PL/v8
=====

PL/v8 is a trusted procedural language that is safe to use, fast to run and
easy to develop, powered by V8 JavaScript Engine.

Contents
--------

Implemented features are as follows.

- Scalar function calls
- Set returing function calls
- Trigger function calls
- Inline statement calls
- Auto mapping between JS and database built-in types
- Database access via SPI including prepared statements and cursors
- Subtransaction
- Utility functions
- Window function API
- Typed array
- Remote debugger
- Runtime environment separation across users in the same session
- Start-up procedure
- Dialects

Scalar function calls
---------------------

In PL/v8, you can write your SQL invoked function in JavaScript.  Use usual
CREATE FUNCTION statement with a JS function body.  Here is an example of a
scalar function call.

    CREATE FUNCTION plv8_test(keys text[], vals text[]) RETURNS text AS $$
        var o = {};
        for(var i=0; i<keys.length; i++){
            o[keys[i]] = vals[i];
        }
        return JSON.stringify(o);
    $$ LANGUAGE plv8 IMMUTABLE STRICT;
    
    SELECT plv8_test(ARRAY['name', 'age'], ARRAY['Tom', '29']);
    SELECT plv8_test(ARRAY['name', 'age'], ARRAY['Tom', '29']);
             plv8_test        
    ---------------------------
     {"name":"Tom","age":"29"}
    (1 row)

The function will be internally defined such that

    (function(arg1, arg2, ..){
       $funcbody$
    })

where $funcbody$ is the script source you specify in the CREATE FUNCTION AS
clause.  The argument names are inherited from the CREATE FUNCTION statement,
or they will be named as $1, $2 if the names are omitted.

Set returning function calls
----------------------------

PL/v8 supports set returning function calls.

    CREATE TYPE rec AS (i integer, t text);
    CREATE FUNCTION set_of_records() RETURNS SETOF rec AS
    $$
        // plv8.return_next() stores records in an internal tuplestore,
        // and return all of them at the end of function.
        plv8.return_next( { "i": 1, "t": "a" } );
        plv8.return_next( { "i": 2, "t": "b" } );
    
        // You can also return records with an array of JSON.
        return [ { "i": 3, "t": "c" }, { "i": 4, "t": "d" } ];
    $$
    LANGUAGE plv8;
    
    SELECT * FROM set_of_records();
     i | t
    ---+---
     1 | a
     2 | b
     3 | c
     4 | d
    (4 rows)

If the function is declared as RETURNS SETOF, PL/v8 prepares a tuplestore every
time called.  You can call plv8.return_next() function to add as many results as
you like to return rows from this function.  Alternatively, you can just return
a JS array to add set of records, a JS object to add a record, or a scalar value
to add a scalar to the tuplestore.  Unlike other PLs, PL/v8 does not support
the per-value return strategy, but it always uses the tuplestore strategy.
If the argument object has extra properties that are not defined by the argument,
return_next raises an error.

Trigger function calls
----------------------

PL/v8 supports trigger function calls.

    CREATE FUNCTION test_trigger() RETURNS trigger AS
    $$
        plv8.elog(NOTICE, "NEW = ", JSON.stringify(NEW));
        plv8.elog(NOTICE, "OLD = ", JSON.stringify(OLD));
        plv8.elog(NOTICE, "TG_OP = ", TG_OP);
        plv8.elog(NOTICE, "TG_ARGV = ", TG_ARGV);
        if (TG_OP == "UPDATE") {
            NEW.i = 102;
            return NEW;
        }
    $$
    LANGUAGE "plv8";
    
    CREATE TRIGGER test_trigger
        BEFORE INSERT OR UPDATE OR DELETE
        ON test_tbl FOR EACH ROW
        EXECUTE PROCEDURE test_trigger('foo', 'bar');

If the trigger type is an INSERT or UPDATE, you can assign properties of NEW
variable to change the actual tuple stored by this operation.

