200 KiB

Исходник Ответственный История

Name

ngx_lua - Embed the power of Lua into Nginx

This module is not distributed with the Nginx source. See the installation instructions.

Status

This module is under active development and is production ready.

Version

This document describes ngx_lua v0.5.14 released on 13 August 2012.

Synopsis

# set search paths for pure Lua external libraries (';;' is the default path):
lua_package_path '/foo/bar/?.lua;/blah/?.lua;;';

# set search paths for Lua external libraries written in C (can also use ';;'):
lua_package_cpath '/bar/baz/?.so;/blah/blah/?.so;;';

server {
    location /inline_concat {
        # MIME type determined by default_type:
        default_type 'text/plain';

        set $a "hello";
        set $b "world";
        # inline Lua script
        set_by_lua $res "return ngx.arg[1]..ngx.arg[2]" $a $b;
        echo $res;
    }

    location /rel_file_concat {
        set $a "foo";
        set $b "bar";
        # script path relative to nginx prefix
        # $ngx_prefix/conf/concat.lua contents:
        #
        #    return ngx.arg[1]..ngx.arg[2]
        #
        set_by_lua_file $res conf/concat.lua $a $b;
        echo $res;
    }

    location /abs_file_concat {
        set $a "fee";
        set $b "baz";
        # absolute script path not modified
        set_by_lua_file $res /usr/nginx/conf/concat.lua $a $b;
        echo $res;
    }

    location /lua_content {
        # MIME type determined by default_type:
        default_type 'text/plain';

        content_by_lua "ngx.say('Hello,world!')";
    }

     location /nginx_var {
        # MIME type determined by default_type:
        default_type 'text/plain';

        # try access /nginx_var?a=hello,world
        content_by_lua "ngx.print(ngx.var['arg_a'], '\\n')";
    }

    location /request_body {
         # force reading request body (default off)
         lua_need_request_body on;
         client_max_body_size 50k;
         client_body_buffer_size 50k;

         content_by_lua 'ngx.print(ngx.var.request_body)';
    }

    # transparent non-blocking I/O in Lua via subrequests
    location /lua {
        # MIME type determined by default_type:
        default_type 'text/plain';

        content_by_lua '
            local res = ngx.location.capture("/some_other_location")
            if res.status == 200 then
                ngx.print(res.body)
            end';
    }

    # GET /recur?num=5
    location /recur {
        # MIME type determined by default_type:
        default_type 'text/plain';

        content_by_lua '
           local num = tonumber(ngx.var.arg_num) or 0

           if num > 50 then
               ngx.say("num too big")
               return
           end

           ngx.say("num is: ", num)

           if num > 0 then
               res = ngx.location.capture("/recur?num=" .. tostring(num - 1))
               ngx.print("status=", res.status, " ")
               ngx.print("body=", res.body)
           else
               ngx.say("end")
           end
           ';
    }

    location /foo {
        rewrite_by_lua '
            res = ngx.location.capture("/memc",
                { args = { cmd = 'incr', key = ngx.var.uri } }
            )
        ';

        proxy_pass http://blah.blah.com;
    }

    location /blah {
        access_by_lua '
            local res = ngx.location.capture("/auth")

            if res.status == ngx.HTTP_OK then
                return
            end

            if res.status == ngx.HTTP_FORBIDDEN then
                ngx.exit(res.status)
            end

            ngx.exit(ngx.HTTP_INTERNAL_SERVER_ERROR)
        ';

        # proxy_pass/fastcgi_pass/postgres_pass/...
    }

    location /mixed {
        rewrite_by_lua_file /path/to/rewrite.lua;
        access_by_lua_file /path/to/access.lua;
        content_by_lua_file /path/to/content.lua;
    }

    # use nginx var in code path
    # WARN: contents in nginx var must be carefully filtered,
    # otherwise there'll be great security risk!
    location ~ ^/app/(.+) {
            content_by_lua_file /path/to/lua/app/root/$1.lua;
    }

    location / {
       lua_need_request_body on;

       client_max_body_size 100k;
       client_body_buffer_size 100k;

       access_by_lua '
           -- check the client IP address is in our black list
           if ngx.var.remote_addr == "132.5.72.3" then
               ngx.exit(ngx.HTTP_FORBIDDEN)
           end

           -- check if the request body contains bad words
           if ngx.var.request_body and
                    string.match(ngx.var.request_body, "fsck")
           then
               return ngx.redirect("/terms_of_use.html")
           end

           -- tests passed
       ';

       # proxy_pass/fastcgi_pass/etc settings
    }
}

Description

This module embeds Lua, via the standard Lua interpreter or LuaJIT, into Nginx and by leveraging Nginx's subrequests, allows the integration of the powerful Lua threads (Lua coroutines) into the Nginx event model.

Unlike Apache's mod_lua and Lighttpd's mod_magnet, Lua code executed using this module can be 100% non-blocking on network traffic as long as the Nginx API for Lua provided by this module is used to handle requests to upstream services such as mysql, postgresql, memcached, redis, or upstream http web services. (See ngx.location.capture, ngx.location.capture_multi, ngx.socket.tcp, HttpDrizzleModule, ngx_postgres, HttpMemcModule, HttpRedis2Module and HttpProxyModule modules for details).

The Lua interpreter or LuaJIT instance is shared across all the requests in a single nginx worker process but request contexts are segregated using lightweight Lua coroutines. Loaded Lua modules persist in the nginx worker process level resulting in a small memory footprint even when under heavy loads.

Directives

lua_code_cache

syntax: lua_code_cache on | off

default: lua_code_cache on

context: main, server, location, location if

Enables or disables the Lua code cache for set_by_lua_file, content_by_lua_file, rewrite_by_lua_file, and access_by_lua_file, and also force Lua module reloading on a per-request basis.

The Lua files referenced in set_by_lua_file, content_by_lua_file, access_by_lua_file, and rewrite_by_lua_file will not be cached and the Lua package.loaded table will be cleared at the entry point of every request (such that Lua modules will not be cached either). With this in place, developers can adopt an edit-and-refresh approach.

Please note however, that Lua code written inline within nginx.conf such as those specified by set_by_lua, content_by_lua, access_by_lua, and rewrite_by_lua will always be cached because only the Nginx config file parser can correctly parse the nginx.conf file and the only ways to to reload the config file are to send a HUP signal or to restart Nginx.

The ngx_lua module does not currently support the stat mode available with the Apache mod_lua module but this is planned for implementation in the future.

Disabling the Lua code cache is strongly discouraged for production use and should only be used during development as it has a significant negative impact on overall performance. In addition, race conditions when reloading Lua modules are common for concurrent requests when the code cache is disabled.

lua_regex_cache_max_entries

syntax: lua_regex_cache_max_entries <num>

default: lua_regex_cache_max_entries 1024

context: http

Specifies the maximum number of entries allowed in the worker process level compiled regex cache.

The regular expressions used in ngx.re.match, ngx.re.gmatch, ngx.re.sub, and ngx.re.gsub will be cached within this cache if the regex option o (i.e., compile-once flag) is specified.

The default number of entries allowed is 1024 and when this limit is reached, new regular expressions will not be cached (as if the o option was not specified) and there will be one, and only one, warning in the error.log file:

2011/08/27 23:18:26 [warn] 31997#0: *1 lua exceeding regex cache max entries (1024), ...

Do not activate the o option for regular expressions (and/or replace string arguments for ngx.re.sub and ngx.re.gsub) that are generated on the fly and give rise to infinite variations to avoid hitting the specified limit.

lua_package_path

syntax: lua_package_path <lua-style-path-str>

default: The content of LUA_PATH environ variable or Lua's compiled-in defaults.

context: main

Sets the Lua module search path used by scripts specified by set_by_lua, content_by_lua and others. The path string is in standard Lua path form, and ;; can be used to stand for the original search paths.

As from the v0.5.0rc29 release, the special notation $prefix or ${prefix} can be used in the search path string to indicate the path of the server prefix usually determined by the -p PATH command-line option while starting the Nginx server.

lua_package_cpath

syntax: lua_package_cpath <lua-style-cpath-str>

default: The content of LUA_CPATH environment variable or Lua's compiled-in defaults.

context: main

Sets the Lua C-module search path used by scripts specified by set_by_lua, content_by_lua and others. The cpath string is in standard Lua cpath form, and ;; can be used to stand for the original cpath.

init_by_lua

syntax: init_by_lua <lua-script-str>

context: http

phase: loading-config

Runs the Lua code specified by the argument <lua-script-str> on the global Lua VM level when the Nginx master process (if any) is loading the Nginx config file.

When Nginx receives the HUP signal and starts reloading the config file, the Lua VM will also be re-created and init_by_lua will run again on the new Lua VM.

