, if the
R object being returned is a JPEG or PNG graphic for use outside of R.
Aggregate Functions
Aggregates in PostgreSQL are extensible via SQL commands. In general, to create a new aggregate, a
state transition function and possibly a final function are specified. The final function is used in case the
desired output of the aggregate is different from the data that needs to be kept in the running state value.
There is more than one way to create a new aggregate using PL/R. A simple aggregate can be defined
using the predefined PostgreSQL C function,plr_array_accum
(see PostgreSQL Support Functions) as a state transition
function, and a PL/R function as a finalizer. For example:
create or replace function r_median(_float8) returns float as ’
median(arg1)
’ language ’plr’;
CREATE AGGREGATE median (
sfunc = plr_array_accum,
basetype = float8,
stype = _float8,
finalfunc = r_median
);
create table foo(f0 int, f1 text, f2 float8);
insert into foo values(1,’cat1’,1.21);
insert into foo values(2,’cat1’,1.24);
insert into foo values(3,’cat1’,1.18);
insert into foo values(4,’cat1’,1.26);
insert into foo values(5,’cat1’,1.15);
insert into foo values(6,’cat2’,1.15);
insert into foo values(7,’cat2’,1.26);
insert into foo values(8,’cat2’,1.32);
insert into foo values(9,’cat2’,1.30);
select f1, median(f2) from foo group by f1 order by f1;
f1 | median
------+--------
cat1 | 1.21
cat2 | 1.28
(2 rows)
A more complex aggregate might be created by using a PL/R functions for both state transition and finalizer.
Window Functions
Starting with version 8.4, PostgreSQL supports WINDOW
functions which provide the ability to perform
calculations across sets of rows that are related to the current query row. This is comparable to the type
of calculation that can be done with an aggregate function. But unlike regular aggregate functions, use of
a window function does not cause rows to become grouped into a single output row; the rows retain their
separate identities. Behind the scenes, the window function is able to access more than just the current
row of the query result. See the PostgreSQL documentation for more general information related to the
use of this capability.
PL/R functions may be defined as WINDOW
. For example:
CREATE OR REPLACE FUNCTION r_regr_slope(float8, float8)
RETURNS float8 AS
$BODY$
slope <- NA
y <- farg1
x <- farg2
if (fnumrows==9) try (slope <- lm(y ~ x)$coefficients[2])
return(slope)
$BODY$
LANGUAGE plr WINDOW;
A number of variables are automatically provided by PL/R to the R interpreter:
fargN
farg1
and farg2
are R vectors containing the current row’s data plus that of the related rows.
fnumrows
The number of rows in the current WINDOW
frame.
prownum
(not shown)
Provides the 1-based row offset of the current row in the current PARTITION
.
A more complete example follows:
-- create test table
CREATE TABLE test_data (
fyear integer,
firm float8,
eps float8
);
-- insert randomly perturbated data for test
INSERT INTO test_data
SELECT (b.f + 1) % 10 + 2000 AS fyear,
floor((b.f+1)/10) + 50 AS firm,
f::float8/100 + random()/10 AS eps
FROM generate_series(-500,499,1) b(f);
CREATE OR REPLACE
FUNCTION r_regr_slope(float8, float8)
RETURNS float8 AS
$BODY$
slope <- NA
y <- farg1
x <- farg2
if (fnumrows==9) try (slope <- lm(y ~ x)$coefficients[2])
return(slope)
$BODY$
LANGUAGE plr WINDOW;
SELECT*, r_regr_slope(eps, lag_eps) OVER w AS slope_R
FROM (SELECT firm, fyear, eps,
lag(eps) OVER (ORDER BY firm, fyear) AS lag_eps
FROM test_data) AS a
WHERE eps IS NOT NULL
WINDOW w AS (ORDER BY firm, fyear ROWS 8 PRECEDING);
In this example, the variables farg1
and farg2
contain the current row value for eps and lag_eps, as well
as the preceding 8 rows which are also in the same WINDOW
frame within the same PARTITION
. In this
case since no PARTITION
is explicitly defined, the PARTITION
is the entire set of rows returned from the
inner sub-select.
Another interesting example follows. The idea of “Winsorizing” is to return either the original value or,
if that value is outside certain bounds, a trimmed value. So for example winsorize(eps, 0.1)
would return the
value at the 10th percentile for values of eps less that that, the value of the 90th percentile for eps greater
than that value, and the unmodified value of eps otherwise.
CREATE OR REPLACE FUNCTION winsorize(float8, float8)
RETURNS float8 AS
$BODY$
library(psych)
return(winsor(as.vector(farg1), arg2)[prownum])
$BODY$ LANGUAGE plr VOLATILE WINDOW;
SELECT fyear, eps,
winsorize(eps, 0.1) OVER (PARTITION BY fyear) AS w_eps
FROM test_data ORDER BY fyear, eps;
In this example, use of the variable prownum
is illustrated.
For optimization reasons, constant expressions are not expanded, and corresponding farg2
in the example above
is passes with NULL value for compatibility reasons not to shift other arguments in functions users created
with previous versions of PL/R.
If the frame is unbound and no row exclusion like below and in winsorize example,
window function calls are optimized as an entire partition is passed on the first call
and saved into a dedicated R environment that can be used to store user data.
Do not use .GlobalEnv
for that if you reuse you window function in the same query.