A plv8 trigger function will have special arguments to pass the trigger state as
following

- NEW
- OLD
- TG_NAME
- TG_WHEN
- TG_LEVEL
- TG_OP
- TG_RELID
- TG_TABLE_NAME
- TG_TABLE_SCHEMA
- TG_ARGV

For each variable semantics, see also the trigger section in PostgreSQL manual.

Inline statement calls
----------------------

PL/v8 supports DO block with PostgreSQL 9.0 and above.

    DO $$ plv8.elog(NOTICE, 'this', 'is', 'inline', 'code') $$ LANGUAGE plv8;

Auto mapping between JS and database built-in types
---------------------------------------------------

For the result and arguments, database types and JS types are mapped
automatically.  If the desired database type is one of

- oid
- bool
- int2
- int4
- int8
- float4
- float8
- numeric
- date
- timestamp
- timestamptz
- bytea
- json (>= 9.2)

and the JS value looks compatible, then the conversion succeeds.  Otherwise,
PL/v8 tries to convert them via cstring representation.  An array type is
supported only if the dimention is one.  A JS object will be mapped to
a tuple when applicable.  In addition to these types, PL/v8 supports
polymorphic types such like anyelement and anyarray.  Conversion of bytea is
a little different story.  See TypedArray section.

Database access via SPI including prepared statements and cursors
-----------------------------------------------------------------

### plv8.execute( sql [, args] ) ###

Executes SQL statements and retrieve the result. The `args` is an optional
argument that replaces $n placeholders in `sql`.  For SELECT queries, the
returned value is an array of objects. Each hash represents each record.
Column names are mapped to object properties.  For non-SELECT commands, the
returned value is an integer that represents number of affected rows.

    var json_result = plv8.execute( 'SELECT * FROM tbl' );
    var num_affected = plv8.execute( 'DELETE FROM tbl WHERE price > $1', [ 1000 ] );

Note this function and similar are not allowed outside of transaction,
which can be the case when using the remote debugger.

### plv8.prepare( sql, [, typenames] ) ###

Opens a prepared statement.  The `typename` parameter is an array where
each element is a string to indicate database type name for bind parameters.
Returned value is an object of PreparedPlan.  This object must be freed by
plan.free() before leaving the function.

    var plan = plv8.prepare( 'SELECT * FROM tbl WHERE col = $1', ['int'] );
    var rows = plan.execute( [1] );
    var sum = 0;
    for (var i = 0; i < rows.length; i++) {
      sum += rows[i].num;
    }
    plan.free();
    
    return sum;

### PreparedPlan.execute( [args] ) ###

Executes the prepared statement.  The `args` parameter is as plv8.execute(), and
can be omitted if the statement does not have parameters at all.  The result
of this method is also as described in plv8.execute().

### PreparedPlan.cursor( [args] ) ###

Opens a cursor from the prepared statement. The `args` parameter is as
plv8.execute(), and can be omitted if the statement does not have parameters
at all.  The returned object is of Cursor.  This must be closed by Cursor.close()
before leaving the function.

    var plan = plv8.prepare( 'SELECT * FROM tbl WHERE col = $1', ['int'] );
    var cursor = plan.cursor( [1] );
    var sum = 0, row;
    while (row = cursor.fetch()) {
      sum += row.num;
    }
    cursor.close();
    plan.free();
    
    return sum;

### PreparedPlan.free() ###

Frees the prepared statement.

### Cursor.fetch( [nrows] ) ###

When `nrows` parameter is omitted, fetches a row from the cursor and return
as an object (note: not an array.)  If specified, fetches as many rows as
the parameters up to exceeding, and returns an array of objects.  A negative
value for this parameter will fetch backwards.

### Cursor.move( [nrows] ) ###

Move the cursor `nrows` rows.  A negative value will move backwards.

### Cursor.close() ###

Closes the cursor.

