Introduction

The various *_by_lua, *_by_lua_block and *_by_lua_file configuration directives serve as gateways to the Lua API within the nginx.conf file. The Nginx Lua API described below can only be called within the user Lua code run in the context of these configuration directives.

The API is exposed to Lua in the form of two standard packages ngx and ndk. These packages are in the default global scope within ngx_lua and are always available within ngx_lua directives.

The packages can be introduced into external Lua modules like this:


 local say = ngx.say

 local _M = {}

 function _M.foo(a)
     say(a)
 end

 return _M

Use of the package.seeall flag is strongly discouraged due to its various bad side-effects.

It is also possible to directly require the packages in external Lua modules:


 local ngx = require "ngx"
 local ndk = require "ndk"

The ability to require these packages was introduced in the v0.2.1rc19 release.

Network I/O operations in user code should only be done through the Nginx Lua API calls as the Nginx event loop may be blocked and performance drop off dramatically otherwise. Disk operations with relatively small amount of data can be done using the standard Lua io library but huge file reading and writing should be avoided wherever possible as they may block the Nginx process significantly. Delegating all network and disk I/O operations to Nginx's subrequests (via the ngx.location.capture method and similar) is strongly recommended for maximum performance.

Back to TOC

ngx.arg

syntax: val = ngx.arg[index]

context: set_by_lua*, body_filter_by_lua*

When this is used in the context of the set_by_lua* directives, this table is read-only and holds the input arguments to the config directives:


 value = ngx.arg[n]

Here is an example


 location /foo {
     set $a 32;
     set $b 56;

     set_by_lua $sum
         'return tonumber(ngx.arg[1]) + tonumber(ngx.arg[2])'
         $a $b;

     echo $sum;
 }

that writes out 88, the sum of 32 and 56.

When this table is used in the context of body_filter_by_lua*, the first element holds the input data chunk to the output filter code and the second element holds the boolean flag for the "eof" flag indicating the end of the whole output data stream.

The data chunk and "eof" flag passed to the downstream Nginx output filters can also be overridden by assigning values directly to the corresponding table elements. When setting nil or an empty Lua string value to ngx.arg[1], no data chunk will be passed to the downstream Nginx output filters at all.

Back to TOC

ngx.var.VARIABLE

syntax: ngx.var.VAR_NAME

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*

Read and write Nginx variable values.


 value = ngx.var.some_nginx_variable_name
 ngx.var.some_nginx_variable_name = value

Note that only already defined nginx variables can be written to. For example:


 location /foo {
     set $my_var ''; # this line is required to create $my_var at config time
     content_by_lua_block {
         ngx.var.my_var = 123;
         ...
     }
 }

That is, nginx variables cannot be created on-the-fly.

Some special nginx variables like $args and $limit_rate can be assigned a value, many others are not, like $query_string, $arg_PARAMETER, and $http_NAME.

Nginx regex group capturing variables $1, $2, $3, and etc, can be read by this interface as well, by writing ngx.var[1], ngx.var[2], ngx.var[3], and etc.

Setting ngx.var.Foo to a nil value will unset the $Foo Nginx variable.


 ngx.var.args = nil

WARNING When reading from an Nginx variable, Nginx will allocate memory in the per-request memory pool which is freed only at request termination. So when you need to read from an Nginx variable repeatedly in your Lua code, cache the Nginx variable value to your own Lua variable, for example,


 local val = ngx.var.some_var
 --- use the val repeatedly later

to prevent (temporary) memory leaking within the current request's lifetime. Another way of caching the result is to use the ngx.ctx table.

Undefined NGINX variables are evaluated to nil while uninitialized (but defined) NGINX variables are evaluated to an empty Lua string.

This API requires a relatively expensive metamethod call and it is recommended to avoid using it on hot code paths.

Back to TOC

Core constants

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, *log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*


   ngx.OK (0)
   ngx.ERROR (-1)
   ngx.AGAIN (-2)
   ngx.DONE (-4)
   ngx.DECLINED (-5)

Note that only three of these constants are utilized by the Nginx API for Lua (i.e., ngx.exit accepts NGX_OK, NGX_ERROR, and NGX_DECLINED as input).


   ngx.null

The ngx.null constant is a NULL light userdata usually used to represent nil values in Lua tables etc and is similar to the lua-cjson library's cjson.null constant. This constant was first introduced in the v0.5.0rc5 release.

The ngx.DECLINED constant was first introduced in the v0.5.0rc19 release.

Back to TOC

HTTP method constants

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

  ngx.HTTP_GET
  ngx.HTTP_HEAD
  ngx.HTTP_PUT
  ngx.HTTP_POST
  ngx.HTTP_DELETE
  ngx.HTTP_OPTIONS   (added in the v0.5.0rc24 release)
  ngx.HTTP_MKCOL     (added in the v0.8.2 release)
  ngx.HTTP_COPY      (added in the v0.8.2 release)
  ngx.HTTP_MOVE      (added in the v0.8.2 release)
  ngx.HTTP_PROPFIND  (added in the v0.8.2 release)
  ngx.HTTP_PROPPATCH (added in the v0.8.2 release)
  ngx.HTTP_LOCK      (added in the v0.8.2 release)
  ngx.HTTP_UNLOCK    (added in the v0.8.2 release)
  ngx.HTTP_PATCH     (added in the v0.8.2 release)
  ngx.HTTP_TRACE     (added in the v0.8.2 release)

These constants are usually used in ngx.location.capture and ngx.location.capture_multi method calls.

Back to TOC

HTTP status constants

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*


   value = ngx.HTTP_CONTINUE (100) (first added in the v0.9.20 release)
   value = ngx.HTTP_SWITCHING_PROTOCOLS (101) (first added in the v0.9.20 release)
   value = ngx.HTTP_OK (200)
   value = ngx.HTTP_CREATED (201)
   value = ngx.HTTP_ACCEPTED (202) (first added in the v0.9.20 release)
   value = ngx.HTTP_NO_CONTENT (204) (first added in the v0.9.20 release)
   value = ngx.HTTP_PARTIAL_CONTENT (206) (first added in the v0.9.20 release)
   value = ngx.HTTP_SPECIAL_RESPONSE (300)
   value = ngx.HTTP_MOVED_PERMANENTLY (301)
   value = ngx.HTTP_MOVED_TEMPORARILY (302)
   value = ngx.HTTP_SEE_OTHER (303)
   value = ngx.HTTP_NOT_MODIFIED (304)
   value = ngx.HTTP_TEMPORARY_REDIRECT (307) (first added in the v0.9.20 release)
   value = ngx.HTTP_BAD_REQUEST (400)
   value = ngx.HTTP_UNAUTHORIZED (401)
   value = ngx.HTTP_PAYMENT_REQUIRED (402) (first added in the v0.9.20 release)
   value = ngx.HTTP_FORBIDDEN (403)
   value = ngx.HTTP_NOT_FOUND (404)
   value = ngx.HTTP_NOT_ALLOWED (405)
   value = ngx.HTTP_NOT_ACCEPTABLE (406) (first added in the v0.9.20 release)
   value = ngx.HTTP_REQUEST_TIMEOUT (408) (first added in the v0.9.20 release)
   value = ngx.HTTP_CONFLICT (409) (first added in the v0.9.20 release)
   value = ngx.HTTP_GONE (410)
   value = ngx.HTTP_UPGRADE_REQUIRED (426) (first added in the v0.9.20 release)
   value = ngx.HTTP_TOO_MANY_REQUESTS (429) (first added in the v0.9.20 release)
   value = ngx.HTTP_CLOSE (444) (first added in the v0.9.20 release)
   value = ngx.HTTP_ILLEGAL (451) (first added in the v0.9.20 release)
   value = ngx.HTTP_INTERNAL_SERVER_ERROR (500)
   value = ngx.HTTP_METHOD_NOT_IMPLEMENTED (501)
   value = ngx.HTTP_BAD_GATEWAY (502) (first added in the v0.9.20 release)
   value = ngx.HTTP_SERVICE_UNAVAILABLE (503)
   value = ngx.HTTP_GATEWAY_TIMEOUT (504) (first added in the v0.3.1rc38 release)
   value = ngx.HTTP_VERSION_NOT_SUPPORTED (505) (first added in the v0.9.20 release)
   value = ngx.HTTP_INSUFFICIENT_STORAGE (507) (first added in the v0.9.20 release)

Back to TOC

Nginx log level constants

context: init_by_lua*, init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*


   ngx.STDERR
   ngx.EMERG
   ngx.ALERT
   ngx.CRIT
   ngx.ERR
   ngx.WARN
   ngx.NOTICE
   ngx.INFO
   ngx.DEBUG

These constants are usually used by the ngx.log method.

Back to TOC

print

syntax: print(...)

context: init_by_lua*, init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Writes argument values into the nginx error.log file with the ngx.NOTICE log level.

It is equivalent to


 ngx.log(ngx.NOTICE, ...)

Lua nil arguments are accepted and result in literal "nil" strings while Lua booleans result in literal "true" or "false" strings. And the ngx.null constant will yield the "null" string output.

There is a hard coded 2048 byte limitation on error message lengths in the Nginx core. This limit includes trailing newlines and leading time stamps. If the message size exceeds this limit, Nginx will truncate the message text accordingly. This limit can be manually modified by editing the NGX_MAX_ERROR_STR macro definition in the src/core/ngx_log.h file in the Nginx source tree.

Back to TOC

ngx.ctx

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*

This table can be used to store per-request Lua context data and has a life time identical to the current request (as with the Nginx variables).

Consider the following example,


 location /test {
     rewrite_by_lua_block {
         ngx.ctx.foo = 76
     }
     access_by_lua_block {
         ngx.ctx.foo = ngx.ctx.foo + 3
     }
     content_by_lua_block {
         ngx.say(ngx.ctx.foo)
     }
 }

Then GET /test will yield the output


 79

That is, the ngx.ctx.foo entry persists across the rewrite, access, and content phases of a request.

Every request, including subrequests, has its own copy of the table. For example:


 location /sub {
     content_by_lua_block {
         ngx.say("sub pre: ", ngx.ctx.blah)
         ngx.ctx.blah = 32
         ngx.say("sub post: ", ngx.ctx.blah)
     }
 }

 location /main {
     content_by_lua_block {
         ngx.ctx.blah = 73
         ngx.say("main pre: ", ngx.ctx.blah)
         local res = ngx.location.capture("/sub")
         ngx.print(res.body)
         ngx.say("main post: ", ngx.ctx.blah)
     }
 }

Then GET /main will give the output


 main pre: 73
 sub pre: nil
 sub post: 32
 main post: 73

Here, modification of the ngx.ctx.blah entry in the subrequest does not affect the one in the parent request. This is because they have two separate versions of ngx.ctx.blah.

Internal redirection will destroy the original request ngx.ctx data (if any) and the new request will have an empty ngx.ctx table. For instance,


 location /new {
     content_by_lua_block {
         ngx.say(ngx.ctx.foo)
     }
 }

 location /orig {
     content_by_lua_block {
         ngx.ctx.foo = "hello"
         ngx.exec("/new")
     }
 }

Then GET /orig will give


 nil

rather than the original "hello" value.

Arbitrary data values, including Lua closures and nested tables, can be inserted into this "magic" table. It also allows the registration of custom meta methods.

Overriding ngx.ctx with a new Lua table is also supported, for example,


 ngx.ctx = { foo = 32, bar = 54 }

When being used in the context of init_worker_by_lua*, this table just has the same lifetime of the current Lua handler.

The ngx.ctx lookup requires relatively expensive metamethod calls and it is much slower than explicitly passing per-request data along by your own function arguments. So do not abuse this API for saving your own function arguments because it usually has quite some performance impact.

Because of the metamethod magic, never "local" the ngx.ctx table outside your Lua function scope on the Lua module level level due to worker-level data sharing. For example, the following is bad:


 -- mymodule.lua
 local _M = {}

 -- the following line is bad since ngx.ctx is a per-request
 -- data while this `ctx` variable is on the Lua module level
 -- and thus is per-nginx-worker.
 local ctx = ngx.ctx

 function _M.main()
     ctx.foo = "bar"
 end

 return _M

Use the following instead:


 -- mymodule.lua
 local _M = {}

 function _M.main(ctx)
     ctx.foo = "bar"
 end

 return _M

That is, let the caller pass the ctx table explicitly via a function argument.

Back to TOC

ngx.location.capture

syntax: res = ngx.location.capture(uri, options?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Issues a synchronous but still non-blocking Nginx Subrequest using uri.

Nginx's subrequests provide a powerful way to make non-blocking internal requests to other locations configured with disk file directory or any other nginx C modules like ngx_proxy, ngx_fastcgi, ngx_memc, ngx_postgres, ngx_drizzle, and even ngx_lua itself and etc etc etc.

Also note that subrequests just mimic the HTTP interface but there is no extra HTTP/TCP traffic nor IPC involved. Everything works internally, efficiently, on the C level.

Subrequests are completely different from HTTP 301/302 redirection (via ngx.redirect) and internal redirection (via ngx.exec).

You should always read the request body (by either calling ngx.req.read_body or configuring lua_need_request_body on) before initiating a subrequest.

This API function (as well as ngx.location.capture_multi) always buffers the whole response body of the subrequest in memory. Thus, you should use cosockets and streaming processing instead if you have to handle large subrequest responses.

Here is a basic example:


 res = ngx.location.capture(uri)

Returns a Lua table with 4 slots: res.status, res.header, res.body, and res.truncated.

res.status holds the response status code for the subrequest response.

res.header holds all the response headers of the subrequest and it is a normal Lua table. For multi-value response headers, the value is a Lua (array) table that holds all the values in the order that they appear. For instance, if the subrequest response headers contain the following lines:


 Set-Cookie: a=3
 Set-Cookie: foo=bar
 Set-Cookie: baz=blah

Then res.header["Set-Cookie"] will be evaluated to the table value {"a=3", "foo=bar", "baz=blah"}.

res.body holds the subrequest's response body data, which might be truncated. You always need to check the res.truncated boolean flag to see if res.body contains truncated data. The data truncation here can only be caused by those unrecoverable errors in your subrequests like the cases that the remote end aborts the connection prematurely in the middle of the response body data stream or a read timeout happens when your subrequest is receiving the response body data from the remote.

URI query strings can be concatenated to URI itself, for instance,


 res = ngx.location.capture('/foo/bar?a=3&b=4')

Named locations like @foo are not allowed due to a limitation in the nginx core. Use normal locations combined with the internal directive to prepare internal-only locations.

An optional option table can be fed as the second argument, which supports the options:

  • method specify the subrequest's request method, which only accepts constants like ngx.HTTP_POST.
  • body specify the subrequest's request body (string value only).
  • args specify the subrequest's URI query arguments (both string value and Lua tables are accepted)
  • ctx specify a Lua table to be the ngx.ctx table for the subrequest. It can be the current request's ngx.ctx table, which effectively makes the parent and its subrequest to share exactly the same context table. This option was first introduced in the v0.3.1rc25 release.
  • vars take a Lua table which holds the values to set the specified Nginx variables in the subrequest as this option's value. This option was first introduced in the v0.3.1rc31 release.
  • copy_all_vars specify whether to copy over all the Nginx variable values of the current request to the subrequest in question. modifications of the nginx variables in the subrequest will not affect the current (parent) request. This option was first introduced in the v0.3.1rc31 release.
  • share_all_vars specify whether to share all the Nginx variables of the subrequest with the current (parent) request. modifications of the Nginx variables in the subrequest will affect the current (parent) request. Enabling this option may lead to hard-to-debug issues due to bad side-effects and is considered bad and harmful. Only enable this option when you completely know what you are doing.
  • always_forward_body when set to true, the current (parent) request's request body will always be forwarded to the subrequest being created if the body option is not specified. The request body read by either ngx.req.read_body() or lua_need_request_body on will be directly forwarded to the subrequest without copying the whole request body data when creating the subrequest (no matter the request body data is buffered in memory buffers or temporary files). By default, this option is false and when the body option is not specified, the request body of the current (parent) request is only forwarded when the subrequest takes the PUT or POST request method.

