PostgreSQL HTTP Client

Motivation

Wouldn't it be nice to be able to write a trigger that called a web service? Either to get back a result, or to poke that service into refreshing itself against the new state of the database?

This extension is for that.

Examples

URL encode a string.

SELECT urlencode('my special string''s & things?');

              urlencode
-------------------------------------
 my+special+string%27s+%26+things%3F
(1 row)

URL encode a JSON associative array.

SELECT urlencode(jsonb_build_object('name','Colin & James','rate','50%'));

              urlencode
-------------------------------------
 name=Colin+%26+James&rate=50%25
(1 row)

Run a GET request and see the content.

SELECT content
  FROM http_get('http://httpbin.org/ip');

           content
-----------------------------
 {"origin":"24.69.186.43"}
(1 row)

Run a GET request with an Authorization header.

SELECT content::json->'headers'->>'Authorization'
  FROM http((
          'GET',
           'http://httpbin.org/headers',
           ARRAY[http_header('Authorization','Bearer eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9')],
           NULL,
           NULL
        )::http_request);

                   content
----------------------------------------------
 Bearer eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9
(1 row)

Read the status and content fields out of a http_response object.

SELECT status, content_type
  FROM http_get('http://httpbin.org/');

 status |       content_type
--------+--------------------------
    200 | text/html; charset=utf-8
(1 row)

Show all the http_header in an http_response object.

SELECT (unnest(headers)).*
  FROM http_get('http://httpbin.org/');

              field               |             value
----------------------------------+-------------------------------
 Connection                       | close
 Server                           | meinheld/0.6.1
 Date                             | Tue, 09 Jan 2018 18:40:30 GMT
 Content-Type                     | text/html; charset=utf-8
 Content-Length                   | 13011
 Access-Control-Allow-Origin      | *
 Access-Control-Allow-Credentials | true
 X-Powered-By                     | Flask
 X-Processed-Time                 | 0.0208520889282
 Via                              | 1.1 vegur

Use the PUT command to send a simple text document to a server.

SELECT status, content_type, content::json->>'data' AS data
  FROM http_put('http://httpbin.org/put', 'some text', 'text/plain');

 status |   content_type   |   data
--------+------------------+-----------
    200 | application/json | some text

Use the PATCH command to send a simple JSON document to a server.

SELECT status, content_type, content::json->>'data' AS data
  FROM http_patch('http://httpbin.org/patch', '{"this":"that"}', 'application/json');

 status |   content_type   |      data
--------+------------------+------------------
    200 | application/json | '{"this":"that"}'

Use the DELETE command to request resource deletion.

SELECT status, content_type, content::json->>'url' AS url
  FROM http_delete('http://httpbin.org/delete');

 status |   content_type   |            url
--------+------------------+---------------------------
    200 | application/json | http://httpbin.org/delete

As a shortcut to send data to a GET request, pass a JSONB data argument.

SELECT status, content::json->'args' AS args
  FROM http_get('http://httpbin.org/get',
                jsonb_build_object('myvar','myval','foo','bar'));

To POST to a URL using a data payload instead of parameters embedded in the URL, encode the data in a JSONB as a data payload.

SELECT status, content::json->'form' AS form
  FROM http_post('http://httpbin.org/post',
                 jsonb_build_object('myvar','myval','foo','bar'));

To access binary content, you must coerce the content from the default varchar representation to a bytea representation using the textsend function. Using the default varchar::bytea cast will not work, as the cast will stop the first time it hits a zero-valued byte (common in binary data).

WITH
  http AS (
    SELECT * FROM http_get('http://httpbin.org/image/png')
  ),
  headers AS (
    SELECT (unnest(headers)).* FROM http
  )
SELECT
  http.content_type,
  length(textsend(http.content)) AS length_binary,
  headers.value AS length_headers
FROM http, headers
WHERE field = 'Content-Length';

 content_type | length_binary | length_headers
--------------+---------------+----------------
 image/png    |          8090 | 8090

To access only the headers you can do a HEAD-Request. This will not follow redirections.

SELECT
    http.status,
    headers.value AS location
FROM
    http_head('http://google.com') AS http
    LEFT OUTER JOIN LATERAL (SELECT value
        FROM unnest(http.headers)
        WHERE field = 'Location') AS headers
        ON true;

 status |                         location
--------+-----------------------------------------------------------
    302 | http://www.google.ch/?gfe_rd=cr&ei=ACESWLy_KuvI8zeghL64Ag

Concepts

Every HTTP call is a made up of an http_request and an http_response.

     Composite type "public.http_request"
    Column    |       Type        | Modifiers
--------------+-------------------+-----------
 method       | http_method       |
 uri          | character varying |
 headers      | http_header[]     |
 content_type | character varying |
 content      | character varying |

    Composite type "public.http_response"
    Column    |       Type        | Modifiers
--------------+-------------------+-----------
 status       | integer           |
 content_type | character varying |
 headers      | http_header[]     |
 content      | character varying |

The utility functions, http_get(), http_post(), http_put(), http_delete() and http_head() are just wrappers around a master function, http(http_request) that returns http_response.