Subtransaction
--------------

### plv8.subtransaction( func ) ###

plv8.execute() creates a subtransaction every time.  If you need an atomic
operation, you will need to call plv8.subtransaction() to create a subtransaction
block.

    try{
      plv8.subtransaction(function(){
        plv8.execute("INSERT INTO tbl VALUES(1)");  -- should be rolled back!
        plv8.execute("INSERT INTO tbl VALUES(1/0)");-- occurs an exception
      });
    } catch(e) {
      ... do fall back plan ...
    }

If one of the SQL execution in the subtransaction block fails, all of operation
within the block is rolled back.  If the process in the block throws JS
exception, it is transported to the outside.  So use try ... catch block to
capture it and do alternative operations when it happens.

Utility functions
-----------------

PL/v8 provides the following utility built-in functions.

- plv8.elog(elevel, msg1[, msg2, ...])
- plv8.quote_literal(str)
- plv8.nullable(str)
- plv8.quote_ident(str)

plv8.elog emits message to the client or the log file.  The elevel is one of

- DEBUG5
- DEBUG4
- DEBUG3
- DEBUG2
- DEBUG1
- LOG
- INFO
- NOTICE
- WARNING
- ERROR

See the PostgreSQL manual for each error level.

Each functionality for quote family is identical to the built-in SQL function
with the same name.

In addition, PL/v8 provides a function to access other plv8 functions that have
been registered in the database.

    CREATE FUNCTION callee(a int) RETURNS int AS $$ return a * a $$ LANGUAGE plv8;
    CREATE FUNCTION caller(a int, t int) RETURNS int AS $$
      var func = plv8.find_function("callee");
      return func(a);
    $$ LANGUAGE plv8;

With plv8.find_function(), you can look up other plv8 functions.  If they are
not the plv8 function, it errors out.  The function signature parameter to
plv8.find_function() is either of regproc (function name only) or regprocedure
(function name with argument types).  You can make use of internal type for
arguments and void type for return type for the pure JavaScript function to
make sure any invocation from SQL statement should not happen.

The plv8 object provides version string as `plv8.version`.  This string
corresponds to plv8 module version.  Note this is not the extension version.

Window function API
-------------------

You can define user-defined window functions with PL/v8.  It wraps C-level
window function API to support full functionality.  To create one, first
obtain window object by plv8.get_window_object(), which provides interfaces
as follows.

### WindowObject.get_current_position() ###

Returns the current position in the partition, starting from 0.

### WindowObject.get_partition_row_count() ###

Returns the number of rows in the partition.

### WindowObject.set_mark_position( pos ) ###

Set mark at the specified row.  Rows above this position will be gone and
not be accessible later.

### WindowObject.rows_are_peers( pos1, pos2 ) ###

Returns true if the rows at `pos1` and `pos2` are peers.

### WindowObject.get_func_arg_in_partition( argno, relpos, seektype, mark_pos ) ###

### WindowObject.get_func_arg_in_frame( argno, relpos, seektype, mark_pos ) ###

Returns the value of the argument in `argno` (starting from 0) to this
function at the `relpos` row from `seektype` in the current partition or
frame.  `seektype` can be either of WindowObject.SEEK_HEAD,
WindowObject.SEEK_CURRENT, or WindowObject.SEEK_TAIL.  If `mark_pos` is true,
the row the argument is fetched from is marked.  If the specified row is
out of the partition/frame, the returned value will be `undefined`.

### WindowObject.get_func_arg_in_current( argno ) ###

Returns the value of the argument in `argno` (starting from 0) to this
function at the current row.  Note that the returned value will be the
same as the argument variable of the function.

### WindowObject.get_partition_local( [size] ) ###

Returns partition-local value, which is released at the end of the current
partition.  If nothing is stored, `undefined` is returned.  `size` argument
(default 1000) is the byte size of the allocated memory in the first call.
Once the memory allocated, the size will not change.