Issuing a POST subrequest, for example, can be done as follows


 res = ngx.location.capture(
     '/foo/bar',
     { method = ngx.HTTP_POST, body = 'hello, world' }
 )

See HTTP method constants methods other than POST. The method option is ngx.HTTP_GET by default.

The args option can specify extra URI arguments, for instance,


 ngx.location.capture('/foo?a=1',
     { args = { b = 3, c = ':' } }
 )

is equivalent to


 ngx.location.capture('/foo?a=1&b=3&c=%3a')

that is, this method will escape argument keys and values according to URI rules and concatenate them together into a complete query string. The format for the Lua table passed as the args argument is identical to the format used in the ngx.encode_args method.

The args option can also take plain query strings:


 ngx.location.capture('/foo?a=1',
     { args = 'b=3&c=%3a' } }
 )

This is functionally identical to the previous examples.

The share_all_vars option controls whether to share nginx variables among the current request and its subrequests. If this option is set to true, then the current request and associated subrequests will share the same Nginx variable scope. Hence, changes to Nginx variables made by a subrequest will affect the current request.

Care should be taken in using this option as variable scope sharing can have unexpected side effects. The args, vars, or copy_all_vars options are generally preferable instead.

This option is set to false by default


 location /other {
     set $dog "$dog world";
     echo "$uri dog: $dog";
 }

 location /lua {
     set $dog 'hello';
     content_by_lua_block {
         res = ngx.location.capture("/other",
             { share_all_vars = true });

         ngx.print(res.body)
         ngx.say(ngx.var.uri, ": ", ngx.var.dog)
     }
 }

Accessing location /lua gives

/other dog: hello world
/lua: hello world

The copy_all_vars option provides a copy of the parent request's Nginx variables to subrequests when such subrequests are issued. Changes made to these variables by such subrequests will not affect the parent request or any other subrequests sharing the parent request's variables.


 location /other {
     set $dog "$dog world";
     echo "$uri dog: $dog";
 }

 location /lua {
     set $dog 'hello';
     content_by_lua_block {
         res = ngx.location.capture("/other",
             { copy_all_vars = true });

         ngx.print(res.body)
         ngx.say(ngx.var.uri, ": ", ngx.var.dog)
     }
 }

Request GET /lua will give the output

/other dog: hello world
/lua: hello

Note that if both share_all_vars and copy_all_vars are set to true, then share_all_vars takes precedence.

In addition to the two settings above, it is possible to specify values for variables in the subrequest using the vars option. These variables are set after the sharing or copying of variables has been evaluated, and provides a more efficient method of passing specific values to a subrequest over encoding them as URL arguments and unescaping them in the Nginx config file.


 location /other {
     content_by_lua_block {
         ngx.say("dog = ", ngx.var.dog)
         ngx.say("cat = ", ngx.var.cat)
     }
 }

 location /lua {
     set $dog '';
     set $cat '';
     content_by_lua_block {
         res = ngx.location.capture("/other",
             { vars = { dog = "hello", cat = 32 }});

         ngx.print(res.body)
     }
 }

Accessing /lua will yield the output

dog = hello
cat = 32

The ctx option can be used to specify a custom Lua table to serve as the ngx.ctx table for the subrequest.


 location /sub {
     content_by_lua_block {
         ngx.ctx.foo = "bar";
     }
 }
 location /lua {
     content_by_lua_block {
         local ctx = {}
         res = ngx.location.capture("/sub", { ctx = ctx })

         ngx.say(ctx.foo);
         ngx.say(ngx.ctx.foo);
     }
 }

Then request GET /lua gives

bar
nil

It is also possible to use this ctx option to share the same ngx.ctx table between the current (parent) request and the subrequest:


 location /sub {
     content_by_lua_block {
         ngx.ctx.foo = "bar";
     }
 }
 location /lua {
     content_by_lua_block {
         res = ngx.location.capture("/sub", { ctx = ngx.ctx })
         ngx.say(ngx.ctx.foo);
     }
 }

Request GET /lua yields the output

bar

Note that subrequests issued by ngx.location.capture inherit all the request headers of the current request by default and that this may have unexpected side effects on the subrequest responses. For example, when using the standard ngx_proxy module to serve subrequests, an "Accept-Encoding: gzip" header in the main request may result in gzipped responses that cannot be handled properly in Lua code. Original request headers should be ignored by setting proxy_pass_request_headers to off in subrequest locations.

When the body option is not specified and the always_forward_body option is false (the default value), the POST and PUT subrequests will inherit the request bodies of the parent request (if any).

There is a hard-coded upper limit on the number of concurrent subrequests possible for every main request. In older versions of Nginx, the limit was 50 concurrent subrequests and in more recent versions, Nginx 1.1.x onwards, this was increased to 200 concurrent subrequests. When this limit is exceeded, the following error message is added to the error.log file:

[error] 13983#0: *1 subrequests cycle while processing "/uri"

The limit can be manually modified if required by editing the definition of the NGX_HTTP_MAX_SUBREQUESTS macro in the nginx/src/http/ngx_http_request.h file in the Nginx source tree.

Please also refer to restrictions on capturing locations configured by subrequest directives of other modules.

Back to TOC

ngx.location.capture_multi

syntax: res1, res2, ... = ngx.location.capture_multi({ {uri, options?}, {uri, options?}, ... })

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Just like ngx.location.capture, but supports multiple subrequests running in parallel.

This function issues several parallel subrequests specified by the input table and returns their results in the same order. For example,


 res1, res2, res3 = ngx.location.capture_multi{
     { "/foo", { args = "a=3&b=4" } },
     { "/bar" },
     { "/baz", { method = ngx.HTTP_POST, body = "hello" } },
 }

 if res1.status == ngx.HTTP_OK then
     ...
 end

 if res2.body == "BLAH" then
     ...
 end

This function will not return until all the subrequests terminate. The total latency is the longest latency of the individual subrequests rather than the sum.

Lua tables can be used for both requests and responses when the number of subrequests to be issued is not known in advance:


 -- construct the requests table
 local reqs = {}
 table.insert(reqs, { "/mysql" })
 table.insert(reqs, { "/postgres" })
 table.insert(reqs, { "/redis" })
 table.insert(reqs, { "/memcached" })

 -- issue all the requests at once and wait until they all return
 local resps = { ngx.location.capture_multi(reqs) }

 -- loop over the responses table
 for i, resp in ipairs(resps) do
     -- process the response table "resp"
 end

The ngx.location.capture function is just a special form of this function. Logically speaking, the ngx.location.capture can be implemented like this


 ngx.location.capture =
     function (uri, args)
         return ngx.location.capture_multi({ {uri, args} })
     end

Please also refer to restrictions on capturing locations configured by subrequest directives of other modules.

Back to TOC

ngx.status

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*

Read and write the current request's response status. This should be called before sending out the response headers.


 ngx.status = ngx.HTTP_CREATED
 status = ngx.status

Setting ngx.status after the response header is sent out has no effect but leaving an error message in your nginx's error log file:

attempt to set ngx.status after sending out response headers

Back to TOC

ngx.header.HEADER

syntax: ngx.header.HEADER = VALUE

syntax: value = ngx.header.HEADER

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*

Set, add to, or clear the current request's HEADER response header that is to be sent.

Underscores (_) in the header names will be replaced by hyphens (-) by default. This transformation can be turned off via the lua_transform_underscores_in_response_headers directive.

The header names are matched case-insensitively.


 -- equivalent to ngx.header["Content-Type"] = 'text/plain'
 ngx.header.content_type = 'text/plain';

 ngx.header["X-My-Header"] = 'blah blah';

Multi-value headers can be set this way:


 ngx.header['Set-Cookie'] = {'a=32; path=/', 'b=4; path=/'}

will yield


 Set-Cookie: a=32; path=/
 Set-Cookie: b=4; path=/

in the response headers.

Only Lua tables are accepted (Only the last element in the table will take effect for standard headers such as Content-Type that only accept a single value).


 ngx.header.content_type = {'a', 'b'}

is equivalent to


 ngx.header.content_type = 'b'

Setting a slot to nil effectively removes it from the response headers:


 ngx.header["X-My-Header"] = nil;

The same applies to assigning an empty table:


 ngx.header["X-My-Header"] = {};

Setting ngx.header.HEADER after sending out response headers (either explicitly with ngx.send_headers or implicitly with ngx.print and similar) will throw out a Lua exception.

Reading ngx.header.HEADER will return the value of the response header named HEADER.

Underscores (_) in the header names will also be replaced by dashes (-) and the header names will be matched case-insensitively. If the response header is not present at all, nil will be returned.

This is particularly useful in the context of header_filter_by_lua*, for example,


 location /test {
     set $footer '';

     proxy_pass http://some-backend;

     header_filter_by_lua_block {
         if ngx.header["X-My-Header"] == "blah" then
             ngx.var.footer = "some value"
         end
     }

     echo_after_body $footer;
 }

For multi-value headers, all of the values of header will be collected in order and returned as a Lua table. For example, response headers

Foo: bar
Foo: baz

will result in


 {"bar", "baz"}

to be returned when reading ngx.header.Foo.

Note that ngx.header is not a normal Lua table and as such, it is not possible to iterate through it using the Lua ipairs function.

For reading request headers, use the ngx.req.get_headers function instead.

Back to TOC

ngx.resp.get_headers

syntax: headers = ngx.resp.get_headers(max_headers?, raw?)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, balancer_by_lua*

Returns a Lua table holding all the current response headers for the current request.


 local h = ngx.resp.get_headers()
 for k, v in pairs(h) do
     ...
 end

This function has the same signature as ngx.req.get_headers except getting response headers instead of request headers.

This API was first introduced in the v0.9.5 release.

Back to TOC

ngx.req.is_internal

syntax: is_internal = ngx.req.is_internal()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*

Returns a boolean indicating whether the current request is an "internal request", i.e., a request initiated from inside the current nginx server instead of from the client side.

Subrequests are all internal requests and so are requests after internal redirects.

This API was first introduced in the v0.9.20 release.

Back to TOC

ngx.req.start_time

syntax: secs = ngx.req.start_time()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*

Returns a floating-point number representing the timestamp (including milliseconds as the decimal part) when the current request was created.

The following example emulates the $request_time variable value (provided by ngx_http_log_module) in pure Lua:


 local request_time = ngx.now() - ngx.req.start_time()

This function was first introduced in the v0.7.7 release.

See also ngx.now and ngx.update_time.

Back to TOC

ngx.req.http_version

syntax: num = ngx.req.http_version()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*

Returns the HTTP version number for the current request as a Lua number.

Current possible values are 2.0, 1.0, 1.1, and 0.9. Returns nil for unrecognized values.

This method was first introduced in the v0.7.17 release.

Back to TOC

ngx.req.raw_header

syntax: str = ngx.req.raw_header(no_request_line?)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*

Returns the original raw HTTP protocol header received by the Nginx server.

By default, the request line and trailing CR LF terminator will also be included. For example,


 ngx.print(ngx.req.raw_header())

gives something like this:

GET /t HTTP/1.1
Host: localhost
Connection: close
Foo: bar

You can specify the optional no_request_line argument as a true value to exclude the request line from the result. For example,


 ngx.print(ngx.req.raw_header(true))

outputs something like this:

Host: localhost
Connection: close
Foo: bar

This method was first introduced in the v0.7.17 release.

This method does not work in HTTP/2 requests yet.

Back to TOC

ngx.req.get_method

syntax: method_name = ngx.req.get_method()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, balancer_by_lua*

Retrieves the current request's request method name. Strings like "GET" and "POST" are returned instead of numerical method constants.

If the current request is an Nginx subrequest, then the subrequest's method name will be returned.

This method was first introduced in the v0.5.6 release.

See also ngx.req.set_method.

Back to TOC

ngx.req.set_method

syntax: ngx.req.set_method(method_id)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*

Overrides the current request's request method with the method_id argument. Currently only numerical method constants are supported, like ngx.HTTP_POST and ngx.HTTP_GET.

If the current request is an Nginx subrequest, then the subrequest's method will be overridden.

This method was first introduced in the v0.5.6 release.

See also ngx.req.get_method.

Back to TOC

ngx.req.set_uri

syntax: ngx.req.set_uri(uri, jump?)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*

Rewrite the current request's (parsed) URI by the uri argument. The uri argument must be a Lua string and cannot be of zero length, or a Lua exception will be thrown.

The optional boolean jump argument can trigger location rematch (or location jump) as ngx_http_rewrite_module's rewrite directive, that is, when jump is true (default to false), this function will never return and it will tell Nginx to try re-searching locations with the new URI value at the later post-rewrite phase and jumping to the new location.

Location jump will not be triggered otherwise, and only the current request's URI will be modified, which is also the default behavior. This function will return but with no returned values when the jump argument is false or absent altogether.

For example, the following nginx config snippet


 rewrite ^ /foo last;

can be coded in Lua like this:


 ngx.req.set_uri("/foo", true)

Similarly, Nginx config


 rewrite ^ /foo break;

can be coded in Lua as


 ngx.req.set_uri("/foo", false)

or equivalently,


 ngx.req.set_uri("/foo")

The jump argument can only be set to true in rewrite_by_lua*. Use of jump in other contexts is prohibited and will throw out a Lua exception.

A more sophisticated example involving regex substitutions is as follows


 location /test {
     rewrite_by_lua_block {
         local uri = ngx.re.sub(ngx.var.uri, "^/test/(.*)", "/$1", "o")
         ngx.req.set_uri(uri)
     }
     proxy_pass http://my_backend;
 }

which is functionally equivalent to


 location /test {
     rewrite ^/test/(.*) /$1 break;
     proxy_pass http://my_backend;
 }

Note that it is not possible to use this interface to rewrite URI arguments and that ngx.req.set_uri_args should be used for this instead. For instance, Nginx config


 rewrite ^ /foo?a=3? last;

can be coded as


 ngx.req.set_uri_args("a=3")
 ngx.req.set_uri("/foo", true)

or


 ngx.req.set_uri_args({a = 3})
 ngx.req.set_uri("/foo", true)

This interface was first introduced in the v0.3.1rc14 release.

Back to TOC

ngx.req.set_uri_args

syntax: ngx.req.set_uri_args(args)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*

Rewrite the current request's URI query arguments by the args argument. The args argument can be either a Lua string, as in


 ngx.req.set_uri_args("a=3&b=hello%20world")

or a Lua table holding the query arguments' key-value pairs, as in


 ngx.req.set_uri_args({ a = 3, b = "hello world" })

where in the latter case, this method will escape argument keys and values according to the URI escaping rule.

Multi-value arguments are also supported:


 ngx.req.set_uri_args({ a = 3, b = {5, 6} })

which will result in a query string like a=3&b=5&b=6.

This interface was first introduced in the v0.3.1rc13 release.

See also ngx.req.set_uri.

Back to TOC

ngx.req.get_uri_args

syntax: args = ngx.req.get_uri_args(max_args?)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, balancer_by_lua*

Returns a Lua table holding all the current request URL query arguments.


 location = /test {
     content_by_lua_block {
         local args = ngx.req.get_uri_args()
         for key, val in pairs(args) do
             if type(val) == "table" then
                 ngx.say(key, ": ", table.concat(val, ", "))
             else
                 ngx.say(key, ": ", val)
             end
         end
     }
 }

Then GET /test?foo=bar&bar=baz&bar=blah will yield the response body


 foo: bar
 bar: baz, blah

Multiple occurrences of an argument key will result in a table value holding all the values for that key in order.

Keys and values are unescaped according to URI escaping rules. In the settings above, GET /test?a%20b=1%61+2 will yield:


 a b: 1a 2

Arguments without the =<value> parts are treated as boolean arguments. GET /test?foo&bar will yield:


 foo: true
 bar: true

That is, they will take Lua boolean values true. However, they are different from arguments taking empty string values. GET /test?foo=&bar= will give something like


 foo:
 bar:

Empty key arguments are discarded. GET /test?=hello&=world will yield an empty output for instance.

Updating query arguments via the nginx variable $args (or ngx.var.args in Lua) at runtime is also supported:


 ngx.var.args = "a=3&b=42"
 local args = ngx.req.get_uri_args()

Here the args table will always look like


 {a = 3, b = 42}

regardless of the actual request query string.