The headers field for requests and response is a PostgreSQL array of type http_header which is just a simple tuple.

  Composite type "public.http_header"
 Column |       Type        | Modifiers
--------+-------------------+-----------
 field  | character varying |
 value  | character varying |

As seen in the examples, you can unspool the array of http_header tuples into a result set using the PostgreSQL unnest() function on the array. From there you select out the particular header you are interested in.

Functions

http_header(field VARCHAR, value VARCHAR) returns http_header
http(request http_request) returns http_response
http_get(uri VARCHAR) returns http_response
http_get(uri VARCHAR, data JSONB) returns http_response
http_post(uri VARCHAR, content VARCHAR, content_type VARCHAR) returns http_response
http_post(uri VARCHAR, data JSONB) returns http_response
http_put(uri VARCHAR, content VARCHAR, content_type VARCHAR) returns http_response
http_patch(uri VARCHAR, content VARCHAR, content_type VARCHAR) returns http_response
http_delete(uri VARCHAR, content VARCHAR, content_type VARCHAR)) returns http_response
http_head(uri VARCHAR) returns http_response
http_set_curlopt(curlopt VARCHAR, value varchar) returns boolean
http_reset_curlopt() returns boolean
http_list_curlopt() returns setof(curlopt text, value text)
urlencode(string VARCHAR) returns text
urlencode(data JSONB) returns text

CURL Options

Select CURL options are available to set using the http_set_curlopt(curlopt VARCHAR, value varchar) function.

For example,

-- Set the PROXYPORT option
SELECT http_set_curlopt('CURLOPT_PROXYPORT', '12345');

-- List all currently set options
SELECT * FROM http_list_curlopt();

Will set the proxy port option for the lifetime of the database connection. You can reset all CURL options to their defaults using the http_reset_curlopt() function.

Using this extension as a background automated process without supervision (e.g as a trigger) may have unintended consequences for other servers. It is considered a best practice to share contact information with your requests, so that administrators can reach you in case your HTTP calls get out of control.

Certain API policies (e.g. Wikimedia User-Agent policy) may even require sharing specific contact information with each request. Others may disallow (via robots.txt) certain agents they don't recognize.

For such cases you can set the CURLOPT_USERAGENT option

SELECT http_set_curlopt('CURLOPT_USERAGENT',
                        'Examplebot/2.1 (+http://www.example.com/bot.html) Contact abuse@example.com');

SELECT status, content::json ->> 'user-agent' FROM http_get('http://httpbin.org/user-agent');

 status |                         user_agent
--------+-----------------------------------------------------------
    200 | Examplebot/2.1 (+http://www.example.com/bot.html) Contact abuse@example.com

Keep-Alive & Timeouts

The http_reset_curlopt() approach described above is recommended. The global variables below will be deprecated and removed over time.

By default each request uses a fresh connection and assures that the connection is closed when the request is done. This behavior reduces the chance of consuming system resources (sockets) as the extension runs over extended periods of time.

High-performance applications may wish to enable keep-alive and connection persistence to reduce latency and enhance throughput. The following GUC variable changes the behavior of the http extension to maintain connections as long as possible:

http.keepalive = 'on'

By default a 5 second timeout is set for the completion of a request. If a different timeout is desired the following GUC variable can be used to set it in milliseconds:

http.timeout_msec = 200

Installation

UNIX

If you have PostgreSQL (>= 9.3) devel packages and CURL devel packages installed (>= 0.7.20), you should have pg_config and curl-config on your path, so you should be able to just run make (or gmake), then make install, then in your database CREATE EXTENSION http.

If you already installed a previous version and you just want to upgrade, then ALTER EXTENSION http UPDATE.

Compiling for Apt based systems, using apt postgresql packages

Refer to https://wiki.postgresql.org/wiki/Apt for pulling packages from apt.postgresql.org repo

# replace the postgresql-server-dev-14 with your current version
sudo apt install postgresql-server-dev-14 libcurl4-openssl-dev make g++
make
sudo make install

Windows

There is a build available at postgresonline, not maintained by me.

Why This is a Bad Idea

"What happens if the web page takes a long time to return?" Your SQL call will just wait there until it does. Make sure your web service fails fast. Or (dangerous in a different way) run your query within pg_background.
"What if the web page returns junk?" Your SQL call will have to test for junk before doing anything with the payload.
"What if the web page never returns?" Set a short timeout, or send a cancel to the request, or just wait forever.
"What if a user queries a page they shouldn't?" Restrict function access, or just don't install a footgun like this extension where users can access it.

To Do

The new http://www.postgresql.org/docs/9.3/static/bgworker.html background worker support could be used to set up an HTTP request queue, so that pgsql-http can register a request and callback and then return immediately.
Inevitably some web server will return gzip content (Content-Encoding) without being asked for it. Handling that gracefully would be good.

liuzhe1947/pgsql-http