### WindowObject.set_partition_local( obj ) ###

Stores the partition-local value, which you can retrieve later by
get_partition_local().  This function internally uses JSON.stringify to
serialize the object, so if you pass value that is not able to be serialized
may end up being unexpected value.  If the size of serialized value is
more than the allocated memory, it will throw an exception.

You can also learn more on how to use these API in sql/window.sql regression
test, which implements most of the native window functions.  For the general
information of the user-defined window function, see the CREATE FUNCTION
page of the PostgreSQL manual.

Typed array
-----------

The typed array is something v8 provides to allow fast access to native memory,
mainly for the purpose of their canvas support in browsers.  PL/v8 uses this
to map bytea and various array types to JavaScript array.  In case of bytea,
you can access each byte as an array of unsigned bytes.  For
int2/int4/float4/float8 array type, PL/v8 provides direct access to each
element by using PL/v8 domain types.

- plv8_int2array maps int2[]
- plv8_int4array maps int4[]
- plv8_float4array maps float4[]
- plv8_float8array maps float8[]

These are only annotations that tell PL/v8 to use fast access method instead of
regular one.  For these typed arrays, only 1-dimensional array without NULL
element.  Also, there is currently no way to create such typed array inside
PL/v8 functions.  Only arguments can be typed array.  You can modify the element
and return the value.  An example for these types are as follows.

  CREATE FUNCTION int4sum(ary plv8_int4array) RETURNS int8 AS $$
    var sum = 0;
    for (var i = 0; i < ary.length; i++) {
      sum += ary[i];
    }
    return sum;
  $$ LANGUAGE plv8 IMMUTABLE STRICT;
  
  SELECT int4sum(ARRAY[1, 2, 3, 4, 5]);
   int4sum 
  ---------
        15
  (1 row)

Remote debugger
---------------

PL/v8 supports v8 remote debugger.  You need to enable it at the compile time
to pass ENABLE_DEBUGGER_SUPPORT to `make`.  `make static` will automatically
turns it on.  If enabled, and once PL/v8 module is loaded (and the execution
engine is initialized, PL/v8 accepts a remote debugger connection.  If you
have `d8` from v8 package, run with `--remote-debug --debug-port=35432` to
attach the functions.  If you want to change the remote debugger port, there
is a GUC `plv8.debugger_port` to set the port number. You can also try
`debugger` statement inside functions to set a breakpoint.  For more details
of v8 remote debugger, see v8 documentation.

Runtime environment separation across users in the same session
---------------------------------------------------------------

In PL/v8, each session has one global JS runtime context.  This enables function
invocations at low cost, and sharing common object among the functions.  However,
for the security reasons, if the user switches to another with SET ROLE command,
a new JS runtime context is initialized and used separately.  This prevents
unexpected information leak risk.

Each plv8 function is invoked as if the function is the property of other object.
This means "this" in each function is a JS object that is created every time
the function is executed in a query.  In other words, the life time and the
visibility of "this" object in a function is only a series of function calls in
a query.  If you need to share some value among different functions, keep it in
`plv8` object because each function invocation has different "this" object.

Start-up procedure
------------------

PL/v8 provides a start up facility, which allows to call initialization plv8
function specified in the GUC variable.

    SET plv8.start_proc = 'plv8_init';
    SELECT plv8_test(10);

If this variable is set when the runtime is initialized, before the function
call of plv8_test() another plv8 function plv8_init() is invoked.  In such
initialization function, you can add any properties to `plv8` object to expose
common values or assign them to "this" property.  In the initialization
function, the receiver "this" is specially pointing to the global object, so
the variables that is assigned to "this" property in this initialization are
visible from any subsequent function as global variables.

Remember CREATE FUNCTION also starts the plv8 runtime environment, so make sure
to SET this GUC before any plv8 actions including CREATE FUNCTION.

Dialects
--------

This module also contains some dialect supports.  Currently, we have two dialects
are supported.

- CoffeeScript (plcoffee)
- LiveScript (plls)

With PostgreSQL 9.1 or above, you are able to load tohse dialects via CREATE
EXTENSION command.