Note that a maximum of 100 request arguments are parsed by default (including those with the same name) and that additional request arguments are silently discarded to guard against potential denial of service attacks.

However, the optional max_args function argument can be used to override this limit:


 local args = ngx.req.get_uri_args(10)

This argument can be set to zero to remove the limit and to process all request arguments received:


 local args = ngx.req.get_uri_args(0)

Removing the max_args cap is strongly discouraged.

Back to TOC

ngx.req.get_post_args

syntax: args, err = ngx.req.get_post_args(max_args?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*

Returns a Lua table holding all the current request POST query arguments (of the MIME type application/x-www-form-urlencoded). Call ngx.req.read_body to read the request body first or turn on the lua_need_request_body directive to avoid errors.


 location = /test {
     content_by_lua_block {
         ngx.req.read_body()
         local args, err = ngx.req.get_post_args()
         if not args then
             ngx.say("failed to get post args: ", err)
             return
         end
         for key, val in pairs(args) do
             if type(val) == "table" then
                 ngx.say(key, ": ", table.concat(val, ", "))
             else
                 ngx.say(key, ": ", val)
             end
         end
     }
 }

Then


 # Post request with the body 'foo=bar&bar=baz&bar=blah'
 $ curl --data 'foo=bar&bar=baz&bar=blah' localhost/test

will yield the response body like


 foo: bar
 bar: baz, blah

Multiple occurrences of an argument key will result in a table value holding all of the values for that key in order.

Keys and values will be unescaped according to URI escaping rules.

With the settings above,


 # POST request with body 'a%20b=1%61+2'
 $ curl -d 'a%20b=1%61+2' localhost/test

will yield:


 a b: 1a 2

Arguments without the =<value> parts are treated as boolean arguments. POST /test with the request body foo&bar will yield:


 foo: true
 bar: true

That is, they will take Lua boolean values true. However, they are different from arguments taking empty string values. POST /test with request body foo=&bar= will return something like


 foo:
 bar:

Empty key arguments are discarded. POST /test with body =hello&=world will yield empty outputs for instance.

Note that a maximum of 100 request arguments are parsed by default (including those with the same name) and that additional request arguments are silently discarded to guard against potential denial of service attacks.

However, the optional max_args function argument can be used to override this limit:


 local args = ngx.req.get_post_args(10)

This argument can be set to zero to remove the limit and to process all request arguments received:


 local args = ngx.req.get_post_args(0)

Removing the max_args cap is strongly discouraged.

Back to TOC

ngx.req.get_headers

syntax: headers = ngx.req.get_headers(max_headers?, raw?)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*

Returns a Lua table holding all the current request headers.


 local h = ngx.req.get_headers()
 for k, v in pairs(h) do
     ...
 end

To read an individual header:


 ngx.say("Host: ", ngx.req.get_headers()["Host"])

Note that the ngx.var.HEADER API call, which uses core $http_HEADER variables, may be more preferable for reading individual request headers.

For multiple instances of request headers such as:


 Foo: foo
 Foo: bar
 Foo: baz

the value of ngx.req.get_headers()["Foo"] will be a Lua (array) table such as:


 {"foo", "bar", "baz"}

Note that a maximum of 100 request headers are parsed by default (including those with the same name) and that additional request headers are silently discarded to guard against potential denial of service attacks.

However, the optional max_headers function argument can be used to override this limit:


 local headers = ngx.req.get_headers(10)

This argument can be set to zero to remove the limit and to process all request headers received:


 local headers = ngx.req.get_headers(0)

Removing the max_headers cap is strongly discouraged.

Since the 0.6.9 release, all the header names in the Lua table returned are converted to the pure lower-case form by default, unless the raw argument is set to true (default to false).

Also, by default, an __index metamethod is added to the resulting Lua table and will normalize the keys to a pure lowercase form with all underscores converted to dashes in case of a lookup miss. For example, if a request header My-Foo-Header is present, then the following invocations will all pick up the value of this header correctly:


 ngx.say(headers.my_foo_header)
 ngx.say(headers["My-Foo-Header"])
 ngx.say(headers["my-foo-header"])

The __index metamethod will not be added when the raw argument is set to true.

Back to TOC

ngx.req.set_header

syntax: ngx.req.set_header(header_name, header_value)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*

Set the current request's request header named header_name to value header_value, overriding any existing ones.

By default, all the subrequests subsequently initiated by ngx.location.capture and ngx.location.capture_multi will inherit the new header.

Here is an example of setting the Content-Type header:


 ngx.req.set_header("Content-Type", "text/css")

The header_value can take an array list of values, for example,


 ngx.req.set_header("Foo", {"a", "abc"})

will produce two new request headers:


 Foo: a
 Foo: abc

and old Foo headers will be overridden if there is any.

When the header_value argument is nil, the request header will be removed. So


 ngx.req.set_header("X-Foo", nil)

is equivalent to


 ngx.req.clear_header("X-Foo")

Back to TOC

ngx.req.clear_header

syntax: ngx.req.clear_header(header_name)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*

Clears the current request's request header named header_name. None of the current request's existing subrequests will be affected but subsequently initiated subrequests will inherit the change by default.

Back to TOC

ngx.req.read_body

syntax: ngx.req.read_body()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Reads the client request body synchronously without blocking the Nginx event loop.


 ngx.req.read_body()
 local args = ngx.req.get_post_args()

If the request body is already read previously by turning on lua_need_request_body or by using other modules, then this function does not run and returns immediately.

If the request body has already been explicitly discarded, either by the ngx.req.discard_body function or other modules, this function does not run and returns immediately.

In case of errors, such as connection errors while reading the data, this method will throw out a Lua exception or terminate the current request with a 500 status code immediately.

The request body data read using this function can be retrieved later via ngx.req.get_body_data or, alternatively, the temporary file name for the body data cached to disk using ngx.req.get_body_file. This depends on

  1. whether the current request body is already larger than the client_body_buffer_size,
  2. and whether client_body_in_file_only has been switched on.

In cases where current request may have a request body and the request body data is not required, The ngx.req.discard_body function must be used to explicitly discard the request body to avoid breaking things under HTTP 1.1 keepalive or HTTP 1.1 pipelining.

This function was first introduced in the v0.3.1rc17 release.

Back to TOC

ngx.req.discard_body

syntax: ngx.req.discard_body()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Explicitly discard the request body, i.e., read the data on the connection and throw it away immediately (without using the request body by any means).

This function is an asynchronous call and returns immediately.

If the request body has already been read, this function does nothing and returns immediately.

This function was first introduced in the v0.3.1rc17 release.

See also ngx.req.read_body.

Back to TOC

ngx.req.get_body_data

syntax: data = ngx.req.get_body_data()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, log_by_lua*

Retrieves in-memory request body data. It returns a Lua string rather than a Lua table holding all the parsed query arguments. Use the ngx.req.get_post_args function instead if a Lua table is required.

This function returns nil if

  1. the request body has not been read,
  2. the request body has been read into disk temporary files,
  3. or the request body has zero size.

If the request body has not been read yet, call ngx.req.read_body first (or turned on lua_need_request_body to force this module to read the request body. This is not recommended however).

If the request body has been read into disk files, try calling the ngx.req.get_body_file function instead.

To force in-memory request bodies, try setting client_body_buffer_size to the same size value in client_max_body_size.

Note that calling this function instead of using ngx.var.request_body or ngx.var.echo_request_body is more efficient because it can save one dynamic memory allocation and one data copy.

This function was first introduced in the v0.3.1rc17 release.

See also ngx.req.get_body_file.

Back to TOC

ngx.req.get_body_file

syntax: file_name = ngx.req.get_body_file()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Retrieves the file name for the in-file request body data. Returns nil if the request body has not been read or has been read into memory.

The returned file is read only and is usually cleaned up by Nginx's memory pool. It should not be manually modified, renamed, or removed in Lua code.

If the request body has not been read yet, call ngx.req.read_body first (or turned on lua_need_request_body to force this module to read the request body. This is not recommended however).

If the request body has been read into memory, try calling the ngx.req.get_body_data function instead.

To force in-file request bodies, try turning on client_body_in_file_only.

This function was first introduced in the v0.3.1rc17 release.

See also ngx.req.get_body_data.

Back to TOC

ngx.req.set_body_data

syntax: ngx.req.set_body_data(data)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Set the current request's request body using the in-memory data specified by the data argument.

If the current request's request body has not been read, then it will be properly discarded. When the current request's request body has been read into memory or buffered into a disk file, then the old request body's memory will be freed or the disk file will be cleaned up immediately, respectively.

This function was first introduced in the v0.3.1rc18 release.

See also ngx.req.set_body_file.

Back to TOC

ngx.req.set_body_file

syntax: ngx.req.set_body_file(file_name, auto_clean?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Set the current request's request body using the in-file data specified by the file_name argument.

If the optional auto_clean argument is given a true value, then this file will be removed at request completion or the next time this function or ngx.req.set_body_data are called in the same request. The auto_clean is default to false.

Please ensure that the file specified by the file_name argument exists and is readable by an Nginx worker process by setting its permission properly to avoid Lua exception errors.

If the current request's request body has not been read, then it will be properly discarded. When the current request's request body has been read into memory or buffered into a disk file, then the old request body's memory will be freed or the disk file will be cleaned up immediately, respectively.

This function was first introduced in the v0.3.1rc18 release.

See also ngx.req.set_body_data.

Back to TOC

ngx.req.init_body

syntax: ngx.req.init_body(buffer_size?)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*

Creates a new blank request body for the current request and inializes the buffer for later request body data writing via the ngx.req.append_body and ngx.req.finish_body APIs.

If the buffer_size argument is specified, then its value will be used for the size of the memory buffer for body writing with ngx.req.append_body. If the argument is omitted, then the value specified by the standard client_body_buffer_size directive will be used instead.

When the data can no longer be hold in the memory buffer for the request body, then the data will be flushed onto a temporary file just like the standard request body reader in the Nginx core.

It is important to always call the ngx.req.finish_body after all the data has been appended onto the current request body. Also, when this function is used together with ngx.req.socket, it is required to call ngx.req.socket before this function, or you will get the "request body already exists" error message.

The usage of this function is often like this:


 ngx.req.init_body(128 * 1024)  -- buffer is 128KB
 for chunk in next_data_chunk() do
     ngx.req.append_body(chunk) -- each chunk can be 4KB
 end
 ngx.req.finish_body()

This function can be used with ngx.req.append_body, ngx.req.finish_body, and ngx.req.socket to implement efficient input filters in pure Lua (in the context of rewrite_by_lua* or access_by_lua*), which can be used with other Nginx content handler or upstream modules like ngx_http_proxy_module and ngx_http_fastcgi_module.

This function was first introduced in the v0.5.11 release.

Back to TOC

ngx.req.append_body

syntax: ngx.req.append_body(data_chunk)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*

Append new data chunk specified by the data_chunk argument onto the existing request body created by the ngx.req.init_body call.

When the data can no longer be hold in the memory buffer for the request body, then the data will be flushed onto a temporary file just like the standard request body reader in the Nginx core.

It is important to always call the ngx.req.finish_body after all the data has been appended onto the current request body.

This function can be used with ngx.req.init_body, ngx.req.finish_body, and ngx.req.socket to implement efficient input filters in pure Lua (in the context of rewrite_by_lua* or access_by_lua*), which can be used with other Nginx content handler or upstream modules like ngx_http_proxy_module and ngx_http_fastcgi_module.

This function was first introduced in the v0.5.11 release.

See also ngx.req.init_body.

Back to TOC

ngx.req.finish_body

syntax: ngx.req.finish_body()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*

Completes the construction process of the new request body created by the ngx.req.init_body and ngx.req.append_body calls.

This function can be used with ngx.req.init_body, ngx.req.append_body, and ngx.req.socket to implement efficient input filters in pure Lua (in the context of rewrite_by_lua* or access_by_lua*), which can be used with other Nginx content handler or upstream modules like ngx_http_proxy_module and ngx_http_fastcgi_module.

This function was first introduced in the v0.5.11 release.

See also ngx.req.init_body.

Back to TOC

ngx.req.socket

syntax: tcpsock, err = ngx.req.socket()

syntax: tcpsock, err = ngx.req.socket(raw)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Returns a read-only cosocket object that wraps the downstream connection. Only receive and receiveuntil methods are supported on this object.

In case of error, nil will be returned as well as a string describing the error.

The socket object returned by this method is usually used to read the current request's body in a streaming fashion. Do not turn on the lua_need_request_body directive, and do not mix this call with ngx.req.read_body and ngx.req.discard_body.

If any request body data has been pre-read into the Nginx core request header buffer, the resulting cosocket object will take care of this to avoid potential data loss resulting from such pre-reading. Chunked request bodies are not yet supported in this API.

Since the v0.9.0 release, this function accepts an optional boolean raw argument. When this argument is true, this function returns a full-duplex cosocket object wrapping around the raw downstream connection socket, upon which you can call the receive, receiveuntil, and send methods.

When the raw argument is true, it is required that no pending data from any previous ngx.say, ngx.print, or ngx.send_headers calls exists. So if you have these downstream output calls previously, you should call ngx.flush(true) before calling ngx.req.socket(true) to ensure that there is no pending output data. If the request body has not been read yet, then this "raw socket" can also be used to read the request body.

You can use the "raw request socket" returned by ngx.req.socket(true) to implement fancy protocols like WebSocket, or just emit your own raw HTTP response header or body data. You can refer to the lua-resty-websocket library for a real world example.

This function was first introduced in the v0.5.0rc1 release.

Back to TOC

ngx.exec

syntax: ngx.exec(uri, args?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Does an internal redirect to uri with args and is similar to the echo_exec directive of the echo-nginx-module.


 ngx.exec('/some-location');
 ngx.exec('/some-location', 'a=3&b=5&c=6');
 ngx.exec('/some-location?a=3&b=5', 'c=6');

The optional second args can be used to specify extra URI query arguments, for example:


 ngx.exec("/foo", "a=3&b=hello%20world")

Alternatively, a Lua table can be passed for the args argument for ngx_lua to carry out URI escaping and string concatenation.


 ngx.exec("/foo", { a = 3, b = "hello world" })

The result is exactly the same as the previous example.

The format for the Lua table passed as the args argument is identical to the format used in the ngx.encode_args method.

Named locations are also supported but the second args argument will be ignored if present and the querystring for the new target is inherited from the referring location (if any).

GET /foo/file.php?a=hello will return "hello" and not "goodbye" in the example below


 location /foo {
     content_by_lua_block {
         ngx.exec("@bar", "a=goodbye");
     }
 }

 location @bar {
     content_by_lua_block {
         local args = ngx.req.get_uri_args()
         for key, val in pairs(args) do
             if key == "a" then
                 ngx.say(val)
             end
         end
     }
 }

Note that the ngx.exec method is different from ngx.redirect in that it is purely an internal redirect and that no new external HTTP traffic is involved.

Also note that this method call terminates the processing of the current request and that it must be called before ngx.send_headers or explicit response body outputs by either ngx.print or ngx.say.

It is recommended that a coding style that combines this method call with the return statement, i.e., return ngx.exec(...) be adopted when this method call is used in contexts other than header_filter_by_lua* to reinforce the fact that the request processing is being terminated.

Back to TOC

ngx.redirect

syntax: ngx.redirect(uri, status?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Issue an HTTP 301 or 302 redirection to uri.

The optional status parameter specifies the HTTP status code to be used. The following status codes are supported right now:

  • 301
  • 302 (default)
  • 307

It is 302 (ngx.HTTP_MOVED_TEMPORARILY) by default.

Here is an example assuming the current server name is localhost and that it is listening on port 1984:


 return ngx.redirect("/foo")

which is equivalent to


 return ngx.redirect("/foo", ngx.HTTP_MOVED_TEMPORARILY)

Redirecting arbitrary external URLs is also supported, for example:


 return ngx.redirect("http://www.google.com")

We can also use the numerical code directly as the second status argument:


 return ngx.redirect("/foo", 301)

This method is similar to the rewrite directive with the redirect modifier in the standard ngx_http_rewrite_module, for example, this nginx.conf snippet


 rewrite ^ /foo? redirect;  # nginx config

is equivalent to the following Lua code


 return ngx.redirect('/foo');  -- Lua code

while


 rewrite ^ /foo? permanent;  # nginx config

is equivalent to


 return ngx.redirect('/foo', ngx.HTTP_MOVED_PERMANENTLY)  -- Lua code

URI arguments can be specified as well, for example:


 return ngx.redirect('/foo?a=3&b=4')

Note that this method call terminates the processing of the current request and that it must be called before ngx.send_headers or explicit response body outputs by either ngx.print or ngx.say.

It is recommended that a coding style that combines this method call with the return statement, i.e., return ngx.redirect(...) be adopted when this method call is used in contexts other than header_filter_by_lua* to reinforce the fact that the request processing is being terminated.

Back to TOC

ngx.send_headers

syntax: ok, err = ngx.send_headers()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Explicitly send out the response headers.

Since v0.8.3 this function returns 1 on success, or returns nil and a string describing the error otherwise.

Note that there is normally no need to manually send out response headers as ngx_lua will automatically send headers out before content is output with ngx.say or ngx.print or when content_by_lua* exits normally.

Back to TOC

ngx.headers_sent

syntax: value = ngx.headers_sent

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*

Returns true if the response headers have been sent (by ngx_lua), and false otherwise.

This API was first introduced in ngx_lua v0.3.1rc6.

Back to TOC

ngx.print

syntax: ok, err = ngx.print(...)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Emits arguments concatenated to the HTTP client (as response body). If response headers have not been sent, this function will send headers out first and then output body data.

Since v0.8.3 this function returns 1 on success, or returns nil and a string describing the error otherwise.

Lua nil values will output "nil" strings and Lua boolean values will output "true" and "false" literal strings respectively.

Nested arrays of strings are permitted and the elements in the arrays will be sent one by one:


 local table = {
     "hello, ",
     {"world: ", true, " or ", false,
         {": ", nil}}
 }
 ngx.print(table)

will yield the output


 hello, world: true or false: nil

Non-array table arguments will cause a Lua exception to be thrown.

The ngx.null constant will yield the "null" string output.

This is an asynchronous call and will return immediately without waiting for all the data to be written into the system send buffer. To run in synchronous mode, call ngx.flush(true) after calling ngx.print. This can be particularly useful for streaming output. See ngx.flush for more details.

Please note that both ngx.print and ngx.say will always invoke the whole Nginx output body filter chain, which is an expensive operation. So be careful when calling either of these two in a tight loop; buffer the data yourself in Lua and save the calls.

Back to TOC

ngx.say

syntax: ok, err = ngx.say(...)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Just as ngx.print but also emit a trailing newline.

Back to TOC

ngx.log

syntax: ngx.log(log_level, ...)

context: init_by_lua*, init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Log arguments concatenated to error.log with the given logging level.

Lua nil arguments are accepted and result in literal "nil" string while Lua booleans result in literal "true" or "false" string outputs. And the ngx.null constant will yield the "null" string output.

The log_level argument can take constants like ngx.ERR and ngx.WARN. Check out Nginx log level constants for details.

There is a hard coded 2048 byte limitation on error message lengths in the Nginx core. This limit includes trailing newlines and leading time stamps. If the message size exceeds this limit, Nginx will truncate the message text accordingly. This limit can be manually modified by editing the NGX_MAX_ERROR_STR macro definition in the src/core/ngx_log.h file in the Nginx source tree.

Back to TOC

ngx.flush

syntax: ok, err = ngx.flush(wait?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Flushes response output to the client.

ngx.flush accepts an optional boolean wait argument (Default: false) first introduced in the v0.3.1rc34 release. When called with the default argument, it issues an asynchronous call (Returns immediately without waiting for output data to be written into the system send buffer). Calling the function with the wait argument set to true switches to synchronous mode.

In synchronous mode, the function will not return until all output data has been written into the system send buffer or until the send_timeout setting has expired. Note that using the Lua coroutine mechanism means that this function does not block the Nginx event loop even in the synchronous mode.

When ngx.flush(true) is called immediately after ngx.print or ngx.say, it causes the latter functions to run in synchronous mode. This can be particularly useful for streaming output.

Note that ngx.flush is not functional when in the HTTP 1.0 output buffering mode. See HTTP 1.0 support.

Since v0.8.3 this function returns 1 on success, or returns nil and a string describing the error otherwise.

Back to TOC

ngx.exit

syntax: ngx.exit(status)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

When status >= 200 (i.e., ngx.HTTP_OK and above), it will interrupt the execution of the current request and return status code to nginx.

When status == 0 (i.e., ngx.OK), it will only quit the current phase handler (or the content handler if the content_by_lua* directive is used) and continue to run later phases (if any) for the current request.

The status argument can be ngx.OK, ngx.ERROR, ngx.HTTP_NOT_FOUND, ngx.HTTP_MOVED_TEMPORARILY, or other HTTP status constants.

To return an error page with custom contents, use code snippets like this:


 ngx.status = ngx.HTTP_GONE
 ngx.say("This is our own content")
 -- to cause quit the whole request rather than the current phase handler
 ngx.exit(ngx.HTTP_OK)

The effect in action:


 $ curl -i http://localhost/test
 HTTP/1.1 410 Gone
 Server: nginx/1.0.6
 Date: Thu, 15 Sep 2011 00:51:48 GMT
 Content-Type: text/plain
 Transfer-Encoding: chunked
 Connection: keep-alive

 This is our own content

Number literals can be used directly as the argument, for instance,


 ngx.exit(501)

Note that while this method accepts all HTTP status constants as input, it only accepts NGX_OK and NGX_ERROR of the core constants.

Also note that this method call terminates the processing of the current request and that it is recommended that a coding style that combines this method call with the return statement, i.e., return ngx.exit(...) be used to reinforce the fact that the request processing is being terminated.

When being used in the contexts of header_filter_by_lua and ssl_session_store_by_lua*, ngx.exit() is an asynchronous operation and will return immediately. This behavior may change in future and it is recommended that users always use return in combination as suggested above.

Back to TOC

ngx.eof

syntax: ok, err = ngx.eof()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Explicitly specify the end of the response output stream. In the case of HTTP 1.1 chunked encoded output, it will just trigger the Nginx core to send out the "last chunk".

When you disable the HTTP 1.1 keep-alive feature for your downstream connections, you can rely on descent HTTP clients to close the connection actively for you when you call this method. This trick can be used do back-ground jobs without letting the HTTP clients to wait on the connection, as in the following example:


 location = /async {
     keepalive_timeout 0;
     content_by_lua_block {
         ngx.say("got the task!")
         ngx.eof()  -- a descent HTTP client will close the connection at this point
         -- access MySQL, PostgreSQL, Redis, Memcached, and etc here...
     }
 }

But if you create subrequests to access other locations configured by Nginx upstream modules, then you should configure those upstream modules to ignore client connection abortions if they are not by default. For example, by default the standard ngx_http_proxy_module will terminate both the subrequest and the main request as soon as the client closes the connection, so it is important to turn on the proxy_ignore_client_abort directive in your location block configured by ngx_http_proxy_module:


 proxy_ignore_client_abort on;

A better way to do background jobs is to use the ngx.timer.at API.

Since v0.8.3 this function returns 1 on success, or returns nil and a string describing the error otherwise.

Back to TOC

ngx.sleep

syntax: ngx.sleep(seconds)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Sleeps for the specified seconds without blocking. One can specify time resolution up to 0.001 seconds (i.e., one milliseconds).

Behind the scene, this method makes use of the Nginx timers.

Since the 0.7.20 release, The 0 time argument can also be specified.

This method was introduced in the 0.5.0rc30 release.

Back to TOC

ngx.escape_uri

syntax: newstr = ngx.escape_uri(str)

context: init_by_lua*, init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Escape str as a URI component.

Back to TOC

ngx.unescape_uri

syntax: newstr = ngx.unescape_uri(str)

context: init_by_lua*, init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*

Unescape str as an escaped URI component.

For example,


 ngx.say(ngx.unescape_uri("b%20r56+7"))

gives the output

b r56 7

Back to TOC

ngx.encode_args

syntax: str = ngx.encode_args(table)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*

Encode the Lua table to a query args string according to the URI encoded rules.

For example,


 ngx.encode_args({foo = 3, ["b r"] = "hello world"})

yields

foo=3&b%20r=hello%20world

The table keys must be Lua strings.

Multi-value query args are also supported. Just use a Lua table for the argument's value, for example:


 ngx.encode_args({baz = {32, "hello"}})

gives

baz=32&baz=hello

If the value table is empty and the effect is equivalent to the nil value.

Boolean argument values are also supported, for instance,


 ngx.encode_args({a = true, b = 1})

yields

a&b=1

If the argument value is false, then the effect is equivalent to the nil value.

This method was first introduced in the v0.3.1rc27 release.

Back to TOC

ngx.decode_args

syntax: table = ngx.decode_args(str, max_args?)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Decodes a URI encoded query-string into a Lua table. This is the inverse function of ngx.encode_args.

The optional max_args argument can be used to specify the maximum number of arguments parsed from the str argument. By default, a maximum of 100 request arguments are parsed (including those with the same name) and that additional URI arguments are silently discarded to guard against potential denial of service attacks.

This argument can be set to zero to remove the limit and to process all request arguments received:


 local args = ngx.decode_args(str, 0)

Removing the max_args cap is strongly discouraged.

This method was introduced in the v0.5.0rc29.

Back to TOC

ngx.encode_base64

syntax: newstr = ngx.encode_base64(str, no_padding?)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Encodes str to a base64 digest.

Since the 0.9.16 release, an optional boolean-typed no_padding argument can be specified to control whether the base64 padding should be appended to the resulting digest (default to false, i.e., with padding enabled).

Back to TOC

ngx.decode_base64

syntax: newstr = ngx.decode_base64(str)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Decodes the str argument as a base64 digest to the raw form. Returns nil if str is not well formed.

Back to TOC

ngx.crc32_short

syntax: intval = ngx.crc32_short(str)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Calculates the CRC-32 (Cyclic Redundancy Code) digest for the str argument.

This method performs better on relatively short str inputs (i.e., less than 30 ~ 60 bytes), as compared to ngx.crc32_long. The result is exactly the same as ngx.crc32_long.

Behind the scene, it is just a thin wrapper around the ngx_crc32_short function defined in the Nginx core.

This API was first introduced in the v0.3.1rc8 release.

Back to TOC

ngx.crc32_long

syntax: intval = ngx.crc32_long(str)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Calculates the CRC-32 (Cyclic Redundancy Code) digest for the str argument.

This method performs better on relatively long str inputs (i.e., longer than 30 ~ 60 bytes), as compared to ngx.crc32_short. The result is exactly the same as ngx.crc32_short.

Behind the scene, it is just a thin wrapper around the ngx_crc32_long function defined in the Nginx core.

This API was first introduced in the v0.3.1rc8 release.

Back to TOC

ngx.hmac_sha1

syntax: digest = ngx.hmac_sha1(secret_key, str)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Computes the HMAC-SHA1 digest of the argument str and turns the result using the secret key <secret_key>.

The raw binary form of the HMAC-SHA1 digest will be generated, use ngx.encode_base64, for example, to encode the result to a textual representation if desired.

For example,


 local key = "thisisverysecretstuff"
 local src = "some string we want to sign"
 local digest = ngx.hmac_sha1(key, src)
 ngx.say(ngx.encode_base64(digest))

yields the output

R/pvxzHC4NLtj7S+kXFg/NePTmk=

This API requires the OpenSSL library enabled in the Nginx build (usually by passing the --with-http_ssl_module option to the ./configure script).

This function was first introduced in the v0.3.1rc29 release.

Back to TOC

ngx.md5

syntax: digest = ngx.md5(str)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns the hexadecimal representation of the MD5 digest of the str argument.

For example,


 location = /md5 {
     content_by_lua_block { ngx.say(ngx.md5("hello")) }
 }

yields the output

5d41402abc4b2a76b9719d911017c592

See ngx.md5_bin if the raw binary MD5 digest is required.

Back to TOC

ngx.md5_bin

syntax: digest = ngx.md5_bin(str)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns the binary form of the MD5 digest of the str argument.

See ngx.md5 if the hexadecimal form of the MD5 digest is required.

Back to TOC

ngx.sha1_bin

syntax: digest = ngx.sha1_bin(str)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns the binary form of the SHA-1 digest of the str argument.

This function requires SHA-1 support in the Nginx build. (This usually just means OpenSSL should be installed while building Nginx).

This function was first introduced in the v0.5.0rc6.

Back to TOC

ngx.quote_sql_str

syntax: quoted_value = ngx.quote_sql_str(raw_value)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns a quoted SQL string literal according to the MySQL quoting rules.

Back to TOC

ngx.today

syntax: str = ngx.today()

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns current date (in the format yyyy-mm-dd) from the nginx cached time (no syscall involved unlike Lua's date library).

This is the local time.

Back to TOC

ngx.time

syntax: secs = ngx.time()

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns the elapsed seconds from the epoch for the current time stamp from the nginx cached time (no syscall involved unlike Lua's date library).

Updates of the Nginx time cache an be forced by calling ngx.update_time first.

Back to TOC

ngx.now

syntax: secs = ngx.now()

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns a floating-point number for the elapsed time in seconds (including milliseconds as the decimal part) from the epoch for the current time stamp from the nginx cached time (no syscall involved unlike Lua's date library).

You can forcibly update the Nginx time cache by calling ngx.update_time first.

This API was first introduced in v0.3.1rc32.

Back to TOC

ngx.update_time

syntax: ngx.update_time()

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Forcibly updates the Nginx current time cache. This call involves a syscall and thus has some overhead, so do not abuse it.

This API was first introduced in v0.3.1rc32.

Back to TOC

ngx.localtime

syntax: str = ngx.localtime()

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns the current time stamp (in the format yyyy-mm-dd hh:mm:ss) of the nginx cached time (no syscall involved unlike Lua's os.date function).

This is the local time.

Back to TOC

ngx.utctime

syntax: str = ngx.utctime()

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns the current time stamp (in the format yyyy-mm-dd hh:mm:ss) of the nginx cached time (no syscall involved unlike Lua's os.date function).

This is the UTC time.

Back to TOC

ngx.cookie_time

syntax: str = ngx.cookie_time(sec)

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns a formatted string can be used as the cookie expiration time. The parameter sec is the time stamp in seconds (like those returned from ngx.time).


 ngx.say(ngx.cookie_time(1290079655))
     -- yields "Thu, 18-Nov-10 11:27:35 GMT"

Back to TOC

ngx.http_time

syntax: str = ngx.http_time(sec)

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns a formated string can be used as the http header time (for example, being used in Last-Modified header). The parameter sec is the time stamp in seconds (like those returned from ngx.time).


 ngx.say(ngx.http_time(1290079655))
     -- yields "Thu, 18 Nov 2010 11:27:35 GMT"

Back to TOC

ngx.parse_http_time

syntax: sec = ngx.parse_http_time(str)

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Parse the http time string (as returned by ngx.http_time) into seconds. Returns the seconds or nil if the input string is in bad forms.


 local time = ngx.parse_http_time("Thu, 18 Nov 2010 11:27:35 GMT")
 if time == nil then
     ...
 end

Back to TOC

ngx.is_subrequest

syntax: value = ngx.is_subrequest

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*

Returns true if the current request is an nginx subrequest, or false otherwise.

Back to TOC

ngx.re.match

syntax: captures, err = ngx.re.match(subject, regex, options?, ctx?, res_table?)

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Matches the subject string using the Perl compatible regular expression regex with the optional options.

Only the first occurrence of the match is returned, or nil if no match is found. In case of errors, like seeing a bad regular expression or exceeding the PCRE stack limit, nil and a string describing the error will be returned.

When a match is found, a Lua table captures is returned, where captures[0] holds the whole substring being matched, and captures[1] holds the first parenthesized sub-pattern's capturing, captures[2] the second, and so on.


 local m, err = ngx.re.match("hello, 1234", "[0-9]+")
 if m then
     -- m[0] == "1234"

 else
     if err then
         ngx.log(ngx.ERR, "error: ", err)
         return
     end

     ngx.say("match not found")
 end

 local m, err = ngx.re.match("hello, 1234", "([0-9])[0-9]+")
 -- m[0] == "1234"
 -- m[1] == "1"

Named captures are also supported since the v0.7.14 release and are returned in the same Lua table as key-value pairs as the numbered captures.


 local m, err = ngx.re.match("hello, 1234", "([0-9])(?<remaining>[0-9]+)")
 -- m[0] == "1234"
 -- m[1] == "1"
 -- m[2] == "234"
 -- m["remaining"] == "234"

Unmatched subpatterns will have false values in their captures table fields.


 local m, err = ngx.re.match("hello, world", "(world)|(hello)|(?<named>howdy)")
 -- m[0] == "hello"
 -- m[1] == false
 -- m[2] == "hello"
 -- m[3] == false
 -- m["named"] == false

Specify options to control how the match operation will be performed. The following option characters are supported:

a             anchored mode (only match from the beginning)

d             enable the DFA mode (or the longest token match semantics).
              this requires PCRE 6.0+ or else a Lua exception will be thrown.
              first introduced in ngx_lua v0.3.1rc30.

D             enable duplicate named pattern support. This allows named
              subpattern names to be repeated, returning the captures in
              an array-like Lua table. for example,
                local m = ngx.re.match("hello, world",
                                       "(?<named>\w+), (?<named>\w+)",
                                       "D")
                -- m["named"] == {"hello", "world"}
              this option was first introduced in the v0.7.14 release.
              this option requires at least PCRE 8.12.

i             case insensitive mode (similar to Perl's /i modifier)

j             enable PCRE JIT compilation, this requires PCRE 8.21+ which
              must be built with the --enable-jit option. for optimum performance,
              this option should always be used together with the 'o' option.
              first introduced in ngx_lua v0.3.1rc30.

J             enable the PCRE Javascript compatible mode. this option was
              first introduced in the v0.7.14 release. this option requires
              at least PCRE 8.12.

m             multi-line mode (similar to Perl's /m modifier)

o             compile-once mode (similar to Perl's /o modifier),
              to enable the worker-process-level compiled-regex cache

s             single-line mode (similar to Perl's /s modifier)

u             UTF-8 mode. this requires PCRE to be built with
              the --enable-utf8 option or else a Lua exception will be thrown.

U             similar to "u" but disables PCRE's UTF-8 validity check on
              the subject string. first introduced in ngx_lua v0.8.1.

x             extended mode (similar to Perl's /x modifier)

These options can be combined:


 local m, err = ngx.re.match("hello, world", "HEL LO", "ix")
 -- m[0] == "hello"

 local m, err = ngx.re.match("hello, 美好生活", "HELLO, (.{2})", "iu")
 -- m[0] == "hello, 美好"
 -- m[1] == "美好"

The o option is useful for performance tuning, because the regex pattern in question will only be compiled once, cached in the worker-process level, and shared among all requests in the current Nginx worker process. The upper limit of the regex cache can be tuned via the lua_regex_cache_max_entries directive.

The optional fourth argument, ctx, can be a Lua table holding an optional pos field. When the pos field in the ctx table argument is specified, ngx.re.match will start matching from that offset (starting from 1). Regardless of the presence of the pos field in the ctx table, ngx.re.match will always set this pos field to the position after the substring matched by the whole pattern in case of a successful match. When match fails, the ctx table will be left intact.


 local ctx = {}
 local m, err = ngx.re.match("1234, hello", "[0-9]+", "", ctx)
      -- m[0] = "1234"
      -- ctx.pos == 5

 local ctx = { pos = 2 }
 local m, err = ngx.re.match("1234, hello", "[0-9]+", "", ctx)
      -- m[0] = "34"
      -- ctx.pos == 5

The ctx table argument combined with the a regex modifier can be used to construct a lexer atop ngx.re.match.

Note that, the options argument is not optional when the ctx argument is specified and that the empty Lua string ("") must be used as placeholder for options if no meaningful regex options are required.

This method requires the PCRE library enabled in Nginx. (Known Issue With Special Escaping Sequences).

To confirm that PCRE JIT is enabled, activate the Nginx debug log by adding the --with-debug option to Nginx or OpenResty's ./configure script. Then, enable the "debug" error log level in error_log directive. The following message will be generated if PCRE JIT is enabled:

pcre JIT compiling result: 1

Starting from the 0.9.4 release, this function also accepts a 5th argument, res_table, for letting the caller supply the Lua table used to hold all the capturing results. Starting from 0.9.6, it is the caller's responsibility to ensure this table is empty. This is very useful for recycling Lua tables and saving GC and table allocation overhead.

This feature was introduced in the v0.2.1rc11 release.

Back to TOC

ngx.re.find

syntax: from, to, err = ngx.re.find(subject, regex, options?, ctx?, nth?)

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Similar to ngx.re.match but only returns the beginning index (from) and end index (to) of the matched substring. The returned indexes are 1-based and can be fed directly into the string.sub API function to obtain the matched substring.

In case of errors (like bad regexes or any PCRE runtime errors), this API function returns two nil values followed by a string describing the error.

If no match is found, this function just returns a nil value.

Below is an example:


 local s = "hello, 1234"
 local from, to, err = ngx.re.find(s, "([0-9]+)", "jo")
 if from then
     ngx.say("from: ", from)
     ngx.say("to: ", to)
     ngx.say("matched: ", string.sub(s, from, to))
 else
     if err then
         ngx.say("error: ", err)
         return
     end
     ngx.say("not matched!")
 end

This example produces the output

from: 8
to: 11
matched: 1234

Because this API function does not create new Lua strings nor new Lua tables, it is much faster than ngx.re.match. It should be used wherever possible.

Since the 0.9.3 release, an optional 5th argument, nth, is supported to specify which (submatch) capture's indexes to return. When nth is 0 (which is the default), the indexes for the whole matched substring is returned; when nth is 1, then the 1st submatch capture's indexes are returned; when nth is 2, then the 2nd submatch capture is returned, and so on. When the specified submatch does not have a match, then two nil values will be returned. Below is an example for this:


 local str = "hello, 1234"
 local from, to = ngx.re.find(str, "([0-9])([0-9]+)", "jo", nil, 2)
 if from then
     ngx.say("matched 2nd submatch: ", string.sub(str, from, to))  -- yields "234"
 end

This API function was first introduced in the v0.9.2 release.

Back to TOC

ngx.re.gmatch

syntax: iterator, err = ngx.re.gmatch(subject, regex, options?)

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Similar to ngx.re.match, but returns a Lua iterator instead, so as to let the user programmer iterate all the matches over the <subject> string argument with the PCRE regex.

In case of errors, like seeing an ill-formed regular expression, nil and a string describing the error will be returned.

Here is a small example to demonstrate its basic usage:


 local iterator, err = ngx.re.gmatch("hello, world!", "([a-z]+)", "i")
 if not iterator then
     ngx.log(ngx.ERR, "error: ", err)
     return
 end

 local m
 m, err = iterator()    -- m[0] == m[1] == "hello"
 if err then
     ngx.log(ngx.ERR, "error: ", err)
     return
 end

 m, err = iterator()    -- m[0] == m[1] == "world"
 if err then
     ngx.log(ngx.ERR, "error: ", err)
     return
 end

 m, err = iterator()    -- m == nil
 if err then
     ngx.log(ngx.ERR, "error: ", err)
     return
 end

More often we just put it into a Lua loop:


 local it, err = ngx.re.gmatch("hello, world!", "([a-z]+)", "i")
 if not it then
     ngx.log(ngx.ERR, "error: ", err)
     return
 end

 while true do
     local m, err = it()
     if err then
         ngx.log(ngx.ERR, "error: ", err)
         return
     end

     if not m then
         -- no match found (any more)
         break
     end

     -- found a match
     ngx.say(m[0])
     ngx.say(m[1])
 end

The optional options argument takes exactly the same semantics as the ngx.re.match method.

The current implementation requires that the iterator returned should only be used in a single request. That is, one should not assign it to a variable belonging to persistent namespace like a Lua package.

This method requires the PCRE library enabled in Nginx. (Known Issue With Special Escaping Sequences).

This feature was first introduced in the v0.2.1rc12 release.

Back to TOC

ngx.re.sub

syntax: newstr, n, err = ngx.re.sub(subject, regex, replace, options?)

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Substitutes the first match of the Perl compatible regular expression regex on the subject argument string with the string or function argument replace. The optional options argument has exactly the same meaning as in ngx.re.match.

This method returns the resulting new string as well as the number of successful substitutions. In case of failures, like syntax errors in the regular expressions or the <replace> string argument, it will return nil and a string describing the error.

When the replace is a string, then it is treated as a special template for string replacement. For example,


 local newstr, n, err = ngx.re.sub("hello, 1234", "([0-9])[0-9]", "[$0][$1]")
 if newstr then
     -- newstr == "hello, [12][1]34"
     -- n == 1
 else
     ngx.log(ngx.ERR, "error: ", err)
     return
 end

where $0 referring to the whole substring matched by the pattern and $1 referring to the first parenthesized capturing substring.

Curly braces can also be used to disambiguate variable names from the background string literals:


 local newstr, n, err = ngx.re.sub("hello, 1234", "[0-9]", "${0}00")
     -- newstr == "hello, 100234"
     -- n == 1

Literal dollar sign characters ($) in the replace string argument can be escaped by another dollar sign, for instance,


 local newstr, n, err = ngx.re.sub("hello, 1234", "[0-9]", "$$")
     -- newstr == "hello, $234"
     -- n == 1

Do not use backlashes to escape dollar signs; it will not work as expected.

When the replace argument is of type "function", then it will be invoked with the "match table" as the argument to generate the replace string literal for substitution. The "match table" fed into the replace function is exactly the same as the return value of ngx.re.match. Here is an example:


 local func = function (m)
     return "[" .. m[0] .. "][" .. m[1] .. "]"
 end
 local newstr, n, err = ngx.re.sub("hello, 1234", "( [0-9] ) [0-9]", func, "x")
     -- newstr == "hello, [12][1]34"
     -- n == 1

The dollar sign characters in the return value of the replace function argument are not special at all.

This method requires the PCRE library enabled in Nginx. (Known Issue With Special Escaping Sequences).

This feature was first introduced in the v0.2.1rc13 release.

Back to TOC

ngx.re.gsub

syntax: newstr, n, err = ngx.re.gsub(subject, regex, replace, options?)

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Just like ngx.re.sub, but does global substitution.

Here is some examples:


 local newstr, n, err = ngx.re.gsub("hello, world", "([a-z])[a-z]+", "[$0,$1]", "i")
 if newstr then
     -- newstr == "[hello,h], [world,w]"
     -- n == 2
 else
     ngx.log(ngx.ERR, "error: ", err)
     return
 end

 local func = function (m)
     return "[" .. m[0] .. "," .. m[1] .. "]"
 end
 local newstr, n, err = ngx.re.gsub("hello, world", "([a-z])[a-z]+", func, "i")
     -- newstr == "[hello,h], [world,w]"
     -- n == 2

This method requires the PCRE library enabled in Nginx. (Known Issue With Special Escaping Sequences).

This feature was first introduced in the v0.2.1rc15 release.

Back to TOC

ngx.shared.DICT

syntax: dict = ngx.shared.DICT

syntax: dict = ngx.shared[name_var]

context: init_by_lua*, init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Fetching the shm-based Lua dictionary object for the shared memory zone named DICT defined by the lua_shared_dict directive.

Shared memory zones are always shared by all the nginx worker processes in the current nginx server instance.

The resulting object dict has the following methods:

Here is an example:


 http {
     lua_shared_dict dogs 10m;
     server {
         location /set {
             content_by_lua_block {
                 local dogs = ngx.shared.dogs
                 dogs:set("Jim", 8)
                 ngx.say("STORED")
             }
         }
         location /get {
             content_by_lua_block {
                 local dogs = ngx.shared.dogs
                 ngx.say(dogs:get("Jim"))
             }
         }
     }
 }

Let us test it:


 $ curl localhost/set
 STORED

 $ curl localhost/get
 8

 $ curl localhost/get
 8

The number 8 will be consistently output when accessing /get regardless of how many Nginx workers there are because the dogs dictionary resides in the shared memory and visible to all of the worker processes.

The shared dictionary will retain its contents through a server config reload (either by sending the HUP signal to the Nginx process or by using the -s reload command-line option).

The contents in the dictionary storage will be lost, however, when the Nginx server quits.

This feature was first introduced in the v0.3.1rc22 release.

Back to TOC

ngx.shared.DICT.get

syntax: value, flags = ngx.shared.DICT:get(key)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Retrieving the value in the dictionary ngx.shared.DICT for the key key. If the key does not exist or has expired, then nil will be returned.

In case of errors, nil and a string describing the error will be returned.

The value returned will have the original data type when they were inserted into the dictionary, for example, Lua booleans, numbers, or strings.

The first argument to this method must be the dictionary object itself, for example,


 local cats = ngx.shared.cats
 local value, flags = cats.get(cats, "Marry")

or use Lua's syntactic sugar for method calls:


 local cats = ngx.shared.cats
 local value, flags = cats:get("Marry")

These two forms are fundamentally equivalent.

If the user flags is 0 (the default), then no flags value will be returned.

This feature was first introduced in the v0.3.1rc22 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.get_stale

syntax: value, flags, stale = ngx.shared.DICT:get_stale(key)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Similar to the get method but returns the value even if the key has already expired.

Returns a 3rd value, stale, indicating whether the key has expired or not.

Note that the value of an expired key is not guaranteed to be available so one should never rely on the availability of expired items.

This method was first introduced in the 0.8.6 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.set

syntax: success, err, forcible = ngx.shared.DICT:set(key, value, exptime?, flags?)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Unconditionally sets a key-value pair into the shm-based dictionary ngx.shared.DICT. Returns three values:

  • success: boolean value to indicate whether the key-value pair is stored or not.
  • err: textual error message, can be "no memory".
  • forcible: a boolean value to indicate whether other valid items have been removed forcibly when out of storage in the shared memory zone.

The value argument inserted can be Lua booleans, numbers, strings, or nil. Their value type will also be stored into the dictionary and the same data type can be retrieved later via the get method.

The optional exptime argument specifies expiration time (in seconds) for the inserted key-value pair. The time resolution is 0.001 seconds. If the exptime takes the value 0 (which is the default), then the item will never expire.

The optional flags argument specifies a user flags value associated with the entry to be stored. It can also be retrieved later with the value. The user flags is stored as an unsigned 32-bit integer internally. Defaults to 0. The user flags argument was first introduced in the v0.5.0rc2 release.

When it fails to allocate memory for the current key-value item, then set will try removing existing items in the storage according to the Least-Recently Used (LRU) algorithm. Note that, LRU takes priority over expiration time here. If up to tens of existing items have been removed and the storage left is still insufficient (either due to the total capacity limit specified by lua_shared_dict or memory segmentation), then the err return value will be no memory and success will be false.

If this method succeeds in storing the current item by forcibly removing other not-yet-expired items in the dictionary via LRU, the forcible return value will be true. If it stores the item without forcibly removing other valid items, then the return value forcible will be false.

The first argument to this method must be the dictionary object itself, for example,


 local cats = ngx.shared.cats
 local succ, err, forcible = cats.set(cats, "Marry", "it is a nice cat!")

or use Lua's syntactic sugar for method calls:


 local cats = ngx.shared.cats
 local succ, err, forcible = cats:set("Marry", "it is a nice cat!")

These two forms are fundamentally equivalent.

This feature was first introduced in the v0.3.1rc22 release.

Please note that while internally the key-value pair is set atomically, the atomicity does not go across the method call boundary.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.safe_set

syntax: ok, err = ngx.shared.DICT:safe_set(key, value, exptime?, flags?)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Similar to the set method, but never overrides the (least recently used) unexpired items in the store when running out of storage in the shared memory zone. In this case, it will immediately return nil and the string "no memory".

This feature was first introduced in the v0.7.18 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.add

syntax: success, err, forcible = ngx.shared.DICT:add(key, value, exptime?, flags?)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Just like the set method, but only stores the key-value pair into the dictionary ngx.shared.DICT if the key does not exist.

If the key argument already exists in the dictionary (and not expired for sure), the success return value will be false and the err return value will be "exists".

This feature was first introduced in the v0.3.1rc22 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.safe_add

syntax: ok, err = ngx.shared.DICT:safe_add(key, value, exptime?, flags?)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Similar to the add method, but never overrides the (least recently used) unexpired items in the store when running out of storage in the shared memory zone. In this case, it will immediately return nil and the string "no memory".

This feature was first introduced in the v0.7.18 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.replace

syntax: success, err, forcible = ngx.shared.DICT:replace(key, value, exptime?, flags?)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Just like the set method, but only stores the key-value pair into the dictionary ngx.shared.DICT if the key does exist.

If the key argument does not exist in the dictionary (or expired already), the success return value will be false and the err return value will be "not found".

This feature was first introduced in the v0.3.1rc22 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.delete

syntax: ngx.shared.DICT:delete(key)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Unconditionally removes the key-value pair from the shm-based dictionary ngx.shared.DICT.

It is equivalent to ngx.shared.DICT:set(key, nil).

This feature was first introduced in the v0.3.1rc22 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.incr

syntax: newval, err, forcible? = ngx.shared.DICT:incr(key, value, init?)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Increments the (numerical) value for key in the shm-based dictionary ngx.shared.DICT by the step value value. Returns the new resulting number if the operation is successfully completed or nil and an error message otherwise.

When the key does not exist or has already expired in the shared dictionary,

  1. if the init argument is not specified or takes the value nil, this method will return nil and the error string "not found", or
  2. if the init argument takes a number value, this method will create a new key with the value init + value.

Like the add method, it also overrides the (least recently used) unexpired items in the store when running out of storage in the shared memory zone.

The forcible return value will always be nil when the init argument is not specified.

If this method succeeds in storing the current item by forcibly removing other not-yet-expired items in the dictionary via LRU, the forcible return value will be true. If it stores the item without forcibly removing other valid items, then the return value forcible will be false.

If the original value is not a valid Lua number in the dictionary, it will return nil and "not a number".

The value argument and init argument can be any valid Lua numbers, like negative numbers or floating-point numbers.

This method was first introduced in the v0.3.1rc22 release.

The optional init parameter was first added in the v0.10.6 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.lpush

syntax: length, err = ngx.shared.DICT:lpush(key, value)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Inserts the specified (numerical or string) value at the head of the list named key in the shm-based dictionary ngx.shared.DICT. Returns the number of elements in the list after the push operation.

If key does not exist, it is created as an empty list before performing the push operation. When the key already takes a value that is not a list, it will return nil and "value not a list".

It never overrides the (least recently used) unexpired items in the store when running out of storage in the shared memory zone. In this case, it will immediately return nil and the string "no memory".

This feature was first introduced in the v0.10.6 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.rpush

syntax: length, err = ngx.shared.DICT:rpush(key, value)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Similar to the lpush method, but inserts the specified (numerical or string) value at the tail of the list named key.

This feature was first introduced in the v0.10.6 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.lpop

syntax: val, err = ngx.shared.DICT:lpop(key)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Removes and returns the first element of the list named key in the shm-based dictionary ngx.shared.DICT.

If key does not exist, it will return nil. When the key already takes a value that is not a list, it will return nil and "value not a list".

This feature was first introduced in the v0.10.6 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.rpop

syntax: val, err = ngx.shared.DICT:rpop(key)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Removes and returns the last element of the list named key in the shm-based dictionary ngx.shared.DICT.

If key does not exist, it will return nil. When the key already takes a value that is not a list, it will return nil and "value not a list".

This feature was first introduced in the v0.10.6 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.llen

syntax: len, err = ngx.shared.DICT:llen(key)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns the number of elements in the list named key in the shm-based dictionary ngx.shared.DICT.

If key does not exist, it is interpreted as an empty list and 0 is returned. When the key already takes a value that is not a list, it will return nil and "value not a list".

This feature was first introduced in the v0.10.6 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.flush_all

syntax: ngx.shared.DICT:flush_all()

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Flushes out all the items in the dictionary. This method does not actuall free up all the memory blocks in the dictionary but just marks all the existing items as expired.

This feature was first introduced in the v0.5.0rc17 release.

See also ngx.shared.DICT.flush_expired and ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.flush_expired

syntax: flushed = ngx.shared.DICT:flush_expired(max_count?)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Flushes out the expired items in the dictionary, up to the maximal number specified by the optional max_count argument. When the max_count argument is given 0 or not given at all, then it means unlimited. Returns the number of items that have actually been flushed.

Unlike the flush_all method, this method actually free up the memory used by the expired items.

This feature was first introduced in the v0.6.3 release.

See also ngx.shared.DICT.flush_all and ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.get_keys

syntax: keys = ngx.shared.DICT:get_keys(max_count?)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Fetch a list of the keys from the dictionary, up to <max_count>.

By default, only the first 1024 keys (if any) are returned. When the <max_count> argument is given the value 0, then all the keys will be returned even there is more than 1024 keys in the dictionary.

WARNING Be careful when calling this method on dictionaries with a really huge number of keys. This method may lock the dictionary for quite a while and block all the nginx worker processes that are trying to access the dictionary.

This feature was first introduced in the v0.7.3 release.

Back to TOC

ngx.socket.udp

syntax: udpsock = ngx.socket.udp()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Creates and returns a UDP or datagram-oriented unix domain socket object (also known as one type of the "cosocket" objects). The following methods are supported on this object:

It is intended to be compatible with the UDP API of the LuaSocket library but is 100% nonblocking out of the box.

This feature was first introduced in the v0.5.7 release.

See also ngx.socket.tcp.

Back to TOC

udpsock:setpeername

syntax: ok, err = udpsock:setpeername(host, port)

syntax: ok, err = udpsock:setpeername("unix:/path/to/unix-domain.socket")

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Attempts to connect a UDP socket object to a remote server or to a datagram unix domain socket file. Because the datagram protocol is actually connection-less, this method does not really establish a "connection", but only just set the name of the remote peer for subsequent read/write operations.

Both IP addresses and domain names can be specified as the host argument. In case of domain names, this method will use Nginx core's dynamic resolver to parse the domain name without blocking and it is required to configure the resolver directive in the nginx.conf file like this:


 resolver 8.8.8.8;  # use Google's public DNS nameserver

If the nameserver returns multiple IP addresses for the host name, this method will pick up one randomly.

In case of error, the method returns nil followed by a string describing the error. In case of success, the method returns 1.

Here is an example for connecting to a UDP (memcached) server:


 location /test {
     resolver 8.8.8.8;

     content_by_lua_block {
         local sock = ngx.socket.udp()
         local ok, err = sock:setpeername("my.memcached.server.domain", 11211)
         if not ok then
             ngx.say("failed to connect to memcached: ", err)
             return
         end
         ngx.say("successfully connected to memcached!")
         sock:close()
     }
 }

Since the v0.7.18 release, connecting to a datagram unix domain socket file is also possible on Linux:


 local sock = ngx.socket.udp()
 local ok, err = sock:setpeername("unix:/tmp/some-datagram-service.sock")
 if not ok then
     ngx.say("failed to connect to the datagram unix domain socket: ", err)
     return
 end

assuming the datagram service is listening on the unix domain socket file /tmp/some-datagram-service.sock and the client socket will use the "autobind" feature on Linux.

Calling this method on an already connected socket object will cause the original connection to be closed first.

This method was first introduced in the v0.5.7 release.

Back to TOC

udpsock:send

syntax: ok, err = udpsock:send(data)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Sends data on the current UDP or datagram unix domain socket object.

In case of success, it returns 1. Otherwise, it returns nil and a string describing the error.

The input argument data can either be a Lua string or a (nested) Lua table holding string fragments. In case of table arguments, this method will copy all the string elements piece by piece to the underlying Nginx socket send buffers, which is usually optimal than doing string concatenation operations on the Lua land.

This feature was first introduced in the v0.5.7 release.

Back to TOC

udpsock:receive

syntax: data, err = udpsock:receive(size?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Receives data from the UDP or datagram unix domain socket object with an optional receive buffer size argument, size.

This method is a synchronous operation and is 100% nonblocking.

In case of success, it returns the data received; in case of error, it returns nil with a string describing the error.

If the size argument is specified, then this method will use this size as the receive buffer size. But when this size is greater than 8192, then 8192 will be used instead.

If no argument is specified, then the maximal buffer size, 8192 is assumed.

Timeout for the reading operation is controlled by the lua_socket_read_timeout config directive and the settimeout method. And the latter takes priority. For example:


 sock:settimeout(1000)  -- one second timeout
 local data, err = sock:receive()
 if not data then
     ngx.say("failed to read a packet: ", err)
     return
 end
 ngx.say("successfully read a packet: ", data)

It is important here to call the settimeout method before calling this method.

This feature was first introduced in the v0.5.7 release.

Back to TOC

udpsock:close

syntax: ok, err = udpsock:close()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Closes the current UDP or datagram unix domain socket. It returns the 1 in case of success and returns nil with a string describing the error otherwise.

Socket objects that have not invoked this method (and associated connections) will be closed when the socket object is released by the Lua GC (Garbage Collector) or the current client HTTP request finishes processing.

This feature was first introduced in the v0.5.7 release.

Back to TOC

udpsock:settimeout

syntax: udpsock:settimeout(time)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Set the timeout value in milliseconds for subsequent socket operations (like receive).

Settings done by this method takes priority over those config directives, like lua_socket_read_timeout.

This feature was first introduced in the v0.5.7 release.

Back to TOC

ngx.socket.stream

Just an alias to ngx.socket.tcp. If the stream-typed cosocket may also connect to a unix domain socket, then this API name is preferred.

This API function was first added to the v0.10.1 release.

Back to TOC

ngx.socket.tcp

syntax: tcpsock = ngx.socket.tcp()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Creates and returns a TCP or stream-oriented unix domain socket object (also known as one type of the "cosocket" objects). The following methods are supported on this object:

It is intended to be compatible with the TCP API of the LuaSocket library but is 100% nonblocking out of the box. Also, we introduce some new APIs to provide more functionalities.

The cosocket object created by this API function has exactly the same lifetime as the Lua handler creating it. So never pass the cosocket object to any other Lua handler (including ngx.timer callback functions) and never share the cosocket object between different NGINX requests.

For every cosocket object's underlying connection, if you do not explicitly close it (via close) or put it back to the connection pool (via setkeepalive), then it is automatically closed when one of the following two events happens:

  • the current request handler completes, or
  • the Lua cosocket object value gets collected by the Lua GC.

Fatal errors in cosocket operations always automatically close the current connection (note that, read timeout error is the only error that is not fatal), and if you call close on a closed connection, you will get the "closed" error.

Starting from the 0.9.9 release, the cosocket object here is full-duplex, that is, a reader "light thread" and a writer "light thread" can operate on a single cosocket object simultaneously (both "light threads" must belong to the same Lua handler though, see reasons above). But you cannot have two "light threads" both reading (or writing or connecting) the same cosocket, otherwise you might get an error like "socket busy reading" when calling the methods of the cosocket object.

This feature was first introduced in the v0.5.0rc1 release.

See also ngx.socket.udp.

Back to TOC

tcpsock:connect

syntax: ok, err = tcpsock:connect(host, port, options_table?)

syntax: ok, err = tcpsock:connect("unix:/path/to/unix-domain.socket", options_table?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Attempts to connect a TCP socket object to a remote server or to a stream unix domain socket file without blocking.

Before actually resolving the host name and connecting to the remote backend, this method will always look up the connection pool for matched idle connections created by previous calls of this method (or the ngx.socket.connect function).

Both IP addresses and domain names can be specified as the host argument. In case of domain names, this method will use Nginx core's dynamic resolver to parse the domain name without blocking and it is required to configure the resolver directive in the nginx.conf file like this:


 resolver 8.8.8.8;  # use Google's public DNS nameserver

If the nameserver returns multiple IP addresses for the host name, this method will pick up one randomly.

In case of error, the method returns nil followed by a string describing the error. In case of success, the method returns 1.

Here is an example for connecting to a TCP server:


 location /test {
     resolver 8.8.8.8;

     content_by_lua_block {
         local sock = ngx.socket.tcp()
         local ok, err = sock:connect("www.google.com", 80)
         if not ok then
             ngx.say("failed to connect to google: ", err)
             return
         end
         ngx.say("successfully connected to google!")
         sock:close()
     }
 }

Connecting to a Unix Domain Socket file is also possible:


 local sock = ngx.socket.tcp()
 local ok, err = sock:connect("unix:/tmp/memcached.sock")
 if not ok then
     ngx.say("failed to connect to the memcached unix domain socket: ", err)
     return
 end

assuming memcached (or something else) is listening on the unix domain socket file /tmp/memcached.sock.

Timeout for the connecting operation is controlled by the lua_socket_connect_timeout config directive and the settimeout method. And the latter takes priority. For example:


 local sock = ngx.socket.tcp()
 sock:settimeout(1000)  -- one second timeout
 local ok, err = sock:connect(host, port)

It is important here to call the settimeout method before calling this method.

Calling this method on an already connected socket object will cause the original connection to be closed first.

An optional Lua table can be specified as the last argument to this method to specify various connect options:

  • pool specify a custom name for the connection pool being used. If omitted, then the connection pool name will be generated from the string template "<host>:<port>" or "<unix-socket-path>".

The support for the options table argument was first introduced in the v0.5.7 release.

This method was first introduced in the v0.5.0rc1 release.

Back to TOC

tcpsock:sslhandshake

syntax: session, err = tcpsock:sslhandshake(reused_session?, server_name?, ssl_verify?, send_status_req?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Does SSL/TLS handshake on the currently established connection.

The optional reused_session argument can take a former SSL session userdata returned by a previous sslhandshake call for exactly the same target. For short-lived connections, reusing SSL sessions can usually speed up the handshake by one order by magnitude but it is not so useful if the connection pool is enabled. This argument defaults to nil. If this argument takes the boolean false value, no SSL session userdata would return by this call and only a Lua boolean will be returned as the first return value; otherwise the current SSL session will always be returned as the first argument in case of successes.

The optional server_name argument is used to specify the server name for the new TLS extension Server Name Indication (SNI). Use of SNI can make different servers share the same IP address on the server side. Also, when SSL verification is enabled, this server_name argument is also used to validate the server name specified in the server certificate sent from the remote.

The optional ssl_verify argument takes a Lua boolean value to control whether to perform SSL verification. When set to true, the server certificate will be verified according to the CA certificates specified by the lua_ssl_trusted_certificate directive. You may also need to adjust the lua_ssl_verify_depth directive to control how deep we should follow along the certificate chain. Also, when the ssl_verify argument is true and the server_name argument is also specified, the latter will be used to validate the server name in the server certificate.

The optional send_status_req argument takes a boolean that controls whether to send the OCSP status request in the SSL handshake request (which is for requesting OCSP stapling).

For connections that have already done SSL/TLS handshake, this method returns immediately.

This method was first introduced in the v0.9.11 release.

Back to TOC

tcpsock:send

syntax: bytes, err = tcpsock:send(data)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Sends data without blocking on the current TCP or Unix Domain Socket connection.

This method is a synchronous operation that will not return until all the data has been flushed into the system socket send buffer or an error occurs.

In case of success, it returns the total number of bytes that have been sent. Otherwise, it returns nil and a string describing the error.

The input argument data can either be a Lua string or a (nested) Lua table holding string fragments. In case of table arguments, this method will copy all the string elements piece by piece to the underlying Nginx socket send buffers, which is usually optimal than doing string concatenation operations on the Lua land.

Timeout for the sending operation is controlled by the lua_socket_send_timeout config directive and the settimeout method. And the latter takes priority. For example:


 sock:settimeout(1000)  -- one second timeout
 local bytes, err = sock:send(request)

It is important here to call the settimeout method before calling this method.

In case of any connection errors, this method always automatically closes the current connection.

This feature was first introduced in the v0.5.0rc1 release.

Back to TOC

tcpsock:receive

syntax: data, err, partial = tcpsock:receive(size)

syntax: data, err, partial = tcpsock:receive(pattern?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Receives data from the connected socket according to the reading pattern or size.

This method is a synchronous operation just like the send method and is 100% nonblocking.

In case of success, it returns the data received; in case of error, it returns nil with a string describing the error and the partial data received so far.

If a number-like argument is specified (including strings that look like numbers), then it is interpreted as a size. This method will not return until it reads exactly this size of data or an error occurs.

If a non-number-like string argument is specified, then it is interpreted as a "pattern". The following patterns are supported:

  • '*a': reads from the socket until the connection is closed. No end-of-line translation is performed;
  • '*l': reads a line of text from the socket. The line is terminated by a Line Feed (LF) character (ASCII 10), optionally preceded by a Carriage Return (CR) character (ASCII 13). The CR and LF characters are not included in the returned line. In fact, all CR characters are ignored by the pattern.

If no argument is specified, then it is assumed to be the pattern '*l', that is, the line reading pattern.

Timeout for the reading operation is controlled by the lua_socket_read_timeout config directive and the settimeout method. And the latter takes priority. For example:


 sock:settimeout(1000)  -- one second timeout
 local line, err, partial = sock:receive()
 if not line then
     ngx.say("failed to read a line: ", err)
     return
 end
 ngx.say("successfully read a line: ", line)

It is important here to call the settimeout method before calling this method.

Since the v0.8.8 release, this method no longer automatically closes the current connection when the read timeout error happens. For other connection errors, this method always automatically closes the connection.

This feature was first introduced in the v0.5.0rc1 release.

Back to TOC

tcpsock:receiveuntil

syntax: iterator = tcpsock:receiveuntil(pattern, options?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

This method returns an iterator Lua function that can be called to read the data stream until it sees the specified pattern or an error occurs.

Here is an example for using this method to read a data stream with the boundary sequence --abcedhb:


 local reader = sock:receiveuntil("\r\n--abcedhb")
 local data, err, partial = reader()
 if not data then
     ngx.say("failed to read the data stream: ", err)
 end
 ngx.say("read the data stream: ", data)

When called without any argument, the iterator function returns the received data right before the specified pattern string in the incoming data stream. So for the example above, if the incoming data stream is 'hello, world! -agentzh\r\n--abcedhb blah blah', then the string 'hello, world! -agentzh' will be returned.

In case of error, the iterator function will return nil along with a string describing the error and the partial data bytes that have been read so far.

The iterator function can be called multiple times and can be mixed safely with other cosocket method calls or other iterator function calls.

The iterator function behaves differently (i.e., like a real iterator) when it is called with a size argument. That is, it will read that size of data on each invocation and will return nil at the last invocation (either sees the boundary pattern or meets an error). For the last successful invocation of the iterator function, the err return value will be nil too. The iterator function will be reset after the last successful invocation that returns nil data and nil error. Consider the following example:


 local reader = sock:receiveuntil("\r\n--abcedhb")

 while true do
     local data, err, partial = reader(4)
     if not data then
         if err then
             ngx.say("failed to read the data stream: ", err)
             break
         end

         ngx.say("read done")
         break
     end
     ngx.say("read chunk: [", data, "]")
 end

Then for the incoming data stream 'hello, world! -agentzh\r\n--abcedhb blah blah', we shall get the following output from the sample code above:

read chunk: [hell]
read chunk: [o, w]
read chunk: [orld]
read chunk: [! -a]
read chunk: [gent]
read chunk: [zh]
read done

Note that, the actual data returned might be a little longer than the size limit specified by the size argument when the boundary pattern has ambiguity for streaming parsing. Near the boundary of the data stream, the data string actually returned could also be shorter than the size limit.

Timeout for the iterator function's reading operation is controlled by the lua_socket_read_timeout config directive and the settimeout method. And the latter takes priority. For example:


 local readline = sock:receiveuntil("\r\n")

 sock:settimeout(1000)  -- one second timeout
 line, err, partial = readline()
 if not line then
     ngx.say("failed to read a line: ", err)
     return
 end
 ngx.say("successfully read a line: ", line)

It is important here to call the settimeout method before calling the iterator function (note that the receiveuntil call is irrelevant here).

As from the v0.5.1 release, this method also takes an optional options table argument to control the behavior. The following options are supported:

  • inclusive

The inclusive takes a boolean value to control whether to include the pattern string in the returned data string. Default to false. For example,


 local reader = tcpsock:receiveuntil("_END_", { inclusive = true })
 local data = reader()
 ngx.say(data)

Then for the input data stream "hello world _END_ blah blah blah", then the example above will output hello world _END_, including the pattern string _END_ itself.

Since the v0.8.8 release, this method no longer automatically closes the current connection when the read timeout error happens. For other connection errors, this method always automatically closes the connection.

This method was first introduced in the v0.5.0rc1 release.

Back to TOC

tcpsock:close

syntax: ok, err = tcpsock:close()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Closes the current TCP or stream unix domain socket. It returns the 1 in case of success and returns nil with a string describing the error otherwise.

Note that there is no need to call this method on socket objects that have invoked the setkeepalive method because the socket object is already closed (and the current connection is saved into the built-in connection pool).

Socket objects that have not invoked this method (and associated connections) will be closed when the socket object is released by the Lua GC (Garbage Collector) or the current client HTTP request finishes processing.

This feature was first introduced in the v0.5.0rc1 release.

Back to TOC

tcpsock:settimeout

syntax: tcpsock:settimeout(time)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Set the timeout value in milliseconds for subsequent socket operations (connect, receive, and iterators returned from receiveuntil).

Settings done by this method takes priority over those config directives, i.e., lua_socket_connect_timeout, lua_socket_send_timeout, and lua_socket_read_timeout.

Note that this method does not affect the lua_socket_keepalive_timeout setting; the timeout argument to the setkeepalive method should be used for this purpose instead.

This feature was first introduced in the v0.5.0rc1 release.

Back to TOC

tcpsock:settimeouts

syntax: tcpsock:settimeouts(connect_timeout, send_timeout, read_timeout)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Sets the connect timeout thresold, send timeout threshold, and read timeout threshold, respetively, in milliseconds, for subsequent socket operations (connect, send, receive, and iterators returned from receiveuntil).

Settings done by this method takes priority over those config directives, i.e., lua_socket_connect_timeout, lua_socket_send_timeout, and lua_socket_read_timeout.

You are recommended to use settimeouts instead of settimeout.

Note that this method does not affect the lua_socket_keepalive_timeout setting; the timeout argument to the setkeepalive method should be used for this purpose instead.

This feature was first introduced in the v0.10.7 release.

Back to TOC

tcpsock:setoption

syntax: tcpsock:setoption(option, value?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

This function is added for LuaSocket API compatibility and does nothing for now. Its functionality will be implemented in future.

This feature was first introduced in the v0.5.0rc1 release.

Back to TOC

tcpsock:setkeepalive

syntax: ok, err = tcpsock:setkeepalive(timeout?, size?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Puts the current socket's connection immediately into the cosocket built-in connection pool and keep it alive until other connect method calls request it or the associated maximal idle timeout is expired.

The first optional argument, timeout, can be used to specify the maximal idle timeout (in milliseconds) for the current connection. If omitted, the default setting in the lua_socket_keepalive_timeout config directive will be used. If the 0 value is given, then the timeout interval is unlimited.

The second optional argument, size, can be used to specify the maximal number of connections allowed in the connection pool for the current server (i.e., the current host-port pair or the unix domain socket file path). Note that the size of the connection pool cannot be changed once the pool is created. When this argument is omitted, the default setting in the lua_socket_pool_size config directive will be used.

When the connection pool exceeds the available size limit, the least recently used (idle) connection already in the pool will be closed to make room for the current connection.

Note that the cosocket connection pool is per Nginx worker process rather than per Nginx server instance, so the size limit specified here also applies to every single Nginx worker process.

Idle connections in the pool will be monitored for any exceptional events like connection abortion or unexpected incoming data on the line, in which cases the connection in question will be closed and removed from the pool.

In case of success, this method returns 1; otherwise, it returns nil and a string describing the error.

When the system receive buffer for the current connection has unread data, then this method will return the "connection in dubious state" error message (as the second return value) because the previous session has unread data left behind for the next session and the connection is not safe to be reused.

This method also makes the current cosocket object enter the "closed" state, so there is no need to manually call the close method on it afterwards.

This feature was first introduced in the v0.5.0rc1 release.

Back to TOC

tcpsock:getreusedtimes

syntax: count, err = tcpsock:getreusedtimes()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

This method returns the (successfully) reused times for the current connection. In case of error, it returns nil and a string describing the error.

If the current connection does not come from the built-in connection pool, then this method always returns 0, that is, the connection has never been reused (yet). If the connection comes from the connection pool, then the return value is always non-zero. So this method can also be used to determine if the current connection comes from the pool.

This feature was first introduced in the v0.5.0rc1 release.

Back to TOC

ngx.socket.connect

syntax: tcpsock, err = ngx.socket.connect(host, port)

syntax: tcpsock, err = ngx.socket.connect("unix:/path/to/unix-domain.socket")

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*

This function is a shortcut for combining ngx.socket.tcp() and the connect() method call in a single operation. It is actually implemented like this:


 local sock = ngx.socket.tcp()
 local ok, err = sock:connect(...)
 if not ok then
     return nil, err
 end
 return sock

There is no way to use the settimeout method to specify connecting timeout for this method and the lua_socket_connect_timeout directive must be set at configure time instead.

This feature was first introduced in the v0.5.0rc1 release.

Back to TOC

ngx.get_phase

syntax: str = ngx.get_phase()

context: init_by_lua*, init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Retrieves the current running phase name. Possible return values are

This API was first introduced in the v0.5.10 release.

Back to TOC

ngx.thread.spawn

syntax: co = ngx.thread.spawn(func, arg1, arg2, ...)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Spawns a new user "light thread" with the Lua function func as well as those optional arguments arg1, arg2, and etc. Returns a Lua thread (or Lua coroutine) object represents this "light thread".

"Light threads" are just a special kind of Lua coroutines that are scheduled by the ngx_lua module.

Before ngx.thread.spawn returns, the func will be called with those optional arguments until it returns, aborts with an error, or gets yielded due to I/O operations via the Nginx API for Lua (like tcpsock:receive).

After ngx.thread.spawn returns, the newly-created "light thread" will keep running asynchronously usually at various I/O events.

All the Lua code chunks running by rewrite_by_lua, access_by_lua, and content_by_lua are in a boilerplate "light thread" created automatically by ngx_lua. Such boilerplate "light thread" are also called "entry threads".

By default, the corresponding Nginx handler (e.g., rewrite_by_lua handler) will not terminate until

  1. both the "entry thread" and all the user "light threads" terminates,
  2. a "light thread" (either the "entry thread" or a user "light thread" aborts by calling ngx.exit, ngx.exec, ngx.redirect, or ngx.req.set_uri(uri, true), or
  3. the "entry thread" terminates with a Lua error.

When the user "light thread" terminates with a Lua error, however, it will not abort other running "light threads" like the "entry thread" does.

Due to the limitation in the Nginx subrequest model, it is not allowed to abort a running Nginx subrequest in general. So it is also prohibited to abort a running "light thread" that is pending on one ore more Nginx subrequests. You must call ngx.thread.wait to wait for those "light thread" to terminate before quitting the "world". A notable exception here is that you can abort pending subrequests by calling ngx.exit with and only with the status code ngx.ERROR (-1), 408, 444, or 499.

The "light threads" are not scheduled in a pre-emptive way. In other words, no time-slicing is performed automatically. A "light thread" will keep running exclusively on the CPU until

  1. a (nonblocking) I/O operation cannot be completed in a single run,
  2. it calls coroutine.yield to actively give up execution, or
  3. it is aborted by a Lua error or an invocation of ngx.exit, ngx.exec, ngx.redirect, or ngx.req.set_uri(uri, true).

For the first two cases, the "light thread" will usually be resumed later by the ngx_lua scheduler unless a "stop-the-world" event happens.

User "light threads" can create "light threads" themselves. And normal user coroutines created by coroutine.create can also create "light threads". The coroutine (be it a normal Lua coroutine or a "light thread") that directly spawns the "light thread" is called the "parent coroutine" for the "light thread" newly spawned.

The "parent coroutine" can call ngx.thread.wait to wait on the termination of its child "light thread".

You can call coroutine.status() and coroutine.yield() on the "light thread" coroutines.

The status of the "light thread" coroutine can be "zombie" if

  1. the current "light thread" already terminates (either successfully or with an error),
  2. its parent coroutine is still alive, and
  3. its parent coroutine is not waiting on it with ngx.thread.wait.

The following example demonstrates the use of coroutine.yield() in the "light thread" coroutines to do manual time-slicing:


 local yield = coroutine.yield

 function f()
     local self = coroutine.running()
     ngx.say("f 1")
     yield(self)
     ngx.say("f 2")
     yield(self)
     ngx.say("f 3")
 end

 local self = coroutine.running()
 ngx.say("0")
 yield(self)

 ngx.say("1")
 ngx.thread.spawn(f)

 ngx.say("2")
 yield(self)

 ngx.say("3")
 yield(self)

 ngx.say("4")

Then it will generate the output

0
1
f 1
2
f 2
3
f 3
4

"Light threads" are mostly useful for doing concurrent upstream requests in a single Nginx request handler, kinda like a generalized version of ngx.location.capture_multi that can work with all the Nginx API for Lua. The following example demonstrates parallel requests to MySQL, Memcached, and upstream HTTP services in a single Lua handler, and outputting the results in the order that they actually return (very much like the Facebook BigPipe model):


 -- query mysql, memcached, and a remote http service at the same time,
 -- output the results in the order that they
 -- actually return the results.

 local mysql = require "resty.mysql"
 local memcached = require "resty.memcached"

 local function query_mysql()
     local db = mysql:new()
     db:connect{
                 host = "127.0.0.1",
                 port = 3306,
                 database = "test",
                 user = "monty",
                 password = "mypass"
               }
     local res, err, errno, sqlstate =
             db:query("select * from cats order by id asc")
     db:set_keepalive(0, 100)
     ngx.say("mysql done: ", cjson.encode(res))
 end

 local function query_memcached()
     local memc = memcached:new()
     memc:connect("127.0.0.1", 11211)
     local res, err = memc:get("some_key")
     ngx.say("memcached done: ", res)
 end

 local function query_http()
     local res = ngx.location.capture("/my-http-proxy")
     ngx.say("http done: ", res.body)
 end

 ngx.thread.spawn(query_mysql)      -- create thread 1
 ngx.thread.spawn(query_memcached)  -- create thread 2
 ngx.thread.spawn(query_http)       -- create thread 3

This API was first enabled in the v0.7.0 release.

Back to TOC

ngx.thread.wait

syntax: ok, res1, res2, ... = ngx.thread.wait(thread1, thread2, ...)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Waits on one or more child "light threads" and returns the results of the first "light thread" that terminates (either successfully or with an error).

The arguments thread1, thread2, and etc are the Lua thread objects returned by earlier calls of ngx.thread.spawn.

The return values have exactly the same meaning as coroutine.resume, that is, the first value returned is a boolean value indicating whether the "light thread" terminates successfully or not, and subsequent values returned are the return values of the user Lua function that was used to spawn the "light thread" (in case of success) or the error object (in case of failure).

Only the direct "parent coroutine" can wait on its child "light thread", otherwise a Lua exception will be raised.

The following example demonstrates the use of ngx.thread.wait and ngx.location.capture to emulate ngx.location.capture_multi:


 local capture = ngx.location.capture
 local spawn = ngx.thread.spawn
 local wait = ngx.thread.wait
 local say = ngx.say

 local function fetch(uri)
     return capture(uri)
 end

 local threads = {
     spawn(fetch, "/foo"),
     spawn(fetch, "/bar"),
     spawn(fetch, "/baz")
 }

 for i = 1, #threads do
     local ok, res = wait(threads[i])
     if not ok then
         say(i, ": failed to run: ", res)
     else
         say(i, ": status: ", res.status)
         say(i, ": body: ", res.body)
     end
 end

Here it essentially implements the "wait all" model.

And below is an example demonstrating the "wait any" model:


 function f()
     ngx.sleep(0.2)
     ngx.say("f: hello")
     return "f done"
 end

 function g()
     ngx.sleep(0.1)
     ngx.say("g: hello")
     return "g done"
 end

 local tf, err = ngx.thread.spawn(f)
 if not tf then
     ngx.say("failed to spawn thread f: ", err)
     return
 end

 ngx.say("f thread created: ", coroutine.status(tf))

 local tg, err = ngx.thread.spawn(g)
 if not tg then
     ngx.say("failed to spawn thread g: ", err)
     return
 end

 ngx.say("g thread created: ", coroutine.status(tg))

 ok, res = ngx.thread.wait(tf, tg)
 if not ok then
     ngx.say("failed to wait: ", res)
     return
 end

 ngx.say("res: ", res)

 -- stop the "world", aborting other running threads
 ngx.exit(ngx.OK)

And it will generate the following output:

f thread created: running
g thread created: running
g: hello
res: g done

This API was first enabled in the v0.7.0 release.

Back to TOC

ngx.thread.kill

syntax: ok, err = ngx.thread.kill(thread)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*

Kills a running "light thread" created by ngx.thread.spawn. Returns a true value when successful or nil and a string describing the error otherwise.

According to the current implementation, only the parent coroutine (or "light thread") can kill a thread. Also, a running "light thread" with pending NGINX subrequests (initiated by ngx.location.capture for example) cannot be killed due to a limitation in the NGINX core.

This API was first enabled in the v0.9.9 release.

Back to TOC

ngx.on_abort

syntax: ok, err = ngx.on_abort(callback)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Registers a user Lua function as the callback which gets called automatically when the client closes the (downstream) connection prematurely.

Returns 1 if the callback is registered successfully or returns nil and a string describing the error otherwise.

All the Nginx API for Lua can be used in the callback function because the function is run in a special "light thread", just as those "light threads" created by ngx.thread.spawn.

The callback function can decide what to do with the client abortion event all by itself. For example, it can simply ignore the event by doing nothing and the current Lua request handler will continue executing without interruptions. And the callback function can also decide to terminate everything by calling ngx.exit, for example,


 local function my_cleanup()
     -- custom cleanup work goes here, like cancelling a pending DB transaction

     -- now abort all the "light threads" running in the current request handler
     ngx.exit(499)
 end

 local ok, err = ngx.on_abort(my_cleanup)
 if not ok then
     ngx.log(ngx.ERR, "failed to register the on_abort callback: ", err)
     ngx.exit(500)
 end

When lua_check_client_abort is set to off (which is the default), then this function call will always return the error message "lua_check_client_abort is off".

According to the current implementation, this function can only be called once in a single request handler; subsequent calls will return the error message "duplicate call".

This API was first introduced in the v0.7.4 release.

See also lua_check_client_abort.

Back to TOC

ngx.timer.at

syntax: ok, err = ngx.timer.at(delay, callback, user_arg1, user_arg2, ...)

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Creates an Nginx timer with a user callback function as well as optional user arguments.

The first argument, delay, specifies the delay for the timer, in seconds. One can specify fractional seconds like 0.001 to mean 1 millisecond here. 0 delay can also be specified, in which case the timer will immediately expire when the current handler yields execution.

The second argument, callback, can be any Lua function, which will be invoked later in a background "light thread" after the delay specified. The user callback will be called automatically by the Nginx core with the arguments premature, user_arg1, user_arg2, and etc, where the premature argument takes a boolean value indicating whether it is a premature timer expiration or not, and user_arg1, user_arg2, and etc, are those (extra) user arguments specified when calling ngx.timer.at as the remaining arguments.

Premature timer expiration happens when the Nginx worker process is trying to shut down, as in an Nginx configuration reload triggered by the HUP signal or in an Nginx server shutdown. When the Nginx worker is trying to shut down, one can no longer call ngx.timer.at to create new timers with nonzero delays and in that case ngx.timer.at will return nil and a string describing the error, that is, "process exiting".

Starting from the v0.9.3 release, it is allowed to create zero-delay timers even when the Nginx worker process starts shutting down.

When a timer expires, the user Lua code in the timer callback is running in a "light thread" detached completely from the original request creating the timer. So objects with the same lifetime as the request creating them, like cosockets, cannot be shared between the original request and the timer user callback function.

Here is a simple example:


 location / {
     ...
     log_by_lua_block {
         local function push_data(premature, uri, args, status)
             -- push the data uri, args, and status to the remote
             -- via ngx.socket.tcp or ngx.socket.udp
             -- (one may want to buffer the data in Lua a bit to
             -- save I/O operations)
         end
         local ok, err = ngx.timer.at(0, push_data,
                                      ngx.var.uri, ngx.var.args, ngx.header.status)
         if not ok then
             ngx.log(ngx.ERR, "failed to create timer: ", err)
             return
         end
     }
 }

One can also create infinite re-occurring timers, for instance, a timer getting triggered every 5 seconds, by calling ngx.timer.at recursively in the timer callback function. Here is such an example,


 local delay = 5
 local handler
 handler = function (premature)
     -- do some routine job in Lua just like a cron job
     if premature then
         return
     end
     local ok, err = ngx.timer.at(delay, handler)
     if not ok then
         ngx.log(ngx.ERR, "failed to create the timer: ", err)
         return
     end
 end

 local ok, err = ngx.timer.at(delay, handler)
 if not ok then
     ngx.log(ngx.ERR, "failed to create the timer: ", err)
     return
 end

Because timer callbacks run in the background and their running time will not add to any client request's response time, they can easily accumulate in the server and exhaust system resources due to either Lua programming mistakes or just too much client traffic. To prevent extreme consequences like crashing the Nginx server, there are built-in limitations on both the number of "pending timers" and the number of "running timers" in an Nginx worker process. The "pending timers" here mean timers that have not yet been expired and "running timers" are those whose user callbacks are currently running.

The maximal number of pending timers allowed in an Nginx worker is constrolled by the lua_max_pending_timers directive. The maximal number of running timers is controlled by the lua_max_running_timers directive.

According to the current implementation, each "running timer" will take one (fake) connection record from the global connection record list configured by the standard worker_connections directive in nginx.conf. So ensure that the worker_connections directive is set to a large enough value that takes into account both the real connections and fake connections required by timer callbacks (as limited by the lua_max_running_timers directive).

A lot of the Lua APIs for Nginx are enabled in the context of the timer callbacks, like stream/datagram cosockets (ngx.socket.tcp and ngx.socket.udp), shared memory dictionaries (ngx.shared.DICT), user coroutines (coroutine.*), user "light threads" (ngx.thread.*), ngx.exit, ngx.now/ngx.time, ngx.md5/ngx.sha1_bin, are all allowed. But the subrequest API (like ngx.location.capture), the ngx.req.* API, the downstream output API (like ngx.say, ngx.print, and ngx.flush) are explicitly disabled in this context.

You can pass most of the standard Lua values (nils, booleans, numbers, strings, tables, closures, file handles, and etc) into the timer callback, either explicitly as user arguments or implicitly as upvalues for the callback closure. There are several exceptions, however: you cannot pass any thread objects returned by coroutine.create and ngx.thread.spawn or any cosocket objects returned by ngx.socket.tcp, ngx.socket.udp, and ngx.req.socket because these objects' lifetime is bound to the request context creating them while the timer callback is detached from the creating request's context (by design) and runs in its own (fake) request context. If you try to share the thread or cosocket objects across the boundary of the creating request, then you will get the "no co ctx found" error (for threads) or "bad request" (for cosockets). It is fine, however, to create all these objects inside your timer callback.

This API was first introduced in the v0.8.0 release.

Back to TOC

ngx.timer.running_count

syntax: count = ngx.timer.running_count()

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns the number of timers currently running.

This directive was first introduced in the v0.9.20 release.

Back to TOC

ngx.timer.pending_count

syntax: count = ngx.timer.pending_count()

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns the number of pending timers.

This directive was first introduced in the v0.9.20 release.

Back to TOC

ngx.config.subsystem

syntax: subsystem = ngx.config.subsystem

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, init_by_lua*, init_worker_by_lua*

This string field indicates the current NGINX subsystem the current Lua environment is based on. For this module, this field always takes the string value "http". For ngx_stream_lua_module, however, this field takes the value "stream".

This field was first introduced in the 0.10.1.

Back to TOC

ngx.config.debug

syntax: debug = ngx.config.debug

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, init_by_lua*, init_worker_by_lua*

This boolean field indicates whether the current Nginx is a debug build, i.e., being built by the ./configure option --with-debug.

This field was first introduced in the 0.8.7.

Back to TOC

ngx.config.prefix

syntax: prefix = ngx.config.prefix()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, init_by_lua*, init_worker_by_lua*

Returns the Nginx server "prefix" path, as determined by the -p command-line option when running the nginx executable, or the path specified by the --prefix command-line option when building Nginx with the ./configure script.

This function was first introduced in the 0.9.2.

Back to TOC

ngx.config.nginx_version

syntax: ver = ngx.config.nginx_version

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, init_by_lua*, init_worker_by_lua*

This field take an integral value indicating the version number of the current Nginx core being used. For example, the version number 1.4.3 results in the Lua number 1004003.

This API was first introduced in the 0.9.3 release.

Back to TOC

ngx.config.nginx_configure

syntax: str = ngx.config.nginx_configure()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, init_by_lua*

This function returns a string for the NGINX ./configure command's arguments string.

This API was first introduced in the 0.9.5 release.

Back to TOC

ngx.config.ngx_lua_version

syntax: ver = ngx.config.ngx_lua_version

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, init_by_lua*

This field take an integral value indicating the version number of the current ngx_lua module being used. For example, the version number 0.9.3 results in the Lua number 9003.

This API was first introduced in the 0.9.3 release.

Back to TOC

ngx.worker.exiting

syntax: exiting = ngx.worker.exiting()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, init_by_lua*, init_worker_by_lua*

This function returns a boolean value indicating whether the current Nginx worker process already starts exiting. Nginx worker process exiting happens on Nginx server quit or configuration reload (aka HUP reload).

This API was first introduced in the 0.9.3 release.

Back to TOC

ngx.worker.pid

syntax: pid = ngx.worker.pid()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, init_by_lua*, init_worker_by_lua*

This function returns a Lua number for the process ID (PID) of the current Nginx worker process. This API is more efficient than ngx.var.pid and can be used in contexts where the ngx.var.VARIABLE API cannot be used (like init_worker_by_lua).

This API was first introduced in the 0.9.5 release.

Back to TOC

ngx.worker.count

syntax: count = ngx.worker.count()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, init_by_lua*, init_worker_by_lua*

Returns the total number of the Nginx worker processes (i.e., the value configured by the worker_processes directive in nginx.conf).

This API was first introduced in the 0.9.20 release.

Back to TOC

ngx.worker.id

syntax: count = ngx.worker.id()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, init_worker_by_lua*

Returns the ordinal number of the current Nginx worker processes (starting from number 0).

So if the total number of workers is N, then this method may return a number between 0 and N - 1 (inclusive).

This function returns meaningful values only for NGINX 1.9.1+. With earlier versions of NGINX, it always returns nil.

See also ngx.worker.count.

This API was first introduced in the 0.9.20 release.

Back to TOC

ngx.semaphore

syntax: local semaphore = require "ngx.semaphore"

This is a Lua module that implements a classic-style semaphore API for efficient synchronizations among different "light threads". Sharing the same semaphore among different "light threads" created in different (request) contexts are also supported as long as the "light threads" reside in the same NGINX worker process and the lua_code_cache directive is turned on (which is the default).

This Lua module does not ship with this ngx_lua module itself rather it is shipped with the lua-resty-core library.

Please refer to the documentation for this ngx.semaphore Lua module in lua-resty-core for more details.

This feature requires at least ngx_lua v0.10.0.

Back to TOC

ngx.balancer

syntax: local balancer = require "ngx.balancer"

This is a Lua module that provides a Lua API to allow defining completely dynamic load balancers in pure Lua.

This Lua module does not ship with this ngx_lua module itself rather it is shipped with the lua-resty-core library.

Please refer to the documentation for this ngx.balancer Lua module in lua-resty-core for more details.

This feature requires at least ngx_lua v0.10.0.

Back to TOC

ngx.ssl

syntax: local ssl = require "ngx.ssl"

This Lua module provides API functions to control the SSL handshake process in contexts like ssl_certificate_by_lua*.

This Lua module does not ship with this ngx_lua module itself rather it is shipped with the lua-resty-core library.

Please refer to the documentation for this ngx.ssl Lua module for more details.

This feature requires at least ngx_lua v0.10.0.

Back to TOC

ngx.ocsp

syntax: local ocsp = require "ngx.ocsp"

This Lua module provides API to perform OCSP queries, OCSP response validations, and OCSP stapling planting.

Usually, this module is used together with the ngx.ssl module in the context of ssl_certificate_by_lua*.

This Lua module does not ship with this ngx_lua module itself rather it is shipped with the lua-resty-core library.

Please refer to the documentation for this ngx.ocsp Lua module for more details.

This feature requires at least ngx_lua v0.10.0.

Back to TOC

ndk.set_var.DIRECTIVE

syntax: res = ndk.set_var.DIRECTIVE_NAME

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

This mechanism allows calling other nginx C modules' directives that are implemented by Nginx Devel Kit (NDK)'s set_var submodule's ndk_set_var_value.

For example, the following set-misc-nginx-module directives can be invoked this way:

For instance,


 local res = ndk.set_var.set_escape_uri('a/b');
 -- now res == 'a%2fb'

Similarly, the following directives provided by encrypted-session-nginx-module can be invoked from within Lua too:

This feature requires the ngx_devel_kit module.

Back to TOC

coroutine.create

syntax: co = coroutine.create(f)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, init_by_lua*, ngx.timer.*, header_filter_by_lua*, body_filter_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Creates a user Lua coroutines with a Lua function, and returns a coroutine object.

Similar to the standard Lua coroutine.create API, but works in the context of the Lua coroutines created by ngx_lua.

This API was first usable in the context of init_by_lua* since the 0.9.2.

This API was first introduced in the v0.6.0 release.

Back to TOC

coroutine.resume

syntax: ok, ... = coroutine.resume(co, ...)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, init_by_lua*, ngx.timer.*, header_filter_by_lua*, body_filter_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Resumes the executation of a user Lua coroutine object previously yielded or just created.

Similar to the standard Lua coroutine.resume API, but works in the context of the Lua coroutines created by ngx_lua.

This API was first usable in the context of init_by_lua* since the 0.9.2.

This API was first introduced in the v0.6.0 release.

Back to TOC

coroutine.yield

syntax: ... = coroutine.yield(...)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, init_by_lua*, ngx.timer.*, header_filter_by_lua*, body_filter_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Yields the execution of the current user Lua coroutine.

Similar to the standard Lua coroutine.yield API, but works in the context of the Lua coroutines created by ngx_lua.

This API was first usable in the context of init_by_lua* since the 0.9.2.

This API was first introduced in the v0.6.0 release.

Back to TOC

coroutine.wrap

syntax: co = coroutine.wrap(f)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, init_by_lua*, ngx.timer.*, header_filter_by_lua*, body_filter_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Similar to the standard Lua coroutine.wrap API, but works in the context of the Lua coroutines created by ngx_lua.

This API was first usable in the context of init_by_lua* since the 0.9.2.

This API was first introduced in the v0.6.0 release.

Back to TOC

coroutine.running

syntax: co = coroutine.running()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, init_by_lua*, ngx.timer.*, header_filter_by_lua*, body_filter_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Identical to the standard Lua coroutine.running API.

This API was first usable in the context of init_by_lua* since the 0.9.2.

This API was first enabled in the v0.6.0 release.

Back to TOC

coroutine.status

syntax: status = coroutine.status(co)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, init_by_lua*, ngx.timer.*, header_filter_by_lua*, body_filter_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Identical to the standard Lua coroutine.status API.

This API was first usable in the context of init_by_lua* since the 0.9.2.

This API was first enabled in the v0.6.0 release.