Usually you can register (true) Lua global variables or pre-load Lua modules at server start-up by means of this hook. Here is an example for pre-loading Lua modules:

init_by_lua 'require "cjson"';

server {
    location = /api {
        content_by_lua '
            ngx.say(cjson.encode({dog = 5, cat = 6}))
        ';
    }
}

You can also initialize the lua_shared_dict shm storage at this phase. Here is an example for this:

lua_shared_dict dogs 1m;

init_by_lua '
    local dogs = ngx.shared.dogs;
    dogs:set("Tom", 56)
';

server {
    location = /api {
        content_by_lua '
            local dogs = ngx.shared.dogs;
            ngx.say(dogs:get("Tom"))
        ';
    }
}

But note that, the lua_shared_dict's shm storage will not be cleared through a config reload (via the HUP signal, for example). So if you do not want to re-initialize the shm storage in your init_by_lua code in this case, then you just need to set a custom flag in the shm storage and always check the flag in your init_by_lua code.

Because the Lua code in this context runs before Nginx forks its worker processes (if any), data or code loaded here will enjoy the Copy-on-write (COW) feature provided by many operating systems among all the worker processes, thus saving a lot of memory.

Only a small set of the Nginx API for Lua is supported in this context:

Logging APIs: ngx.log and print,
Shared Dictionary API: ngx.shared.DICT.

More Nginx APIs for Lua may be supported in this context upon future user requests.

Basically you can safely use Lua libraries that do blocking I/O in this very context because blocking the master process during server start-up is completely okay. Even the Nginx core does blocking I/O (at least on resolving upstream's host names) at the configure-loading phase.

You should be very careful about potential security vulnerabilities in your Lua code registered in this context because the Nginx master process is often run under the root account.

This directive was first introduced in the v0.5.5 release.

init_by_lua_file

syntax: init_by_lua_file <path-to-lua-script-file>

context: http

phase: loading-config

Equivalent to init_by_lua, except that the file specified by <path-to-lua-script-file> contains the Lua code or Lua/LuaJIT bytecode to be executed.

When a relative path like foo/bar.lua is given, they will be turned into the absolute path relative to the server prefix path determined by the -p PATH command-line option while starting the Nginx server.

This directive was first introduced in the v0.5.5 release.

set_by_lua

syntax: set_by_lua $res <lua-script-str> [$arg1 $arg2 ...]

context: server, server if, location, location if

phase: server-rewrite, rewrite

Executes code specified in <lua-script-str> with optional input arguments $arg1 $arg2 ..., and returns string output to $res. The code in <lua-script-str> can make API calls and can retrieve input arguments from the ngx.arg table (index starts from 1 and increases sequentially).

This directive is designed to execute short, fast running code blocks as the Nginx event loop is blocked during code execution. Time consuming code sequences should therefore be avoided.

Note that the following API functions are currently disabled within this context:

Output API functions (e.g., ngx.say and ngx.send_headers)
Control API functions (e.g., ngx.exit)
Subrequest API functions (e.g., ngx.location.capture and ngx.location.capture_multi)
Cosocket API functions (e.g., ngx.socket.tcp and ngx.req.socket).

In addition, note that this directive can only write out a value to a single Nginx variable at a time. However, a workaround is possible using the ngx.var.VARIABLE interface.

location /foo {
    set $diff ''; # we have to predefine the $diff variable here

    set_by_lua $sum '
        local a = 32
        local b = 56

        ngx.var.diff = a - b;  -- write to $diff directly
        return a + b;          -- return the $sum value normally
    ';

    echo "sum = $sum, diff = $diff";
}

This directive can be freely mixed with all directives of the HttpRewriteModule, HttpSetMiscModule, and HttpArrayVarModule modules. All of these directives will run in the same order as they appear in the config file.

set $foo 32;
set_by_lua $bar 'tonumber(ngx.var.foo) + 1';
set $baz "bar: $bar";  # $baz == "bar: 33"

As from the v0.5.0rc29 release, Nginx variable interpolation is disabled in the <lua-script-str> argument of this directive and therefore, the dollar sign character ($) can be used directly.

This directive requires the ngx_devel_kit module.

set_by_lua_file

syntax: set_by_lua_file $res <path-to-lua-script-file> [$arg1 $arg2 ...]

context: server, server if, location, location if

phase: server-rewrite, rewrite

Equivalent to set_by_lua, except that the file specified by <path-to-lua-script-file> contains the Lua code, or, as from the v0.5.0rc32 release, the Lua/LuaJIT bytecode to be executed.

Nginx variable interpolation is supported in the <path-to-lua-script-file> argument string of this directive. But special care must be taken for injection attacks.

When the Lua code cache is turned on (by default), the user code is loaded once at the first request and cached and the Nginx config must be reloaded each time the Lua source file is modified. The Lua code cache can be temporarily disabled during development by switching lua_code_cache off in nginx.conf to avoid reloading Nginx.

This directive requires the ngx_devel_kit module.

content_by_lua

syntax: content_by_lua <lua-script-str>

context: location, location if

phase: content

Acts as a "content handler" and executes Lua code string specified in <lua-script-str> for every request. The Lua code may make API calls and is executed as a new spawned coroutine in an independent global environment (i.e. a sandbox).

Do not use this directive and other content handler directives in the same location. For example, this directive and the proxy_pass directive should not be used in the same location.

content_by_lua_file

syntax: content_by_lua_file <path-to-lua-script-file>

context: location, location if

phase: content

Equivalent to content_by_lua, except that the file specified by <path-to-lua-script-file> contains the Lua code, or, as from the v0.5.0rc32 release, the Lua/LuaJIT bytecode to be executed.

Nginx variables can be used in the <path-to-lua-script-file> string to provide flexibility. This however carries some risks and is not ordinarily recommended.

rewrite_by_lua

syntax: rewrite_by_lua <lua-script-str>

context: http, server, location, location if

phase: rewrite tail

Acts as a rewrite phase handler and executes Lua code string specified in <lua-script-str> for every request. The Lua code may make API calls and is executed as a new spawned coroutine in an independent global environment (i.e. a sandbox).

Note that this handler always runs after the standard HttpRewriteModule. So the following will work as expected:

   location /foo {
       set $a 12; # create and initialize $a
       set $b ""; # create and initialize $b
       rewrite_by_lua 'ngx.var.b = tonumber(ngx.var.a) + 1';
       echo "res = $b";
   }

because set $a 12 and set $b "" run before rewrite_by_lua.

On the other hand, the following will not work as expected:

?  location /foo {
?      set $a 12; # create and initialize $a
?      set $b ''; # create and initialize $b
?      rewrite_by_lua 'ngx.var.b = tonumber(ngx.var.a) + 1';
?      if ($b = '13') {
?         rewrite ^ /bar redirect;
?         break;
?      }
?
?      echo "res = $b";
?  }

because if runs before rewrite_by_lua even if it is placed after rewrite_by_lua in the config.

The right way of doing this is as follows:

location /foo {
    set $a 12; # create and initialize $a
    set $b ''; # create and initialize $b
    rewrite_by_lua '
        ngx.var.b = tonumber(ngx.var.a) + 1
        if tonumber(ngx.var.b) == 13 then
            return ngx.redirect("/bar");
        end
    ';

    echo "res = $b";
}

Note that the ngx_eval module can be approximated by using rewrite_by_lua. For example,

location / {
    eval $res {
        proxy_pass http://foo.com/check-spam;
    }

    if ($res = 'spam') {
        rewrite ^ /terms-of-use.html redirect;
    }

    fastcgi_pass ...;
}

can be implemented in ngx_lua as:

location = /check-spam {
    internal;
    proxy_pass http://foo.com/check-spam;
}

location / {
    rewrite_by_lua '
        local res = ngx.location.capture("/check-spam")
        if res.body == "spam" then
            ngx.redirect("/terms-of-use.html")
        end
    ';

    fastcgi_pass ...;
}

Just as any other rewrite phase handlers, rewrite_by_lua also runs in subrequests.

Note that when calling ngx.exit(ngx.OK) within a rewrite_by_lua handler, the nginx request processing control flow will still continue to the content handler. To terminate the current request from within a rewrite_by_lua handler, calling ngx.exit with status >= 200 (ngx.HTTP_OK) and status < 300 (ngx.HTTP_SPECIAL_RESPONSE) for successful quits and ngx.exit(ngx.HTTP_INTERNAL_SERVER_ERROR) (or its friends) for failures.

If the HttpRewriteModule's rewrite directive is used to change the URI and initiate location re-lookups (internal redirections), then any rewrite_by_lua or rewrite_by_lua_file code sequences within the current location will not be executed. For example,

location /foo {
    rewrite ^ /bar;
    rewrite_by_lua 'ngx.exit(503)';
}
location /bar {
    ...
}

Here the Lua code ngx.exit(503) will never run. This will be the case if rewrite ^ /bar last is used as this will similarly initiate an internal redirection. If the break modifier is used instead, there will be no internal redirection and the rewrite_by_lua code will be executed.

The rewrite_by_lua code will always run at the end of the rewrite request-processing phase unless rewrite_by_lua_no_postpone is turned on.

rewrite_by_lua_file

syntax: rewrite_by_lua_file <path-to-lua-script-file>

context: http, server, location, location if

phase: rewrite tail

Equivalent to rewrite_by_lua, except that the file specified by <path-to-lua-script-file> contains the Lua code, or, as from the v0.5.0rc32 release, the Lua/LuaJIT bytecode to be executed.

Nginx variables can be used in the <path-to-lua-script-file> string to provide flexibility. This however carries some risks and is not ordinarily recommended.

The rewrite_by_lua_file code will always run at the end of the rewrite request-processing phase unless rewrite_by_lua_no_postpone is turned on.

access_by_lua

syntax: access_by_lua <lua-script-str>

context: http, server, location, location if

phase: access tail

Acts as an access phase handler and executes Lua code string specified in <lua-script-str> for every request. The Lua code may make API calls and is executed as a new spawned coroutine in an independent global environment (i.e. a sandbox).

Note that this handler always runs after the standard HttpAccessModule. So the following will work as expected:

location / {
    deny    192.168.1.1;
    allow   192.168.1.0/24;
    allow   10.1.1.0/16;
    deny    all;

    access_by_lua '
        local res = ngx.location.capture("/mysql", { ... })
        ...
    ';

    # proxy_pass/fastcgi_pass/...
}

That is, if a client IP address is in the blacklist, it will be denied before the MySQL query for more complex authentication is executed by access_by_lua.

Note that the ngx_auth_request module can be approximated by using access_by_lua:

location / {
    auth_request /auth;

    # proxy_pass/fastcgi_pass/postgres_pass/...
}

can be implemented in ngx_lua as:

location / {
    access_by_lua '
        local res = ngx.location.capture("/auth")

        if res.status == ngx.HTTP_OK then
            return
        end

        if res.status == ngx.HTTP_FORBIDDEN then
            ngx.exit(res.status)
        end

        ngx.exit(ngx.HTTP_INTERNAL_SERVER_ERROR)
    ';

    # proxy_pass/fastcgi_pass/postgres_pass/...
}

As with other access phase handlers, access_by_lua will not run in subrequests.

Note that when calling ngx.exit(ngx.OK) within a access_by_lua handler, the nginx request processing control flow will still continue to the content handler. To terminate the current request from within a access_by_lua handler, calling ngx.exit with status >= 200 (ngx.HTTP_OK) and status < 300 (ngx.HTTP_SPECIAL_RESPONSE) for successful quits and ngx.exit(ngx.HTTP_INTERNAL_SERVER_ERROR) (or its friends) for failures.

access_by_lua_file

syntax: access_by_lua_file <path-to-lua-script-file>

context: http, server, location, location if

phase: access tail

Equivalent to access_by_lua, except that the file specified by <path-to-lua-script-file> contains the Lua code, or, as from the v0.5.0rc32 release, the Lua/LuaJIT bytecode to be executed.

Nginx variables can be used in the <path-to-lua-script-file> string to provide flexibility. This however carries some risks and is not ordinarily recommended.

header_filter_by_lua

syntax: header_filter_by_lua <lua-script-str>

context: http, server, location, location if

phase: output-header-filter

Uses Lua code specified in <lua-script-str> to define an output header filter.

Note that the following API functions are currently disabled within this context:

Output API functions (e.g., ngx.say and ngx.send_headers)
Control API functions (e.g., ngx.exit and ngx.exec)
Subrequest API functions (e.g., ngx.location.capture and ngx.location.capture_multi)
Cosocket API functions (e.g., ngx.socket.tcp and ngx.req.socket).

Here is an example of overriding a response header (or adding one if absent) in our Lua header filter:

location / {
    proxy_pass http://mybackend;
    header_filter_by_lua 'ngx.header.Foo = "blah"';
}

This directive was first introduced in the v0.2.1rc20 release.

header_filter_by_lua_file

syntax: header_filter_by_lua_file <path-to-lua-script-file>

context: http, server, location, location if

phase: output-header-filter

Equivalent to header_filter_by_lua, except that the file specified by <path-to-lua-script-file> contains the Lua code, or as from the v0.5.0rc32 release, the Lua/LuaJIT bytecode to be executed.

This directive was first introduced in the v0.2.1rc20 release.

body_filter_by_lua

syntax: body_filter_by_lua <lua-script-str>

context: http, server, location, location if

phase: output-body-filter

Uses Lua code specified in <lua-script-str> to define an output body filter.

The input data chunk is passed via ngx.arg[1] (as a Lua string value) and the "eof" flag indicating the end of the response body data stream is passed via ngx.arg[2] (as a Lua boolean value).

Behind the scene, the "eof" flag is just the last_buf flag of the nginx chain link buffers. And in the context of an Nginx subrequest, there is no "eof" flag at all, due to the underlying limitation in the Nginx core.

The output data stream can be aborted immediately by running the following Lua statement:

return ngx.ERROR

This will truncate the response body and usually result in incomplete and also invalid responses.

The Lua code can pass its own modified version of the input data chunk to the downstream Nginx output body filters by overriding ngx.arg[1] with a Lua string or a Lua table of strings. For example, to transform all the lowercase letters in the response body, we can just write:

location / {
    proxy_pass http://mybackend;
    body_filter_by_lua 'ngx.arg[1] = string.upper(ngx.arg[1])';
}

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.

Likewise, new "eof" flag can also be specified by setting a boolean value to ngx.arg[2]. For example,

location /t {
    echo hello world;
    echo hiya globe;

    body_filter_by_lua '
        local chunk = ngx.arg[1]
        if string.match(chunk, "hello") then
            ngx.arg[2] = true  -- new eof
            return
        end

        -- just throw away any remaining chunk data
        ngx.arg[1] = nil
    ';
}

Then GET /t will just return the output

hello world

That is, when the body filter sees a chunk containing the word "hello", then it will set the "eof" flag to true immediately, resulting in truncated but still valid responses.

When the Lua code may change the length of the response body, then it is required to always clear out the Content-Length response header (if any) in a header filter to enforce streaming output, as in

location /foo {
    # fastcgi_pass/proxy_pass/...

    header_filter_by_lua 'ngx.header.content_length = nil';
    body_filter_by_lua 'ngx.arg[1] = {string.len(arg[1]), "\n"}'
}

Note that the following API functions are currently disabled within this context:

Output API functions (e.g., ngx.say and ngx.send_headers)
Control API functions (e.g., ngx.exit and ngx.exec)
Subrequest API functions (e.g., ngx.location.capture and ngx.location.capture_multi)
Cosocket API functions (e.g., ngx.socket.tcp and ngx.req.socket).

Nginx output filters may be called multiple times for a single request because response body may be delivered in chunks. Thus, the Lua code specified by in this directive may also run multiple times in the lifetime of a single HTTP request.

This directive was first introduced in the v0.5.0rc32 release.

body_filter_by_lua_file

syntax: body_filter_by_lua_file <path-to-lua-script-file>

context: http, server, location, location if

phase: output-body-filter

Equivalent to body_filter_by_lua, except that the file specified by <path-to-lua-script-file> contains the Lua code, or, as from the v0.5.0rc32 release, the Lua/LuaJIT bytecode to be executed.

This directive was first introduced in the v0.5.0rc32 release.

log_by_lua

syntax: log_by_lua <lua-script-str>

context: http, server, location, location if

phase: log

Run the Lua source code inlined as the <lua-script-str> at the log request processing phase. This does not replace the current access logs, but runs after.

Note that the following API functions are currently disabled within this context:

Output API functions (e.g., ngx.say and ngx.send_headers)
Control API functions (e.g., ngx.exit)
Subrequest API functions (e.g., ngx.location.capture and ngx.location.capture_multi)
Cosocket API functions (e.g., ngx.socket.tcp and ngx.req.socket).

Here is an example of gathering average data for $upstream_response_time:

lua_shared_dict log_dict 5M;

server {
    location / {
        proxy_pass http://mybackend;

        log_by_lua '
            local log_dict = ngx.shared.log_dict
            local upstream_time = tonumber(ngx.var.upstream_response_time)

            local sum = log_dict:get("upstream_time-sum") or 0
            sum = sum + upstream_time
            log_dict:set("upstream_time-sum", sum)

            local newval, err = log_dict:incr("upstream_time-nb", 1)
            if not newval and err == "not found" then
                log_dict:add("upstream_time-nb", 0)
                log_dict:incr("upstream_time-nb", 1)
            end
        ';
    }

    location = /status {
        content_by_lua '
            local log_dict = ngx.shared.log_dict
            local sum = log_dict:get("upstream_time-sum")
            local nb = log_dict:get("upstream_time-nb")

            if nb and sum then
                ngx.say("average upstream response time: ", sum / nb,
                        " (", nb, " reqs)")
            else
                ngx.say("no data yet")
            end
        ';
    }
}

This directive was first introduced in the v0.5.0rc31 release.

log_by_lua_file

syntax: log_by_lua_file <path-to-lua-script-file>

context: http, server, location, location if

phase: log

Equivalent to log_by_lua, except that the file specified by <path-to-lua-script-file> contains the Lua code, or, as from the v0.5.0rc32 release, the Lua/LuaJIT bytecode to be executed.

This directive was first introduced in the v0.5.0rc31 release.

lua_need_request_body

syntax: lua_need_request_body <on|off>

default: off

context: main | server | location

phase: depends on usage

Determines whether to force the request body data to be read before running rewrite/access/access_by_lua* or not. The Nginx core does not read the client request body by default and if request body data is required, then this directive should be turned on or the ngx.req.read_body function should be called within the Lua code.

To read the request body data within the $request_body variable, client_body_buffer_size must have the same value as client_max_body_size. Because when the content length exceeds client_body_buffer_size but less than client_max_body_size, Nginx will automatically buffer the data into a temporary file on the disk, which will lead to empty value in the $request_body variable.

If the current location includes rewrite_by_lua or rewrite_by_lua_file directives, then the request body will be read just before the rewrite_by_lua or rewrite_by_lua_file code is run (and also at the rewrite phase). Similarly, if only content_by_lua is specified, the request body will not be read until the content handler's Lua code is about to run (i.e., the request body will be read during the content phase).

It is recommended however, to use the ngx.req.read_body and ngx.req.discard_body functions for finer control over the request body reading process instead.

This also applies to access_by_lua and access_by_lua_file.

lua_shared_dict

syntax: lua_shared_dict <name> <size>

default: no

context: http

phase: depends on usage

Declares a shared memory zone, <name>, to serve as storage for the shm based Lua dictionary ngx.shared.<name>.

The <size> argument accepts size units such as k and m:

http {
    lua_shared_dict dogs 10m;
    ...
}

See ngx.shared.DICT for details.

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

lua_socket_connect_timeout

syntax: lua_socket_connect_timeout <time>

default: lua_socket_connect_timeout 60s

context: http, server, location

This directive controls the default timeout value used in TCP/unix-domain socket object's connect method and can be overridden by the settimeout method.

The <time> argument can be an integer, with an optional time unit, like s (second), ms (millisecond), m (minute). The default time unit is s, i.e., "second". The default setting is 60s.

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

lua_socket_send_timeout

syntax: lua_socket_send_timeout <time>

default: lua_socket_send_timeout 60s

context: http, server, location

Controls the default timeout value used in TCP/unix-domain socket object's send method and can be overridden by the settimeout method.

The <time> argument can be an integer, with an optional time unit, like s (second), ms (millisecond), m (minute). The default time unit is s, i.e., "second". The default setting is 60s.

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

lua_socket_send_lowat

syntax: lua_socket_send_lowat <size>

default: lua_socket_send_lowat 0

context: http, server, location

Controls the lowat (low water) value for the cosocket send buffer.

lua_socket_read_timeout

syntax: lua_socket_read_timeout <time>

default: lua_socket_read_timeout 60s

context: http, server, location

phase: depends on usage

This directive controls the default timeout value used in TCP/unix-domain socket object's receive method and iterator functions returned by the receiveuntil method. This setting can be overridden by the settimeout method.

The <time> argument can be an integer, with an optional time unit, like s (second), ms (millisecond), m (minute). The default time unit is s, i.e., "second". The default setting is 60s.

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

lua_socket_buffer_size

syntax: lua_socket_buffer_size <size>

default: lua_socket_buffer_size 4k/8k

context: http, server, location

Specifies the buffer size used by cosocket reading operations.

This buffer does not have to be that big to hold everything at the same time because cosocket supports 100% non-buffered reading and parsing. So even 1 byte buffer size should still work everywhere but the performance could be terrible.

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

lua_socket_pool_size

syntax: lua_socket_pool_size <size>

default: lua_socket_pool_size 30

context: http, server, location

Specifies the size limit (in terms of connection count) for every cosocket connection pool associated with every remote server (i.e., identified by either the host-port pair or the unix domain socket file path).

Default to 30 connections for every pool.

When the connection pool is exceeding the size limit, the least recently used (idle) connection already in the pool will be closed automatically 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 so size limit specified here also applies to every single nginx worker process.

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

lua_socket_keepalive_timeout

syntax: lua_socket_keepalive_timeout <time>

default: lua_socket_keepalive_timeout 60s

context: http, server, location

This directive controls the default maximal idle time of the connections in the cosocket built-in connection pool. When this timeout reaches, idle connections will be closed automatically and removed from the pool. This setting can be overridden by cosocket objects' setkeepalive method.

The <time> argument can be an integer, with an optional time unit, like s (second), ms (millisecond), m (minute). The default time unit is s, i.e., "second". The default setting is 60s.

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

lua_socket_log_errors

syntax: lua_socket_log_errors on|off

default: lua_socket_log_errors on

context: http, server, location

This directive can be used to toggle the automatic error logging when a failure occurs for the TCP or UDP cosockets. If you are already doing proper error handling and logging in your Lua code, then it is recommended to turn this directive off to prevent data flushing in your nginx error log files (which is usually rather expensive).

This directive was first introduced in the v0.5.13 release.

lua_http10_buffering

syntax: lua_http10_buffering on|off

default: lua_http10_buffering on

context: http, server, location, location-if

Enables or disables the automatic response caching for HTTP 1.0 (or older) requests. This buffering mechanism is mainly used for HTTP 1.0 keep-alive which replies on a proper Content-Length response header.

If the Lua code explicitly sets a Content-Length response header before sending the headers (either explicitly via ngx.send_headers or implicitly via the first ngx.say or ngx.print call).

To output very large response data in a streaming fashion (via the ngx.flush call, for example), this directive MUST be turned off to minimize memory usage.

This directive is turned on by default.

This directive was first introduced in the v0.5.0rc19 release.

rewrite_by_lua_no_postpone

syntax: rewrite_by_lua_no_postpone on|off

default: rewrite_by_lua_no_postpone off

context: http, server, location, location-if

Controls whether or not to disable postponing rewrite_by_lua and rewrite_by_lua_file directives to run at the end of the rewrite request-processing phase. By default, this directive is turned off and the Lua code is postponed to run at the end of the rewrite phase.

This directive was first introduced in the v0.5.0rc29 release.

lua_transform_underscores_in_response_headers

syntax: lua_transform_underscores_in_response_headers on|off

default: lua_transform_underscores_in_response_headers on

context: http, server, location, location-if

Controls whether to transform underscores (_) in the response header names specified in the ngx.header.HEADER API to hypens (-).

This directive was first introduced in the v0.5.0rc32 release.

Nginx API for Lua

Introduction

The various *_by_lua 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 by using the package.seeall option:

module("my_module", package.seeall)

function say(a) ngx.say(a) end

Alternatively, they can be imported to external Lua modules by using file scoped local Lua variables:

local ngx = ngx
module("my_module")

function say(a) ngx.say(a) end

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.

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 or set_by_lua_file 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 $res
        '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 or body_filter_by_lua_file, 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.

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 '
        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, some are not, like $arg_PARAMETER.

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

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.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.

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.HTTP_GET
  ngx.HTTP_HEAD
  ngx.HTTP_PUT
  ngx.HTTP_POST
  ngx.HTTP_DELETE
  ngx.HTTP_OPTIONS   (first introduced in the v0.5.0rc24 release)

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

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**

  value = ngx.HTTP_OK (200)
  value = ngx.HTTP_CREATED (201)
  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_BAD_REQUEST (400)
  value = ngx.HTTP_UNAUTHORIZED (401)
  value = ngx.HTTP_FORBIDDEN (403)
  value = ngx.HTTP_NOT_FOUND (404)
  value = ngx.HTTP_NOT_ALLOWED (405)
  value = ngx.HTTP_GONE (410)
  value = ngx.HTTP_INTERNAL_SERVER_ERROR (500)
  value = ngx.HTTP_METHOD_NOT_IMPLEMENTED (501)
  value = ngx.HTTP_SERVICE_UNAVAILABLE (503)
  value = ngx.HTTP_GATEWAY_TIMEOUT (504) (first added in the v0.3.1rc38 release)

Nginx log level constants

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.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.

print

syntax: print(...)

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**

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.

ngx.ctx

context: set_by_lua, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua, log_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 '
        ngx.say("foo = ", ngx.ctx.foo)
        ngx.ctx.foo = 76
    ';
    access_by_lua '
        ngx.ctx.foo = ngx.ctx.foo + 3
    ';
    content_by_lua '
        ngx.say(ngx.ctx.foo)
    ';
}

Then GET /test will yield the output

foo = nil
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 '
        ngx.say("sub pre: ", ngx.ctx.blah)
        ngx.ctx.blah = 32
        ngx.say("sub post: ", ngx.ctx.blah)
    ';
}

location /main {
    content_by_lua '
        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 '
        ngx.say(ngx.ctx.foo)
    ';
}

location /orig {
    content_by_lua '
        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 }

ngx.location.capture

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

context: rewrite_by_lua, access_by_lua*, content_by_lua**

Issue 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).

Here is a basic example:

res = ngx.location.capture(uri)

Returns a Lua table with three slots (res.status, res.header, and res.body).

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"}.

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.

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 automatically escape argument keys and values according to URI rules and concatenating 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 '
        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 '
        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 '
        ngx.say("dog = ", ngx.var.dog)
        ngx.say("cat = ", ngx.var.cat)
    ';
}

location /lua {
    set $dog '';
    set $cat '';
    content_by_lua '
        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 '
        ngx.ctx.foo = "bar";
    ';
}
location /lua {
    content_by_lua '
        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 '
        ngx.ctx.foo = "bar";
    ';
}
location /lua {
    content_by_lua '
        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.

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.

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.

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

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 and header_filter_by_lua_file, for example,

location /test {
    set $footer '';

    proxy_pass http://some-backend;

    header_filter_by_lua '
        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.

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**

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.

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 request_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.

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 HttpRewriteModule'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 can only be set to true in rewrite_by_lua and rewrite_by_lua_file. 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 '
        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)

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

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

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 automatically 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.

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**

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

location = /test {
    content_by_lua '
        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 automatically 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.

ngx.req.get_post_args

syntax: 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 Lua exception errors.

location = /test {
    content_by_lua '
        ngx.req.read_body()
        local args = ngx.req.get_post_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

# 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 automatically 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. 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. 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.

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.

ngx.req.get_headers

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

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 args = ngx.req.get_headers(10)

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

local args = ngx.req.get_headers(0)

Removing the max_headers cap is strongly discouraged.

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. None of the current request's subrequests will be affected.

Here is an example of setting the Content-Length 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")

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

whether the current request body is already larger than the client_body_buffer_size,
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.

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. Please note that ignoring request body is not the right way to discard it, and that this function must be called to avoid breaking things under HTTP 1.1 keepalive or HTTP 1.1 pipelining.

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.

ngx.req.get_body_data

syntax: data = ngx.req.get_body_data()

context: rewrite_by_lua, access_by_lua*, content_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

the request body has not been read,
the request body has been read into disk temporary files,
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 automatically, but this is not recommended).

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.

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 automatically 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 automatically, but this is not recommended).

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.

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 requires patching the Nginx core to function properly because the Nginx core does not allow modifying request bodies by the current design. Here is a patch for Nginx 1.0.11: nginx-1.0.11-allow_request_body_updating.patch, and this patch should be applied cleanly to other releases of Nginx as well.

This patch has already been applied to ngx_openresty 1.0.8.17 and above.

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

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 automatically 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.

This function requires patching the Nginx core to function properly because the Nginx core does not allow modifying request bodies by the current design. Here is a patch for Nginx 1.0.9: nginx-1.0.9-allow_request_body_updating.patch, and this patch should be applied cleanly to other releases of Nginx as well. This patch has already been applied to ngx_openresty 1.0.8.17 and above.

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

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 automatically 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 HttpProxyModule and HttpFastcgiModule.

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

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.

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 HttpProxyModule and HttpFastcgiModule.

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

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 HttpProxyModule and HttpFastcgiModule.

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

ngx.req.socket

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

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 there is any request body data that has been pre-read into the Nginx core's request header buffer, the resulting cosocket object will take care of that automatically. So there will not be any data loss due to potential body data pre-reading.

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

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**

Clear the current request's request header named header_name. None of the current request's subrequests will be affected.

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.

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

Named locations are also supported, but query strings are ignored. For example,

location /foo {
    content_by_lua '
        ngx.exec("@bar");
    ';
}

location @bar {
    ...
}

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 automatically.

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.

Note that this is very different from ngx.redirect in that it is just an internal redirect and no new HTTP traffic is involved.

This method never returns.

This method must be called before ngx.send_headers or explicit response body outputs by either ngx.print or ngx.say.

It is strongly recommended to combine the return statement with this call, i.e., return ngx.exec(...).

This method is similar to the echo_exec directive of the HttpEchoModule.

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 whether 301 or 302 to be used. 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("http://localhost:1984/foo", ngx.HTTP_MOVED_TEMPORARILY)

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

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

This method must be called before ngx.send_headers or explicit response body outputs by either ngx.print or ngx.say.

This method never returns.

This method is very much like the rewrite directive with the redirect modifier in the standard HttpRewriteModule, 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')

It is strongly recommended to combine the return statement with this call, i.e., return ngx.redirect(...).

ngx.send_headers

syntax: ngx.send_headers()

context: rewrite_by_lua, access_by_lua*, content_by_lua**

Explicitly send out the response headers.

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.

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.

ngx.print

syntax: 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.

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.

ngx.say

syntax: ngx.say(...)

context: rewrite_by_lua, access_by_lua*, content_by_lua**

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

ngx.log

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

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**

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.

ngx.flush

syntax: 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 non functional when in the HTTP 1.0 output buffering mode. See HTTP 1.0 support.

ngx.exit

syntax: ngx.exit(status)

context: rewrite_by_lua, access_by_lua*, content_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.

It is strongly recommended to combine the return statement with this call, i.e., return ngx.exit(...).

ngx.eof

syntax: ngx.eof()

context: rewrite_by_lua, access_by_lua*, content_by_lua**

Explicitly specify the end of the response output stream.

ngx.sleep

syntax: ngx.sleep(seconds)

context: rewrite_by_lua, access_by_lua*, content_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.

This method was introduced in the 0.5.0rc30 release.

ngx.escape_uri

syntax: newstr = ngx.escape_uri(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**

Escape str as a URI component.

ngx.unescape_uri

syntax: newstr = ngx.unescape_uri(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**

Unescape str as an escaped URI component.

For example,

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

gives the output

b r56 7

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**

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.

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**

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.

ngx.encode_base64

syntax: newstr = ngx.encode_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**

Encode str to a base64 digest.

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**

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

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**

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.

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**

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.

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**

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.

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**

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

For example,

location = /md5 {
    content_by_lua 'ngx.say(ngx.md5("hello"))';
}

yields the output

5d41402abc4b2a76b9719d911017c592

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

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**

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.

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**

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.

ngx.today

syntax: str = ngx.today()

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 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.

ngx.time

syntax: secs = ngx.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 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.

ngx.now

syntax: secs = ngx.now()

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 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).

Use the Nginx core timer_resolution directive to adjust the accuracy or forcibly update the Nginx time cache by calling ngx.update_time first.

This API was first introduced in v0.3.1rc32.

ngx.update_time

syntax: ngx.update_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**

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.

ngx.localtime

syntax: str = ngx.localtime()

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 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.

ngx.utctime

syntax: str = ngx.utctime()

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 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.

ngx.cookie_time

syntax: str = ngx.cookie_time(sec)

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 formated 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"

ngx.http_time

syntax: str = ngx.http_time(sec)

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 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 10 11:27:35 GMT"

ngx.parse_http_time

syntax: sec = ngx.parse_http_time(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**

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 10 11:27:35 GMT")
if time == nil then
    ...
end

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.

ngx.re.match

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

context: set_by_lua, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_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 fatal errors, like seeing bad UTF-8 sequences in UTF-8 mode, a Lua exception will be raised.

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



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

Unmatched sub-patterns will have nil values in their captures table fields.

local m = ngx.re.match("hello, world", "(world)|(hello)")
-- m[0] == "hello"
-- m[1] == nil
-- m[2] == "hello"

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.

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.

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.

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

These options can be combined:

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



local m = 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. 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 = ngx.re.match("1234, hello", "[0-9]+", "", ctx)
     -- m[0] = "1234"
     -- ctx.pos == 4



local ctx = { pos = 2 }
local m = ngx.re.match("1234, hello", "[0-9]+", "", ctx)
     -- m[0] = "34"
     -- ctx.pos == 4

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 PCRE Sequences).

To confirm that PCRE JIT is enabled, activate the Nginx debug log by adding the --with-debug option to Nginx or ngx_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

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

ngx.re.gmatch

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

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

Here is a small example to demonstrate its basic usage:

local iterator = ngx.re.gmatch("hello, world!", "([a-z]+)", "i")
local m
m = iterator()    -- m[0] == m[1] == "hello"
m = iterator()    -- m[0] == m[1] == "world"
m = iterator()    -- m == nil

More often we just put it into a Lua for loop:

for m in ngx.re.gmatch("hello, world!", "([a-z]+)", "i")
    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 PCRE Sequences).

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

ngx.re.sub

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

context: set_by_lua, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_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, or throw out a Lua exception when an error occurred (syntax errors in the <replace> string argument, for example).

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

local newstr, n = ngx.re.sub("hello, 1234", "([0-9])[0-9]", "[$0][$1]")
    -- newstr == "hello, [12][1]34"
    -- n == 1

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 = ngx.re.sub("hello, 1234", "[0-9]", "${0}00")
    -- newstr == "hello, 10034"
    -- n == 1

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

local newstr, n = 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 = 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 PCRE Sequences).

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

ngx.re.gsub

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

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

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

Here is some examples:

local newstr, n = ngx.re.gsub("hello, world", "([a-z])[a-z]+", "[$0,$1]", "i")
    -- newstr == "[hello,h], [world,w]"
    -- n == 2



local func = function (m)
    return "[" .. m[0] .. "," .. m[1] .. "]"
end
local newstr, n = 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 PCRE Sequences).

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

ngx.shared.DICT

syntax: dict = ngx.shared.DICT

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**

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

The resulting object dict has the following methods:

Here is an example:

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

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**

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

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**

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 be expired.

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.

See also ngx.shared.DICT.

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**

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.

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**

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.

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**

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.

ngx.shared.DICT.incr

syntax: newval, err = ngx.shared.DICT:incr(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**

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.

The key must already exist in the dictionary, otherwise it will return nil and "not found".

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

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

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

See also ngx.shared.DICT.

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**

Flushes out all the items in the dictionary.

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

See also ngx.shared.DICT.

ngx.socket.udp

syntax: udpsock = ngx.socket.udp()

context: rewrite_by_lua, access_by_lua*, content_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.

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**

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 '
        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()
    ';
}

Connecting to a datagram unix domain socket file is also possible:

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.

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.

udpsock:send

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

context: rewrite_by_lua, access_by_lua*, content_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 automatically 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.

udpsock:receive

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

context: rewrite_by_lua, access_by_lua*, content_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: ", data)
    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.

udpsock:close

syntax: ok, err = udpsock:close()

context: rewrite_by_lua, access_by_lua*, content_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.

For socket objects that have not invoked this method, they (and their connections) will be automatically 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.

udpsock:settimeout

syntax: udpsock:settimeout(time)

context: rewrite_by_lua, access_by_lua*, content_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.

ngx.socket.tcp

syntax: tcpsock = ngx.socket.tcp()

context: rewrite_by_lua, access_by_lua*, content_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.

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

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**

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).

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 '
        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 automatically 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.

tcpsock:send

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

context: rewrite_by_lua, access_by_lua*, content_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.

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.

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

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**

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.

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

tcpsock:receiveuntil

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

context: rewrite_by_lua, access_by_lua*, content_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 automatically reset after its last successful invocation that returns nil data and nil error. Consider the following example:

local reader = sock:receiveuntil("\r\n--abcedhb")

while true then
    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.

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

tcpsock:close

syntax: ok, err = tcpsock:close()

context: rewrite_by_lua, access_by_lua*, content_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.

For socket objects that have invoked the setkeepalive method, there is no need to call this method on it because the socket object is already closed (and the current connection is saved into the built-in connection pool).

For socket objects that have not invoked setkeepalive nor this method, they (and their connections) will be automatically 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.

tcpsock:settimeout

syntax: tcpsock:settimeout(time)

context: rewrite_by_lua, access_by_lua*, content_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.

tcpsock:setoption

syntax: tcpsock:setoption(option, value?)

context: rewrite_by_lua, access_by_lua*, content_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.

tcpsock:setkeepalive

syntax: ok, err = tcpsock:setkeepalive(timeout?, size?)

context: rewrite_by_lua, access_by_lua*, content_by_lua**

Puts the current socket's connection 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 is exceeding the size limit, the least recently used (idle) connection already in the pool will be closed automatically 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.

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.

tcpsock:getreusedtimes

syntax: count, err = tcpsock:getreusedtimes()

context: rewrite_by_lua, access_by_lua*, content_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.

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**

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.

ngx.get_phase

syntax: str = ngx.get_phase()

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**

Retrieves the current running phase name. Possible return values are

init for the context of init_by_lua or init_by_lua_file.
set for the context of set_by_lua or set_by_lua_file.
rewrite for the context of rewrite_by_lua or rewrite_by_lua_file.
access for the context of access_by_lua or access_by_lua_file.
content for the context of content_by_lua or content_by_lua_file.
header_filter for the context of header_filter_by_lua or header_filter_by_lua_file.
body_filter for the context of body_filter_by_lua or body_filter_by_lua_file.
log for the context of log_by_lua or log_by_lua_file.

This API was first introduced in the v0.5.10 release.

coroutine.create

syntax: co = coroutine.create(f)

context: rewrite_by_lua, access_by_lua*, content_by_lua**

Creates a user Lua coroutines with a Lua function, and returns a coroutine object.

Just behaves like the standard Lua coroutine.create API, but works in the context of the Lua coroutines created automatically by this Nginx module.

This API was first introduced in the v0.6.0 release.

coroutine.resume

syntax: ok, ... = coroutine.resume(co, ...)

context: rewrite_by_lua, access_by_lua*, content_by_lua**

Resumes the executation of a user Lua coroutine object previously yielded or just created.

Just behaves like the standard Lua coroutine.resume API, but works in the context of the Lua coroutines created automatically by this Nginx module.

This API was first introduced in the v0.6.0 release.

coroutine.yield

syntax: ... = coroutine.yield(co, ...)

context: rewrite_by_lua, access_by_lua*, content_by_lua**

Yields the executation of the current user Lua coroutine.

Just behaves like the standard Lua coroutine.yield API, but works in the context of the Lua coroutines created automatically by this Nginx module.

This API was first introduced in the v0.6.0 release.

coroutine.wrap

syntax: co = coroutine.wrap(f)

context: rewrite_by_lua, access_by_lua*, content_by_lua**

Just behaves like the standard Lua coroutine.wrap API, but works in the context of the Lua coroutines created automatically by this Nginx module.

This API was first introduced in the v0.6.0 release.

coroutine.running

syntax: co = coroutine.running()

context: rewrite_by_lua, access_by_lua*, content_by_lua**

Identical to the standard Lua coroutine.running API.

This API was first enabled in the v0.6.0 release.

coroutine.status

syntax: status = coroutine.status(co)

context: rewrite_by_lua, access_by_lua*, content_by_lua**

Identical to the standard Lua coroutine.status API.

This API was first enabled in the v0.6.0 release.

ndk.set_var.DIRECTIVE

syntax: res = ndk.set_var.DIRECTIVE_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**

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 HttpSetMiscModule 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 HttpEncryptedSessionModule can be invoked from within Lua too:

This feature requires the ngx_devel_kit module.

Lua/LuaJIT bytecode support

As from the v0.5.0rc32 release, all *_by_lua_file configure directives (such as content_by_lua_file) support loading Lua 5.1 and LuaJIT 2.0 raw bytecode files directly.

Please note that the bytecode format used by LuaJIT 2.0 is not compatible with that for the standard Lua 5.1 interpreter. So if using LuaJIT 2.0 with ngx_lua, LuaJIT-compatible bytecode files must be generated as shown:

/path/to/luajit/bin/luajit -b /path/to/input_file.lua /path/to/output_file.luac

The -bg option can be used to include debug information in the LuaJIT bytecode file:

/path/to/luajit/bin/luajit -bg /path/to/input_file.lua /path/to/output_file.luac

Please refer to the official LuaJIT documentation for the -b option for more details:

http://luajit.org/running.html#opt_b

Similarly, if using the standard Lua 5.1 interpreter with ngx_lua, Lua-compatible bytecode files must be generated using the luac command-line utility as shown:

luac -o /path/to/output_file.luac /path/to/input_file.lua

Unlike as with LuaJIT, debug information is included in standard Lua 5.1 bytecode files by default. This can be striped out by specifying the -s option as shown:

luac -s -o /path/to/output_file.luac /path/to/input_file.lua

Attempts to load standard Lua 5.1 bytecode files into ngx_lua instances linked to LuaJIT 2.0 or vice versa, an error message such as that below will be logged in the Nginx error.log file:

[error] 13909#0: *1 failed to load Lua inlined code: bad byte-code header in /path/to/test_file.luac

Loading bytecode files via the Lua primitives like require and dofile should always work as expected.

HTTP 1.0 support

The HTTP 1.0 protocol does not support chunked outputs and always requires an explicit Content-Length header when the response body is non-empty in order to support the HTTP 1.0 keep-alive (as required by the ApacheBench (ab) tool). So when an HTTP 1.0 request is present and the lua_http10_buffering directive is turned on, this module will automatically buffer all the outputs of user calls of ngx.say and ngx.print and postpone sending response headers until it sees all the outputs in the response body, and at that time ngx_lua can calculate the total length of the body and construct a proper Content-Length header for the HTTP 1.0 client.

If the user Lua code sets the Content-Length response header itself, then the automatic buffering will be disabled even if the lua_http10_buffering directive is turned on.

For large streaming output responses, it is important to disable the lua_http10_buffering directive to minimise memory usage.

Note that, common HTTP benchmark tools like ab and http_load always issue HTTP 1.0 requests by default. To force curl to send HTTP 1.0 requests, use the -0 option.

NOTE: This mechanism behaves differently when code cache is turned off and should be considered a DIRTY TRICK. Note that backward compatibility is NOT guaranteed and that there may be other undesirable consequences. A new data sharing mechanism will be designed later.

To globally share data among all the requests handled by the same nginx worker process, encapsulate the shared data into a Lua module, use the Lua require builtin to import the module, and then manipulate the shared data in Lua. This works because required Lua modules are loaded only once and all coroutines will share the same copy of the module. Note however that Lua global variables WILL NOT persist between requests because of the one-coroutine-per-request isolation design.

Here is a complete small example:

-- mydata.lua
module("mydata", package.seeall)

local data = {
    dog = 3,
    cat = 4,
    pig = 5,
}

function get_age(name)
    return data[name]
end

and then accessing it from nginx.conf:

location /lua {
    content_lua_by_lua '
        local mydata = require("mydata")
        ngx.say(mydata.get_age("dog"))
    ';
}

The mydata module in this example will only be loaded and run on the first request to the location /lua, and all subsequent requests to the same nginx worker process will use the reloaded instance of the module as well as the same copy of the data in it, until a HUP signal is sent to the Nginx master process to force a reload. This data sharing technique is essential for high performance Lua applications based on this module.

Note that this data sharing is on a per-worker basis and not on a ''per-server' basis'. That is, when there are multiple nginx worker processes under an Nginx master, data sharing cannot cross the process boundary between these workers.

If server wide data sharing is required:

Use the ngx.shared.DICT API provided by this module.
Use only a single nginx worker and a single server. This is however not recommended when there is a multi core CPU or multiple CPUs in a single machine.
Use data storage mechanisms such as memcached, redis, MySQL or PostgreSQL. The ngx_openresty bundle associated with this module comes with a set of companion Nginx modules that provide interfaces with these data storage mechanisms. See the HttpMemcModule, HttpRedis2Module, HttpDrizzleModule and HttpPostgresModule modules for details

Known Issues

TCP socket connect operation issues

The tcpsock:connect method may indicate success despite connection failures such as with Connection Refused errors.

However, later attempts to manipulate the cosocket object will fail and return the actual error status message generated by the failed connect operation.

This issue is due to limitations in the Nginx event model and only appears to affect Mac OS X.

Lua Coroutine Yielding/Resuming

As the module's predefined Nginx I/O API uses the coroutine yielding/resuming mechanism, user code should not call any Lua modules that use the Lua coroutine mechanism in order to prevent conflicts with the module's predefined Nginx API methods such as ngx.location.capture (Actually, coroutine modules have been masked off in content_by_lua directives and others). This limitation is significant and work is ongoing on an alternative coroutine implementation that can fit into the Nginx event model to address this. When this is done, it will be possible to use the Lua coroutine mechanism freely as it is in standard Lua implementations.
Lua's dofile builtin is implemented as a C function in both Lua 5.1 and LuaJIT 2.0 and when ngx.location.capture is called, ngx.exec, ngx.exit or ngx.req.read_body or similar in the file to be loaded by dofile, a coroutine yield across the C function boundary will be initiated. This however is not normally allowed within ngx_lua and will usually result in error messages like lua handler aborted: runtime error: attempt to yield across C-call boundary. To avoid this, define a real Lua module and use the Lua require builtin instead.
As the standard Lua 5.1 interpreter's VM is not fully resumable, the methods ngx.location.capture, ngx.location.capture_multi, ngx.redirect, ngx.exec, and ngx.exit cannot be used within the context of a Lua pcall() or xpcall() when the standard Lua 5.1 interpreter is used and the attempt to yield across metamethod/C-call boundary error will be produced. Please use LuaJIT 2.0, which supports a fully resumable VM, to avoid this.

Lua Variable Scope

Care should be taken when importing modules and this form should be used:

    local xxx = require('xxx')


instead of the old deprecated form:


    require('xxx')


If the old form is required, force reload the module for every request by using the `package.loaded.<module>` command:


    package.loaded.xxx = nil
    require('xxx')

It is recommended to always place the following piece of code at the end of Lua modules that use the I/O operations to prevent casual use of module-level global variables that are shared among all requests:

getmetatable(foo.bar).__newindex = function (table, key, val)
    error('Attempt to write to undeclared variable "' .. key .. '": '
            .. debug.traceback())
end

Assuming the current Lua module is named foo.bar, this will guarantee that local variables in module foo.bar functions have been declared as "local". It prevents undesirable race conditions while accessing such variables. See Data Sharing within an Nginx Worker for the reasons behind this.

Locations Configured by Subrequest Directives of Other Modules

The ngx.location.capture and ngx.location.capture_multi directives cannot capture locations that include the echo_location, echo_location_async, echo_subrequest, or echo_subrequest_async directives.

location /foo {
    content_by_lua '
        res = ngx.location.capture("/bar")
    ';
}
location /bar {
    echo_location /blah;
}
location /blah {
    echo "Success!";
}



$ curl -i http://example.com/foo

will not work as expected.

Special PCRE Sequences

PCRE sequences such as \d, \s, or \w, require special attention because in string literals, the backslash character, \, is stripped out by both the Lua language parser and by the Nginx config file parser before processing. So the following snippet will not work as expected:

# nginx.conf
? location /test {
?     content_by_lua '
?         local regex = "\d+"  -- THIS IS WRONG!!
?         local m = ngx.re.match("hello, 1234", regex)
?         if m then ngx.say(m[0]) else ngx.say("not matched!") end
?     ';
? }
# evaluates to "not matched!"

To avoid this, double escape the backslash:

# nginx.conf
location /test {
    content_by_lua '
        local regex = "\\\\d+"
        local m = ngx.re.match("hello, 1234", regex)
        if m then ngx.say(m[0]) else ngx.say("not matched!") end
    ';
}
# evaluates to "1234"

Here, \\\\d+ is stripped down to \\d+ by the Nginx config file parser and this is further stripped down to \d+ by the Lua language parser before running.

Alternatively, the regex pattern can be presented as a long-bracketed Lua string literal by encasing it in "long brackets", [[...]], in which case backslashes have to only be escaped once for the Nginx config file parser.

# nginx.conf
location /test {
    content_by_lua '
        local regex = [[\\d+]]
        local m = ngx.re.match("hello, 1234", regex)
        if m then ngx.say(m[0]) else ngx.say("not matched!") end
    ';
}
# evaluates to "1234"

Here, [[\\d+]] is stripped down to [[\d+]] by the Nginx config file parser and this is processed correctly.

Note that a longer from of the long bracket, [=[...]=], may be required if the regex pattern contains [...] sequences. The [=[...]=] form may be used as the default form if desired.

# nginx.conf
location /test {
    content_by_lua '
        local regex = [=[[0-9]+]=]
        local m = ngx.re.match("hello, 1234", regex)
        if m then ngx.say(m[0]) else ngx.say("not matched!") end
    ';
}
# evaluates to "1234"

An alternative approach to escaping PCRE sequences is to ensure that Lua code is placed in external script files and executed using the various *_by_lua_file directives. With this approach, the backslashes are only stripped by the Lua language parser and therefore only need to be escaped once each.

-- test.lua
local regex = "\\d+"
local m = ngx.re.match("hello, 1234", regex)
if m then ngx.say(m[0]) else ngx.say("not matched!") end
-- evaluates to "1234"

Within external script files, PCRE sequences presented as long-bracketed Lua string literals do not require modification.

-- test.lua
local regex = [[\d+]]
local m = ngx.re.match("hello, 1234", regex)
if m then ngx.say(m[0]) else ngx.say("not matched!") end
-- evaluates to "1234"

Typical Uses

Just to name a few:

Mashup'ing and processing outputs of various nginx upstream outputs (proxy, drizzle, postgres, redis, memcached, and etc) in Lua,
doing arbitrarily complex access control and security checks in Lua before requests actually reach the upstream backends,
manipulating response headers in an arbitrary way (by Lua)
fetching backend information from external storage backends (like redis, memcached, mysql, postgresql) and use that information to choose which upstream backend to access on-the-fly,
coding up arbitrarily complex web applications in a content handler using synchronous but still non-blocking access to the database backends and other storage,
doing very complex URL dispatch in Lua at rewrite phase,
using Lua to implement advanced caching mechanism for Nginx's subrequests and arbitrary locations.

The possibilities are unlimited as the module allows bringing together various elements within Nginx as well as exposing the power of the Lua language to the user. The module provides the full flexibility of scripting while offering performance levels comparable with native C language programs both in terms of CPU time as well as memory footprint. This is particularly the case when LuaJIT 2.0 is enabled.

Other scripting language implementations typically struggle to match this performance level.

The Lua state (Lua VM instance) is shared across all the requests handled by a single nginx worker process to minimize memory use.

On a ThinkPad T400 2.80 GHz laptop, the Hello World example readily achieves 28k req/sec using http_load -p 10. By contrast, Nginx + php-fpm 5.2.8 + Unix Domain Socket yields 6k req/sec and Node.js v0.6.1 yields 10.2k req/sec for their Hello World equivalents.

Nginx Compatibility

The module is compatible with the following versions of Nginx:

1.3.x (last tested: 1.3.4)
1.2.x (last tested: 1.2.1)
1.1.x (last tested: 1.1.5)
1.0.x (last tested: 1.0.15)
0.9.x (last tested: 0.9.4)
0.8.x >= 0.8.54 (last tested: 0.8.54)

Code Repository

The code repository of this project is hosted on github at chaoslawful/lua-nginx-module.

Installation

The ngx_openresty bundle can be used to install Nginx, ngx_lua, either one of the standard Lua 5.1 interpreter or LuaJIT 2.0, as well as a package of powerful companion Nginx modules. The basic installation step is a simple ./configure --with-luajit && make && make install.

Alternatively, ngx_lua can be manually compiled into Nginx:

Install LuaJIT 2.0 (Recommended) or Lua 5.1 (Lua 5.2 is not supported yet). LuajIT can be downloaded from the the LuaJIT project website and Lua 5.1, from the Lua project website. Some distribution package managers also distribute LuajIT and/or Lua.
Download the latest version of the ngx_devel_kit (NDK) module HERE.
Download the latest version of ngx_lua HERE.
Download the latest version of Nginx HERE (See Nginx Compatibility)

Build the source with this module:

wget 'http://nginx.org/download/nginx-1.2.1.tar.gz'
tar -xzvf nginx-1.2.1.tar.gz
cd nginx-1.2.1/

# tell nginx's build system where to find LuaJIT:
export LUAJIT_LIB=/path/to/luajit/lib
export LUAJIT_INC=/path/to/luajit/include/luajit-2.0

# or tell where to find Lua if using Lua instead:
#export LUA_LIB=/path/to/lua/lib
#export LUA_INC=/path/to/lua/include

# Here we assume Nginx is to be installed under /opt/nginx/.
./configure --prefix=/opt/nginx \
        --add-module=/path/to/ngx_devel_kit \
        --add-module=/path/to/lua-nginx-module

make -j2
make install

Installation on Ubuntu 11.10

Note that it is recommended to use LuaJIT 2.0 instead of the standard Lua 5.1 interpreter where possible.

If the standard Lua 5.1 interpreter is required however, run the following command to install it from the Ubuntu repository:

apt-get install -y lua5.1 liblua5.1-0 liblua5.1-0-dev

Everything should be installed correctly, except for one small tweak.

Library name liblua.so has been changed in liblua5.1 package, it only comes with liblua5.1.so, which needs to be symlinked to /usr/lib so it could be found during the configuration process.

ln -s /usr/lib/x86_64-linux-gnu/liblua5.1.so /usr/lib/liblua.so

Community

English Mailing List

The openresty-en mailing list is for English speakers.

Chinese Mailing List

The openresty mailing list is for Chinese speakers.

Bugs and Patches

Please report bugs or submit patches by:

Creating a ticket on the GitHub Issue Tracker (Recommended)
Posting to the OpenResty community.

TODO

Short Term

review and apply Brian Akin's patch for the new directive lua_socket_log_errors.
review and apply Brian Akin's patch for the new shdict:flush_expired() API.
implement the SSL cosocket API.
review and apply Jader H. Silva's patch for ngx.re.split().
review and apply vadim-pavlov's patch for ngx.location.capture's extra_headers option
use ngx_hash_t to optimize the built-in header look-up process for ngx.req.set_header, ngx.header.HEADER, and etc.
add configure options for different strategies of handling the cosocket connection exceeding in the pools.
add directives to run Lua codes when nginx stops.
add APIs to access cookies as key/value pairs.
add ignore_resp_headers, ignore_resp_body, and ignore_resp options to ngx.location.capture and ngx.location.capture_multi methods, to allow micro performance tuning on the user side.
implement new directive lua_ignore_client_abort.

Longer Term

add lightweight thread API (i.e., the ngx.thread API) as demonstrated in this sample code.
add Lua code automatic time slicing support by yielding and resuming the Lua VM actively via Lua's debug hooks.
add stat mode similar to mod_lua.

Changes

The changes of every release of this module can be obtained from the ngx_openresty bundle's change logs:

http://openresty.org/#Changes

Test Suite

The following dependencies are required to run the test suite:

Nginx version >= 0.8.54
Perl modules:
- test-nginx: http://github.com/agentzh/test-nginx
Nginx modules:
- echo-nginx-module: http://github.com/agentzh/echo-nginx-module
- drizzle-nginx-module: http://github.com/chaoslawful/drizzle-nginx-module
- rds-json-nginx-module: http://github.com/agentzh/rds-json-nginx-module
- set-misc-nginx-module: http://github.com/agentzh/set-misc-nginx-module
- headers-more-nginx-module: http://github.com/agentzh/headers-more-nginx-module
- memc-nginx-module: http://github.com/agentzh/memc-nginx-module
- srcache-nginx-module: http://github.com/agentzh/srcache-nginx-module
- ngx_auth_request: http://mdounin.ru/hg/ngx_http_auth_request_module/
C libraries:
- yajl: https://github.com/lloyd/yajl
Lua modules:
- lua-yajl: https://github.com/brimworks/lua-yajl
  - Note: the compiled module has to be placed in '/usr/local/lib/lua/5.1/'
Applications:
- mysql: create database 'ngx_test', grant all privileges to user 'ngx_test', password is 'ngx_test'
- memcached

The order in which these modules are added during configuration is important as the position of any filter module in the filtering chain determines the final output. The correct adding order is:

ngx_devel_kit
set-misc-nginx-module
ngx_http_auth_request_module
echo-nginx-module
memc-nginx-module
lua-nginx-module (i.e. this module)
headers-more-nginx-module
srcache-nginx-module
drizzle-nginx-module
rds-json-nginx-module

Copyright and License

This module is licensed under the BSD license.

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

Translations

Chinese

200 KiB Исходник Ответственный История

Name

Status

Version

Synopsis

Description

Directives

lua_code_cache

lua_regex_cache_max_entries

lua_package_path

lua_package_cpath

init_by_lua

init_by_lua_file

set_by_lua

set_by_lua_file

content_by_lua

content_by_lua_file

rewrite_by_lua

rewrite_by_lua_file

access_by_lua

access_by_lua_file

header_filter_by_lua

header_filter_by_lua_file

body_filter_by_lua

body_filter_by_lua_file

log_by_lua

log_by_lua_file

lua_need_request_body

lua_shared_dict

lua_socket_connect_timeout

lua_socket_send_timeout

lua_socket_send_lowat

lua_socket_read_timeout

lua_socket_buffer_size

lua_socket_pool_size

lua_socket_keepalive_timeout

lua_socket_log_errors

lua_http10_buffering

rewrite_by_lua_no_postpone

lua_transform_underscores_in_response_headers

Nginx API for Lua

Introduction

ngx.arg

ngx.var.VARIABLE

Core constants

HTTP method constants

HTTP status constants

Nginx log level constants

print

ngx.ctx

ngx.location.capture

ngx.location.capture_multi

ngx.status

ngx.header.HEADER

ngx.req.get_method

ngx.req.set_method

ngx.req.set_uri

ngx.req.set_uri_args

ngx.req.get_uri_args

ngx.req.get_post_args

ngx.req.get_headers

ngx.req.set_header

ngx.req.read_body

ngx.req.discard_body

ngx.req.get_body_data

ngx.req.get_body_file

ngx.req.set_body_data

ngx.req.set_body_file

ngx.req.init_body

ngx.req.append_body

ngx.req.finish_body

ngx.req.socket

ngx.req.clear_header

ngx.exec

ngx.redirect

ngx.send_headers

ngx.headers_sent

ngx.print

ngx.say

ngx.log

200 KiB

Исходник Ответственный История