CREATE OR REPLACE FUNCTION PLR_ClusterDBSCAN(x double precision, y double precision, eps double precision, minpoints integer)
RETURNS int
AS $$
if (1 == prownum) {
c <- dbscan(cbind(farg1, farg2), eps=eps, MinPts=minpoints)
assign("cluster", c$cluster, env = parent.frame())
}
return(cluster[prownum])
$$ WINDOW LANGUAGE PLR;
SELECT x, y, PLR_ClusterDBSCAN(x, y, 300, 15) OVER() AS cid FROM mtl_crimes
Loading R Modules at Startup
PL/R has support for auto-loading R code during interpreter initialization. It uses a special table,
plr_modules
, which is presumed to contain modules of R code. If this table exists, the modules defined
are fetched from the table and loaded into the R interpreter immediately after creation.
The definition of the table plr_modules
is as follows:
CREATE TABLE plr_modules (
modseq int4,
modsrc text
);
The column modseq
is used to control the order of installation. The column modsrc
contains the full text of
the R code to be executed, including assignment if that is desired. Consider, for example, the following
statement:
INSERT INTO plr_modules
VALUES (0, ’pg.test.module.load <-function(msg) {print(msg)}’);
This statement will cause an R function namedpg.test.module.load to be created in the R interpreter on initialization. A PL/R function may now simply reference the function directly as follows:
create or replace function pg_test_module_load(text) returns text as ’
pg.test.module.load(arg1)
’ language ’plr’;
select pg_test_module_load(’hello world’);
pg_test_module_load
---------------------
hello world
(1 row)
The table plr_modules
must be readable by all, but it is wise to make it owned and writable only by the
database administrator.
R Function Names
In PostgreSQL, a function name can be used for different functions (overloaded) as long as the number
of arguments or their types differ. R, however, requires all function names to be distinct. PL/R deals with
this by constructing the internal R function names as a concatenation of the string “PLR” with the object
ID of the procedure’s pg_proc
. Thus, PostgreSQL functions with the same name and different argument
types will be different R functions too. This is not normally a concern for a PL/R programmer, but it might
be visible when debugging. If a specific, known, function name is needed so that an R function can be
referenced by one or more PL/R functions, the install_rcmd(text)
command can be used.
See Using Global Data.
Trigger Procedures
Trigger procedures can be written in PL/R. PostgreSQL requires that a procedure that is to be called as a
trigger must be declared as a function with no arguments and a return type of trigger
.
The information from the trigger manager is passed to the procedure body in the following variables:
pg.tg.name
The name of the trigger from the CREATE TRIGGER
statement.
pg.tg.relid
The object ID of the table that caused the trigger procedure to be invoked.
pg.tg.relname
The name of the table that caused the trigger procedure to be invoked.
pg.tg.when
The string BEFORE
or AFTER
depending on the type of trigger call.
pg.tg.level
The string ROW
or STATEMENT
depending on the type of trigger call.
pg.tg.op
The string INSERT
,UPDATE
, or DELETE
depending on the type of trigger call.
pg.tg.new
When the trigger is defined FOR EACH ROW
, a data.frame containing the values of the new table
row for INSERT
or UPDATE
actions. For triggers defined FOR EACH STATEMENT
and for DELETE
actions, set to NULL
. The attribute names are the table’s column names. Columns that are null will be
represented as NA
.
pg.tg.old
When the trigger is defined FOR EACH ROW
, a data.frame containing the values of the old table
row for DELETE
or UPDATE
actions. For triggers defined FOR EACH STATEMENT
and for INSERT
actions, set to NULL
. The attribute names are the table’s column names. Columns that are null will be
represented as NA
.
pg.tg.args
A vector of the arguments to the procedure as given in the CREATE TRIGGER
statement.
The return value from a trigger procedure can be NULL
or a one row data.frame matching the number
and type of columns in the trigger table. NULL
tells the trigger manager to silently suppress the operation
for this row. If a one row data.frame is returned, it tells PL/R to return a possibly modified row to the
trigger manager that will be inserted instead of the one given in pg.tg.new
. This works for INSERT
and
UPDATE
only. Needless to say that all this is only meaningful when the trigger is rBEFORE
and FOR EACH
ROW
; otherwise the return value is ignored.
Here’s a little example trigger procedure that forces an integer value in a table to keep track of the number of updates that are performed on the row. For new rows inserted, the value is initialized to 0 and then incremented on every update operation.
CREATE FUNCTION trigfunc_modcount() RETURNS trigger AS ’
if (pg.tg.op == "INSERT")
{
retval <- pg.tg.new
retval[pg.tg.args[1]] <- 0
}
if (pg.tg.op == "UPDATE")
{
retval <- pg.tg.new
retval[pg.tg.args[1]] <- pg.tg.old[pg.tg.args[1]] + 1
}
if (pg.tg.op == "DELETE")
retval <- pg.tg.old
return(retval)
’ LANGUAGE plr;
CREATE TABLE mytab (num integer, description text, modcnt integer);
CREATE TRIGGER trig_mytab_modcount BEFORE INSERT OR UPDATE ON mytab
FOR EACH ROW EXECUTE PROCEDURE trigfunc_modcount(’modcnt’);
Notice that the trigger procedure itself does not know the column name; that’s supplied from the trigger arguments. This lets the trigger procedure be reused with different tables.
License
License: GPL version 2 or newer. http://www.gnu.org/copyleft/gpl.html This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA