Name
    ngx_echo - Brings "echo", "sleep", "time", "exec" and more shell-style
    goodies to Nginx config file.

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

Version
    This document describes echo-nginx-module v0.34
    (<http://github.com/agentzh/echo-nginx-module/tarball/v0.34 >) released
    on September 14, 2010.

Synopsis
      location /hello {
        echo "hello, world!";
      }

      location /hello {
        echo -n "hello, "
        echo "world!";
      }

      location /timed_hello {
        echo_reset_timer;
        echo hello world;
        echo "'hello world' takes about $echo_timer_elapsed sec.";
        echo hiya igor;
        echo "'hiya igor' takes about $echo_timer_elapsed sec.";
      }

      location /echo_with_sleep {
        echo hello;
        echo_flush;  # ensure the client can see previous output immediately
        echo_sleep   2.5;  # in sec
        echo world;
      }

      # in the following example, accessing /echo yields
      #   hello
      #   world
      #   blah
      #   hiya
      #   igor
      location /echo {
          echo_before_body hello;
          echo_before_body world;
          proxy_pass $scheme://127.0.0.1:$server_port$request_uri/more;
          echo_after_body hiya;
          echo_after_body igor;
      }
      location /echo/more {
          echo blah;
      }

      # the output of /main might be
      #   hello
      #   world
      #   took 0.000 sec for total.
      # and the whole request would take about 2 sec to complete.
      location /main {
          echo_reset_timer;

          # subrequests in parallel
          echo_location_async /sub1;
          echo_location_async /sub2;

          echo "took $echo_timer_elapsed sec for total.";
      }
      location /sub1 {
          echo_sleep 2;
          echo hello;
      }
      location /sub2 {
          echo_sleep 1;
          echo world;
      }

      # the output of /main might be
      #   hello
      #   world
      #   took 3.003 sec for total.
      # and the whole request would take about 3 sec to complete.
      location /main {
          echo_reset_timer;

          # subrequests in series (chained by CPS)
          echo_location /sub1;
          echo_location /sub2;

          echo "took $echo_timer_elapsed sec for total.";
      }
      location /sub1 {
          echo_sleep 2;
          echo hello;
      }
      location /sub2 {
          echo_sleep 1;
          echo world;
      }

      # Accessing /dup gives
      #   ------ END ------
      location /dup {
        echo_duplicate 3 "--";
        echo_duplicate 1 " END ";
        echo_duplicate 3 "--";
        echo;
      }   

      # /bighello will generate 1000,000,000 hello's.
      location /bighello {
        echo_duplicate 1000_000_000 'hello';
      }

      # echo back the client request
      location /echoback {
        echo_duplicate 1 $echo_client_request_headers;
        echo "\r";

        echo_read_request_body;

        echo_request_body;
      }

      # GET /multi will yields
      #   querystring: foo=Foo
      #   method: POST
      #   body: hi
      #   content length: 2
      #   ///
      #   querystring: bar=Bar
      #   method: PUT
      #   body: hello
      #   content length: 5
      #   ///
      location /multi {
          echo_subrequest_async POST '/sub' -q 'foo=Foo' -b 'hi';
          echo_subrequest_async PUT '/sub' -q 'bar=Bar' -b 'hello';
      }
      location /sub {
          echo "querystring: $query_string";
          echo "method: $echo_request_method";
          echo "body: $echo_request_body";
          echo "content length: $http_content_length";
          echo '///';
      }

      # GET /merge?/foo.js&/bar/blah.js&/yui/baz.js will merge the .js resources together
      location /merge {
          default_type 'text/javascript';
          echo_foreach_split '&' $query_string;
              echo "/* JS File $echo_it */";
              echo_location_async $echo_it;
              echo;
          echo_end;
      }

      # accessing /if?val=abc yields the "hit" output
      # while /if?val=bcd yields "miss":
      location ^~ /if {
          set $res miss;
          if ($arg_val ~* '^a') {
              set $res hit;
              echo $res;
          }
          echo $res;
      }

Description
    This module wraps lots of Nginx internal APIs for streaming input and
    output, parallel/sequential subrequests, timers and sleeping, as well as
    various meta data accessing.

    Basically it provides various utilities that help testing and debugging
    of other modules by trivially emulating different kinds of faked
    subrequest locations.

    People will also find it useful in real-world applications that need to

    1.  serve static contents directly from memory (loading from the Nginx
        config file).

    2.  wrap the upstream response with custom header and footer (kinda like
        the addition module but with contents read directly from the config
        file and Nginx variables).

    3.  merge contents of various "Nginx locations" (i.e., subrequests)
        together in a single main request (using echo_location and its
        friends).

    This is a special dual-role module that can *lazily* serve as a content
    handler or register itself as an output filter only upon demand. By
    default, this module does not do anything at all.

    Use of any of this module's directives (no matter content handler
    directives or filter directives) will force the chunked encoding to be
    used for the HTTP response due to the streaming nature of this module
    (unless HTTP 1.0 is enforced by the client and the Content-Length header
    will be set to the size of the first handler directive that generates
    contents).

    Technially, this module has also demonstrated the following techniques
    that might be helpful for module writers:

    1.  Issue parallel subreqeusts directly from content handler.

    2.  Issue chained subrequests directly from content handler, by passing
        continuation along the subrequest chain.

    3.  Issue subrequests with all HTTP 1.1 methods and even an optional
        faked HTTP request body.

    4.  Interact with the Nginx event model directly from content handler
        using custom events and timers, and resume the content handler back
        if necessary.

    5.  Dual-role module that can (lazily) serve as a content handler or an
        output filter or both.

    6.  Nginx config file variable creation and interpolation.

    7.  Streaming output control using output_chain, flush and its friends.

    8.  Read client request body from the content handler, and returns back
        (asynchronously) to the content handler after completion.

    9.  Use Perl-based declarative test suite to drive the development of
        Nginx C modules.

Content Handler Directives
    Use of the following directives register this module to the current
    Nginx location as a content handler. If you want to use another module,
    like the standard proxy module, as the content handler, use the filter
    directives provided by this module.

    All the content handler directives can be mixed together in a single
    Nginx location and they're supposed to run sequentially just as in the
    Bash scripting language.

    Every content handler directive supports variable interpolation in its
    arguments (if any).

    The MIME type set by the standard default_type directive is respected by
    this module, as in:

      location /hello {
        default_type text/plain;
        echo hello;
      }

    Then on the client side:

      $ curl -I 'http://localhost/echo'
      HTTP/1.1 200 OK
      Server: nginx/0.8.20
      Date: Sat, 17 Oct 2009 03:40:19 GMT
      Content-Type: text/plain
      Connection: keep-alive

    Since the v0.22 release, all of the directives are allowed in the
    rewrite module's if directive block, for instance:

        location ^~ /if {
            set $res miss;
            if ($arg_val ~* '^a') {
                set $res hit;
                echo $res;
            }
            echo $res;
        }

  echo
    syntax: *echo [options] <string>...*

    default: *no*

    context: *location*

    Sends arguments joined by spaces, along with a trailing newline, out to
    the client.

    Note that the data might be buffered by Nginx's underlying buffer. To
    force the output data flushed immediately, use the echo_flush command
    just after "echo", as in

       echo hello world;
       echo_flush;

    When no argument is specified, *echo* emits the trailing newline alone,
    just like the *echo* command in shell.

    Variables may appear in the arguments. An example is

       echo The current request uri is $request_uri;

    where $request_uri is a variable exposed by the [[NginxHttpCoreModule]].

    This command can be used multiple times in a single location
    configuration, as in

        location /echo {
            echo hello;
            echo world;
        }

    The output on the client side looks like this

        $ curl 'http://localhost/echo'
        hello
        world

    Special characters like newlines ("\n") and tabs ("\t") can be escaped
    using C-style escaping sequences. But a notable exception is the dollar
    sign ("$"). As of Nginx 0.8.20, there's still no clean way to esacpe
    this characters. (A work-around might be to use a $echo_dollor variable
    that is always evaluated to the constant "$" character. This feature
    will possibly be introduced in a future version of this module.)

    As of the echo v0.28 release, one can suppress the trailing newline
    character in the output by using the "-n" option, as in

        location /echo {
            echo -n "hello, ";
            echo "world";
        }

    Accessing "/echo" gives

        $ curl 'http://localhost/echo'
        hello, world

    Leading "-n" in variable values won't take effect and will be emitted
    literally, as in

        location /echo {
            set $opt -n;
            echo $opt "hello,";
            echo "world";
        }

    This gives the following output

        $ curl 'http://localhost/echo'
        -n hello,
        world

    One can output leading "-n" literals and other options using the special
    "--" option like this

        location /echo {
            echo -- -n is an option;
        }

    which yields

        $ curl 'http://localhost/echo'
        -n is an option

  echo_duplicate
    syntax: *echo_duplicate <count> <string>*

    default: *no*

    context: *location*

    Outputs duplication of a string indicated by the second argument, using
    the times specified in the first argument.

    For instance,

      location /dup {
          echo_duplicate 3 "abc";
      }

    will lead to an output of "abcabcabc".

    Underscores are allowed in the count number, just like in Perl. For
    example, to emit 1000,000,000 instances of "hello, world":

      location /many_hellos {
          echo_duplicate 1000_000_000 "hello, world";
      }

    The "count" argument could be zero, but not negative. The second
    "string" argument could be an empty string ("") likewise.

    Unlike the echo directive, no trailing newline is appended to the
    result. So it's possible to "abuse" this directive as a
    no-trailing-newline version of echo by using "count" 1, as in

      location /echo_art {
          echo_duplicate 2 '---';
          echo_duplicate 1 ' END ';  # we don't want a trailing newline here
          echo_duplicate 2 '---';
          echo;  # we want a trailing newline here...
      }

    You get

      ------ END ------

    This directive was first introduced in version 0.11.

  echo_flush
    syntax: *echo_flush*

    default: *no*

    context: *location*

    Forces the data potentially buffered by underlying Nginx output filters
    to send immediately to the client side via socket.

    Note that techically the command just emits a ngx_buf_t object with
    "flush" slot set to 1, so certain weird third-party output filter module
    could still block it before it reaches Nginx's (last) write filter.

    This directive does not take any argument.

    Consider the following example:

      location /flush {
         echo hello;

         echo_flush;

         echo_sleep 1;
         echo world;
      }

    Then on the client side, using curl to access "/flush", you'll see the
    "hello" line immediately, but only after 1 second, the last "world"
    line. Without calling "echo_flush" in the example above, you'll most
    likely see no output until 1 second is elapsed due to the internal
    buffering of Nginx.

    This directive will fail to flush the output buffer in case of
    subrequests get involved. Consider the following example:

      location /main {
          echo_location_async /sub;
          echo hello;
          echo_flush;
      }
      location /sub {
          echo_sleep 1;
      }

    Then the client won't see "hello" appear even if "echo_flush" has been
    executed before the subrequest to "/sub" has actually started executing.
    The outputs of "/main" that are sent *after* echo_location_async will be
    postponed and buffered firmly.

    This does *not* apply to outputs sent before the subrequest initiated.
    For a modified version of the example given above:

      location /main {
          echo hello;
          echo_flush;
          echo_location_async /sub;
      }
      location /sub {
          echo_sleep 1;
      }

    The client will immediately see "hello" before "/sub" enters sleeping.

    See also echo, echo_sleep, and echo_location_async.

  echo_sleep
    syntax: *echo_sleep <seconds>*

    default: *no*

    context: *location*

    Sleeps for the time period specified by the argument, which is in
    seconds.

    This operation is non-blocking on server side, so unlike the
    echo_blocking_sleep directive, it won't block the whole Nginx worker
    process.

    The period might takes three digits after the decimal point and must be
    greater than 0.001.

    An example is

       location /echo_after_sleep {
           echo_sleep 1.234;
           echo resumed!;
       }

    Behind the scene, it sets up a per-request "sleep" ngx_event_t object,
    and adds a timer using that custom event to the Nginx event model and
    just waits for a timeout on that event. Because the "sleep" event is
    per-request, this directive can work in parallel subrequests.

  echo_blocking_sleep
    syntax: *echo_blocking_sleep <seconds>*

    default: *no*

    context: *location*

    This is a blocking version of the echo_sleep directive.

    See the documentation of echo_sleep for more detail.

    Behind the curtain, it calls the ngx_msleep macro provided by the Nginx
    core which maps to usleep on POSIX-compliant systems.

    Note that this directive will block the current Nginx worker process
    completely while being executed, so never use it in production
    environment.

  echo_reset_timer
    syntax: *echo_reset_timer*

    default: *no*

    context: *location*

    Reset the timer begin time to *now*, i.e., the time when this command is
    executed during request.

    The timer begin time is default to the starting time of the current
    request and can be overridden by this directive, potentially multiple
    times in a single location. For example:

      location /timed_sleep {
          echo_sleep 0.03;
          echo "$echo_timer_elapsed sec elapsed.";

          echo_reset_timer;

          echo_sleep 0.02;
          echo "$echo_timer_elapsed sec elapsed.";
      }

    The output on the client side might be

        $ curl 'http://localhost/timed_sleep'
        0.032 sec elapsed.
        0.020 sec elapsed.

    The actual figures you get on your side may vary a bit due to your
    system's current activities.

    Invocation of this directive will force the underlying Nginx timer to
    get updated to the current system time (regardless the timer resolution
    specified elsewhere in the config file). Furthermore, references of the
    $echo_timer_elapsed variable will also trigger timer update forcibly.

    See also echo_sleep and $echo_timer_elapsed.

  echo_read_request_body
    Explicitly reads request body so that the $request_body variable will
    always have non-empty values (unless the body is so big that it has been
    saved by Nginx to a local temporary file).

    Note that this might not be the original client request body because the
    current request might be a subrequest with a "artificial" body specified
    by its parent.

    This directive does not generate any output itself, just like
    echo_sleep.

    Here's an example for echo'ing back the original HTTP client request
    (both headers and body are included):

      location /echoback {
        echo_duplicate 1 $echo_client_request_headers;
        echo "\r";
        echo_read_request_body;
        echo $request_body;
      }

    The content of "/echoback" looks like this on my side (I was using
    Perl's LWP utility to access this location on the server):

      $ (echo hello; echo world) | lwp-request -m POST 'http://localhost/echoback'
      POST /echoback HTTP/1.1
      TE: deflate,gzip;q=0.3
      Connection: TE, close
      Host: localhost
      User-Agent: lwp-request/5.818 libwww-perl/5.820
      Content-Length: 12
      Content-Type: application/x-www-form-urlencoded

      hello
      world

    Because "/echoback" is the main request, $request_body holds the
    original client request body.

    Before Nginx 0.7.56, it makes no sense to use this directive because
    $request_body was first introduced in Nginx 0.7.58.

    This directive itself was first introduced in the echo module's v0.14
    release.

  echo_location_async
    syntax: *echo_location_async <location> [<url_args>]*

    default: *no*

    context: *location*

    Issue GET subrequest to the location specified (first argument) with
    optional url arguments specified in the second argument.

    As of Nginx 0.8.20, the "location" argument does *not* support named
    location, due to a limitation in the "ngx_http_subrequest" function. The
    same is true for its brother, the echo_location directive.

    A very simple example is

        location /main {
            echo_location_async /sub;
            echo world;
        }
        location /sub {
            echo hello;
        }

    Accessing "/main" gets

      hello
      world

    Calling multiple locations in parallel is also possible:

        location /main {
            echo_reset_timer;
            echo_location_async /sub1;
            echo_location_async /sub2;
            echo "took $echo_timer_elapsed sec for total.";
        }
        location /sub1 {
            echo_sleep 2; # sleeps 2 sec
            echo hello;
        }
        location /sub2 {
            echo_sleep 1; # sleeps 1 sec
            echo world;
        }

    Accessing "/main" yields

      $ time curl 'http://localhost/main'
      hello
      world
      took 0.000 sec for total.

      real  0m2.006s
      user  0m0.000s
      sys   0m0.004s

    You can see that the main handler "/main" does *not* wait the
    subrequests "/sub1" and "/sub2" to complete and quickly goes on, hence
    the "0.000 sec" timing result. The whole request, however takes
    approximately 2 sec in total to complete because "/sub1" and "/sub2" run
    in parallel (or "concurrently" to be more accurate).

    If you use echo_blocking_sleep in the previous example instead, then
    you'll get the same output, but with 3 sec total response time, because
    "blocking sleep" blocks the whole Nginx worker process.

    Locations can also take an optional querystring argument, for instance

        location /main {
            echo_location_async /sub 'foo=Foo&bar=Bar';
        }
        location /sub {
            echo $arg_foo $arg_bar;
        }

    Accessing "/main" yields

      $ curl 'http://localhost/main'
      Foo Bar

    Querystrings is *not* allowed to be concatenated onto the "location"
    argument with "?" directly, for example, "/sub?foo=Foo&bar=Bar" is an
    invalid location, and shouldn't be fed as the first argument to this
    directive.

    Due to an unknown bug in Nginx (it still exists in Nginx 0.8.20), the
    standard SSI module is required to ensure that the contents of the
    subrequests issued by this directive are correctly merged into the
    output chains of the main one. Fortunately, the SSI module is enabled by
    default during Nginx's "configure" process.

    If calling this directive without SSI module enabled, you'll get
    truncated response without contents of any subrequests and get an alert
    message in your Nginx's "error.log", like this:

      [alert] 24212#0: *1 the http output chain is empty, client: 127.0.0.1, ...

    Technically speaking, this directive is an example that Nginx content
    handler issues one or more subrequests directly. AFAIK, the fancyindex
    module (<https://connectical.com/projects/ngx-fancyindex/wiki >) also
    does such kind of things ;)

    Nginx named locations like @foo is *not* supported here.

    This directive is logically equivalent to the GET version of
    echo_subrequest_async. For example,

      echo_location_async /foo 'bar=Bar';

    is logically equivalent to

      echo_subrequest_async GET /foo -q 'bar=Bar';

    But calling this directive is slightly faster than calling
    echo_subrequest_async using "GET" because we don't have to parse the
    HTTP method names like "GET" and options like "-q".

    This directive is first introduced in version 0.09 of this module and
    requires at least Nginx 0.7.46.

  echo_location
    syntax: *echo_location <location> [<url_args>]*

    default: *no*

    context: *location*

    Just like the echo_location_async directive, but "echo_location" issues
    subrequests *in series* rather than in parallel. That is, the content
    handler directives following this directive won't be executed until the
    subrequest issued by this directive completes.

    The final response body is almost always equivalent to the case when
    echo_location_async is used instead, only if timing variables is used in
    the outputs.

    Consider the following example:

        location /main {
            echo_reset_timer;
            echo_location /sub1;
            echo_location /sub2;
            echo "took $echo_timer_elapsed sec for total.";
        }
        location /sub1 {
            echo_sleep 2;
            echo hello;
        }
        location /sub2 {
            echo_sleep 1;
            echo world;
        }

    The location "/main" above will take for total 3 sec to complete
    (compared to 2 sec if echo_location_async is used instead here). Here's
    the result in action on my machine:

      $ curl 'http://localhost/main'
      hello
      world
      took 3.003 sec for total.

      real  0m3.027s
      user  0m0.020s
      sys   0m0.004s

    This directive is logically equivalent to the GET version of
    echo_subrequest. For example,

      echo_location /foo 'bar=Bar';

    is logically equivalent to

      echo_subrequest GET /foo -q 'bar=Bar';

    But calling this directive is slightly faster than calling
    echo_subrequest using "GET" because we don't have to parse the HTTP
    method names like "GET" and options like "-q".

    Behind the scene, it creates an "ngx_http_post_subrequest_t" object as a
    *continuation* and passes it into the "ngx_http_subrequest" function
    call. Nginx will later reopen this "continuation" in the subrequest's
    "ngx_http_finalize_request" function call. We resumes the execution of
    the parent-request's content handler and starts to run the next
    directive (command) if any.

    Nginx named locations like @foo is *not* supported here.

    This directive was first introduced in the release v0.12.

    See also echo_location_async for more details about the meaning of the
    arguments.

  echo_subrequest_async
    syntax: *echo_subrequest_async <HTTP_method> <location> [-q <url_args>]
    [-b <request_body>]*

    default: *no*

    context: *location*

    Initiate an asynchronous subrequest using HTTP method, an optional url
    arguments (or querystring), and an option request body.

    This directive is very much like a generalized version of the
    echo_location_async directive.

    Here's a small example demonstrating its usage:

        location /multi {
            echo_subrequest_async POST '/sub' -q 'foo=Foo' -b 'hi';
            echo_subrequest_async PUT '/sub' -q 'bar=Bar' -b 'hello';
        }
        location /sub {
            echo "querystring: $query_string";
            echo "method: $echo_request_method";
            echo "body: $echo_request_body";
            echo "content length: $http_content_length";
            echo '///';
        }

    Then on the client side:

      $ curl 'http://localhost/multi'
      querystring: foo=Foo
      method: POST
      body: hi
      content length: 2
      ///
      querystring: bar=Bar
      method: PUT
      body: hello
      content length: 5
      ///

    Here's more funny example using the standard proxy module to handle the
    subrequest:

        location /main {
            echo_subrequest_async POST /sub -b 'hello, world';
        }
        location /sub {
            proxy_pass $scheme://127.0.0.1:$server_port/proxied;
        }
        location /proxied {
            echo "method: $echo_request_method.";

            # we need to read body explicitly here...or $echo_request_body
            #   will evaluate to empty ("")
            echo_read_request_body;

            echo "body: $echo_request_body.";
        }

    Then on the client side, we can see that

      $ curl 'http://localhost/main'
      method: POST.
      body: hello, world.

    Nginx named locations like @foo is *not* supported here.

    This directive was first introduced in the release v0.15.

    See also the echo_subrequest and echo_location_async directives.

  echo_subrequest
    syntax: *echo_subrequest_async <HTTP_method> <location> [-q <url_args>]
    [-b <request_body>]*

    default: *no*

    context: *location*

    This is the synchronous version of the echo_subrequest_async directive.
    And just like echo_location, it does not block the Nginx worker process
    (while echo_blocking_sleep does), rather, it uses continuation to pass
    control along the subrequest chain.

    See echo_subrequest_async for more details.

    Nginx named locations like @foo is *not* supported here.

    This directive was first introduced in the release v0.15.

  echo_foreach_split
    syntax: *echo_foreach_split <delimiter> <string>*

    default: *no*

    context: *location*

    Split the second argument "string" using the delimiter specified in the
    first argument, and then iterate through the resulting items. For
    instance:

      location /loop {
        echo_foreach_split ',' $arg_list;
          echo "item: $echo_it";
        echo_end;
      }

    Accessing /main yields

      $ curl 'http://localhost/loop?list=cat,dog,mouse'
      item: cat
      item: dog
      item: mouse

    As seen in the previous example, this directive should always be
    accompanied by an echo_end directive.

    Parallel "echo_foreach_split" loops are allowed, but nested ones are
    currently forbidden.

    The "delimiter" argument could contain *multiple* arbitrary characters,
    like

      echo_foreach_split '-a-' 'cat-a-dog-a-mouse';
        echo $echo_it;
      echo_end;

    Logically speaking, this looping structure is just the "foreach" loop
    combined with a "split" function call in Perl (using the previous
    example):

       foreach (split ',', $arg_list) {
           print "item $_\n";
       }

    People will also find it useful in merging multiple ".js" or ".css"
    resources into a whole. Here's an example:

      location /merge {
          default_type 'text/javascript';

          echo_foreach_split '&' $query_string;
              echo "/* JS File $echo_it */";
              echo_location_async $echo_it;
              echo;
          echo_end;
      }

    Then accessing /merge to merge the ".js" resources specified in the
    query string:

      $ curl 'http://localhost/merge?/foo/bar.js&/yui/blah.js&/baz.js'

    One can also use third-party Nginx cache module to cache the merged
    response generated by the "/merge" location in the previous example.

    This directive was first introduced in the release v0.17.

  echo_end
    syntax: *echo_end*

    default: *no*

    context: *location*

    This directive is used to terminate the body of looping and conditional
    control structures like echo_foreach_split.

    This directive was first introduced in the release v0.17.

  echo_request_body
    syntax: *echo_request_body*

    default: *no*

    context: *location*

    Outputs the contents of the request body previous read.

    Behind the scene, it's implemented roughly like this:

      if (r->request_body && r->request_body->bufs) {
          return ngx_http_output_filter(r, r->request_body->bufs);
      }

    Unlike the $echo_request_body and $request_body variables, this
    directive will show the whole request body even if some parts or all
    parts of it are saved in temporary files on the disk.

    It is a "no-op" if no request body has been read yet.

    This directive was first introduced in the release v0.18.

    See also echo_read_request_body and the chunkin module.

  echo_exec
    syntax: *echo_exec <location> [<query_string>]*

    syntax: *echo_exec <named_location>*

    default: *no*

    context: *location*

    Does an internal redirect to the location specified. An optional query
    string can be specified for normal locations, as in

      location /foo {
          echo_exec /bar weight=5;
      }
      location /bar {
          echo $arg_weight;
      }

    Or equivalently

      location /foo {
          echo_exec /bar?weight=5;
      }
      location /bar {
          echo $arg_weight;
      }

    Named locations are also supported. Here's an example:

      location /foo {
          echo_exec @bar;
      }
      location @bar {
          # you'll get /foo rather than @bar
          #  due to a potential bug in nginx.
          echo $echo_request_uri;
      }

    But query string (if any) will always be ignored for named location
    redirects due to a limitation in the "ngx_http_named_location" function.

    Never try to echo things before the "echo_exec" directive or you won't
    see the proper response of the location you want to redirect to. Because
    any echoing will cause the original location handler to send HTTP
    headers before the redirection happens.

    Technically speaking, this directive exposes the Nginx internal API
    functions "ngx_http_internal_redirect" and "ngx_http_named_location".

    This directive was first introduced in the v0.21 release.

Filter Directives
    Use of the following directives trigger the filter registration of this
    module. By default, no filter will be registered by this module.

    Every filter directive supports variable interpolation in its arguments
    (if any).

  echo_before_body
    syntax: *echo_before_body [options] [argument]...*

    default: *no*

    context: *location*

    It's the filter version of the echo directive, and prepends its output
    to the beginning of the original outputs generated by the underlying
    content handler.

    An example is

        location /echo {
            echo_before_body hello;
            proxy_pass $scheme://127.0.0.1:$server_port$request_uri/more;
        }
        location /echo/more {
            echo world
        }

    Accessing "/echo" from the client side yields

      hello
      world

    In the previous sample, we borrow the standard proxy module to serve as
    the underlying content handler that generates the "main contents".

    Multiple instances of this filter directive are also allowed, as in:

        location /echo {
            echo_before_body hello;
            echo_before_body world;
            echo !;
        }

    On the client side, the output is like

      $ curl 'http://localhost/echo'
      hello
      world
      !

    In this example, we also use the content handler directives provided by
    this module as the underlying content handler.

    This directive also supports the "-n" and "--" options like the echo
    directive.

    This directive can be mixed with its brother directive echo_after_body.

  echo_after_body
    syntax: *echo_after_body [argument]...*

    default: *no*

    context: *location*

    WARNING this directive does not work for nginx >= 0.7.65.

    It's very much like the echo_before_body directive, but *appends* its
    output to the end of the original outputs generated by the underlying
    content handler.

    Here's a simple example:

        location /echo {
            echo_after_body hello;
            proxy_pass http://127.0.0.1:$server_port$request_uri/more;
        }
        location /echo/more {
            echo world
        }

    Accessing "/echo" from the client side yields

      world
      hello

    Multiple instances are allowed, as in:

        location /echo {
            echo_after_body hello;
            echo_after_body world;
            echo i;
            echo say;
        }

    The output on the client side while accessing the "/echo" location looks
    like

      i
      say
      hello
      world

    This directive also supports the "-n" and "--" options like the echo
    directive.

    When this directive is used in a location accessed by a subrequest, it
    replies on the "sync" flag set in a chain buffer to indicate the end of
    the output for nginx >= 0.8.7. This is a hack because Nginx does not
    provide a reliable way to determine the end of the output chain in a
    subrequest's output filter. Use it in subrequests with care.

    This directive can be mixed with its brother directive echo_before_body.

Variables
  $echo_it
    This is a "topic variable" used by echo_foreach_split, just like the $_
    variable in Perl.

  $echo_timer_elapsed
    This variable holds the seconds elapsed since the start of the current
    request (might be a subrequest though) or the last invocation of the
    echo_reset_timer command.

    The timing result takes three digits after the decimal point.

    References of this variable will force the underlying Nginx timer to
    update to the current system time, regardless the timer resolution
    settings elsewhere in the config file, just like the echo_reset_timer
    directive.

  $echo_request_body
    Evaluates to the current (sub)request's request body previously read if
    no part of the body has been saved to a temporary file. To always show
    the request body even if it's very large, use the echo_request_body
    directive.

  $echo_request_method
    Evaluates to the HTTP request method of the current request (it can be a
    subrequest).

    Behind the scene, it just takes the string data stored in
    "r->method_name".

    Compare it to the $echo_client_request_method variable.

    At least for Nginx 0.8.20 and older, the $request_method variable
    provided by the http core module is actually doing what our
    $echo_client_request_method is doing.

    This variable was first introduced in our v0.15 release.

  $echo_client_request_method
    Always evaluates to the main request's HTTP method even if the current
    request is a subrequest.

    Behind the scene, it just takes the string data stored in
    "r->main->method_name".

    Compare it to the $echo_request_method variable.

    This variable was first introduced in our v0.15 release.

  $echo_client_request_headers
    Evaluates to the original client request's headers.

    Just as the name suggests, it will always take the main request (or the
    client request) even if it's currently executed in a subrequest.

    A simple example is below:

      location /echoback {
         echo "headers are:"
         echo $echo_client_request_headers;
      }

    Accessing "/echoback" yields

      $ curl 'http://localhost/echoback'
      headers are
      GET /echoback HTTP/1.1
      User-Agent: curl/7.18.2 (i486-pc-linux-gnu) libcurl/7.18.2 OpenSSL/0.9.8g
      Host: localhost:1984
      Accept: */*

    Behind the scene, it recovers "r->main->header_in" on the C level and
    does not construct the headers itself by traversing parsed results in
    the request object, and strips the last (trailing) CRLF.

    This variable was first introduced in version 0.15.

  $echo_cacheable_request_uri
    Evaluates to the parsed form of the URI (usually led by "/") of the
    current (sub-)request. Unlike the $echo_request_uri variable, it is
    cacheable.

    See $echo_request_uri for more details.

    This variable was first introduced in version 0.17.

  $echo_request_uri
    Evaluates to the parsed form of the URI (usually led by "/") of the
    current (sub-)request. Unlike the $echo_cacheable_request_uri variable,
    it is *not* cacheable.

    This is quite different from the $request_uri variable exported by the
    [[NginxHttpCoreModule]], because $request_uri is the *unparsed* form of
    the current request's URI.

    This variable was first introduced in version 0.17.

  $echo_incr
    It is a counter that always generate the current counting number,
    starting from 1. The counter is always associated with the main request
    even if it is accessed within a subrequest.

    Consider the following example

        location /main {
            echo "main pre: $echo_incr";
            echo_location_async /sub;
            echo_location_async /sub;
            echo "main post: $echo_incr";
        }
        location /sub {
            echo "sub: $echo_incr";
        }

    Accessing "/main" yields

        main pre: 1
        sub: 3
        sub: 4
        main post: 2

    This directive was first introduced in the v0.18 release.

  $echo_response_status
    Evaluates to the status code of the current (sub)request, null if not
    any.

    Behind the scene, it's just the textual representation of
    "r->headers_out->status".

    This directive was first introduced in the v0.23 release.

Installation
    Grab the nginx source code from nginx.net (<http://nginx.net/ >), for
    example, the version 0.8.41 (see nginx compatibility), and then build
    the source with this module:

        $ wget 'http://sysoev.ru/nginx/nginx-0.8.41.tar.gz'
        $ tar -xzvf nginx-0.8.41.tar.gz
        $ cd nginx-0.8.41/

        # Here we assume you would install you nginx under /opt/nginx/.
        $ ./configure --prefix=/opt/nginx \
            --add-module=/path/to/echo-nginx-module

        $ make -j2
        $ make install

    Download the latest version of the release tarball of this module from
    echo-nginx-module file list
    (<http://github.com/agentzh/echo-nginx-module/downloads >).

Compatibility
    The following versions of Nginx should work with this module:

    *   0.8.x (last tested version is 0.8.40)

    *   0.7.x >= 0.7.21 (last tested version is 0.7.66)

    In particular,

    *   the directive echo_location_async and its brother
        echo_subrequest_async do *not* work with 0.7.x < 0.7.46.

    *   the echo_after_body directive does *not* work at all with nginx <
        0.8.7.

    *   the echo_sleep directive cannot be used after echo_location or
        echo_subrequest for nginx < 0.8.11.

    Earlier versions of Nginx like 0.6.x and 0.5.x will *not* work at all.

    If you find that any particular version of Nginx above 0.7.21 does not
    work with this module, please consider reporting a bug.

Modules that use this module for testing
    The following modules take advantage of this "echo" module in their test
    suite:

    *   The memc module that supports almost the whole memcached TCP
        protocol.

    *   The chunkin module that adds HTTP 1.1 chunked input support to
        Nginx.

    *   The headers_more module that allows you to add, set, and clear input
        and output headers under the conditions that you specify.

    *   The "echo" module itself.

    Please mail me other modules that use "echo" in any form and I'll add
    them to the list above :)

Report Bugs
    Although a lot of effort has been put into testing and code tuning,
    there must be some serious bugs lurking somewhere in this module. So
    whenever you are bitten by any quirks, please don't hesitate to

    1.  send a bug report or even patches to <[email protected]>,

    2.  or create a ticket on the issue tracking interface
        (<http://github.com/agentzh/echo-nginx-module/issues >) provided by
        GitHub.

Source Repository
    Available on github at agentzh/echo-nginx-module
    (<http://github.com/agentzh/echo-nginx-module >).

ChangeLog
  v0.34
    *   we no longer use the problematic "ngx_strXcmp" macros in our source
        because it may cause invalid reads and thus segmentation faults.
        thanks Piotr Sikora.

  v0.33
    *   fixed compatibility with nginx 0.7.66+ because the ngx_time_update
        macro's parameter list has changed. Thanks Guang Feng (蔡镜明).

  v0.32
    *   we should have used "ngx_calloc_buf" instead of "ngx_alloc_buf" for
        the last chunk generated for echo_after_body. thanks valgrind's
        memcheck tool.

    *   we should initialize flags before feeding it into
        "ngx_http_parse_unsafe_uri". thanks valgrind's memcheck tool.

    *   fixed a minor issue in the echo_location/echo_subrequest
        implementation, which used to have race conditions.

  v0.31
    *   the echo wev handler should not proceed if it is still waiting for
        some sequential subrequest or has just processed one to avoid
        bouncing issues.

    *   fixed a segfault for echo_exec for 0.7.x: we should check "r->done"
        before proceeding.

    *   no longer explicitly set "r->write_event_handler" to
        "ngx_http_request_empty_handler" because it's totally wrong for the
        state machine.

    *   fixed the sequential subrequest model bugs: we should ensure the
        "pr->write_event_handler" gets called immediately after the
        "post_subrequest" callback when the subrequest finalizes.

  v0.30
    *   fixed the echo_exec directive for nginx >= 0.8.11. we didn't get the
        "r->main->count" right in the previous version.

  v0.29
    *   refactored the core of this module. now the implementation of
        echo_location, echo_subrequest, echo_sleep, and
        echo_read_request_body finally fit well with the nginx event model
        and Igor Sysoev's way of thinking.

  v0.28
    *   added support for the "-n" and "--" options to the echo,
        echo_before_body, and echo_after_body directives.

  v0.27
    *   applied the patch from Sergey A. Osokin to work with nginx 0.8.35.

  v0.26
    *   bug fix: we should bypass upstream filters in our echo filters. an
        output filter should ever call "ngx_http_output_filter" nor
        "ngx_http_send_special".

  v0.25
    *   now we register a request cleanup handler to ensure our sleep
        event's timer will always get properly deleted even if the request
        is quit prematurely. this affects the echo_sleep directive.

    *   use ngx_null_string whenever possible in the source.

    *   sync'd the bundled test scaffold to Test::Nginx 0.07.

  v0.24
    *   various source file name and coding style fixes. (the code now looks
        more like Igor Sysoev's.)

  v0.23
    *   now the subrequest can read the client request body directly (for
        the main request) because we made subrequests inherit its parent's
        "r->header_in" as well. This affects the echo_read_request_body
        directive.

    *   fixed echo_after_body in subrequests by using a hack (checking
        "cl->buf->sync" for the last buf) for nginx 0.8.7+ only.

    *   added new varaible $echo_response_status to help testing the status
        code of a subrequest. (The memc module makes use of it.)

    *   use the "ngx_calloc_buf" macro to allocate new bufs in the code
        rather than explicit "ngx_pcalloc" calls for safety.

  v0.22
    *   Now we allowed all the directives appear in the rewrite module's if
        block. But so far I've only tested the echo directive.

  v0.21
    *   Added a new directive named echo_exec which does internal redirect
        to other (named) locations.

  v0.20
    *   Fixed a bug in echo_sleep's "r->main->count" handling for nginx
        0.8.x. This bug will cause the server to hang when proxing a
        location with echo_sleep.

    *   Applied the "ngx_str3cmp", "ngx_str4cmp", and "ngx_str6cmp"
        optimizing macros to the "parse_method_name" function, as suggested
        by Marcus Clyne.

    *   Added TODO items regarding $echo_random and "echo_repeat" suggested
        by Marcus Clyne.

  v0.19
    *   Fixed the CPS-style chained subrequest model for the echo_location
        and echo_subrequest directives. they are now working perfectly and
        will not hang the server with the recent nginx 0.8.21 ~ 0.8.27
        releases. To be specifically, the chained subrequest should call
        "ngx_http_finalize_request" on its parent request if the content
        handler of the parent request does not return "NGX_DONE".

    *   Undeprecated the echo_location and echo_subrequest directives.

  v0.18
    *   Fixed the "zero size buf in output" alerts in error.log.

    *   Added the new directive echo_request_body.

    *   Now we use the "ngx_http_parse_unsafe_uri" function to check the
        locations to echo_location_async and its friends. Thanks Arvind
        Jayaprakash for suggesting this fix.

    *   Deprecated the echo_location and echo_subrequest directives.

    *   For HTTP 1.0 clients, use the buf length of the first chain link as
        the output header Content-Length.

    *   Implemented new variable $echo_incr.

  v0.17
    *   Added new directives echo_foreach_split and echo_end. Also
        introduced a "topic variable" named $echo_it.

    *   Added new variables $echo_request_uri and
        $echo_cacheable_request_uri.

  v0.16
    *   Now the subrequests issued by the echo_location_async and
        echo_location directives no longer inherit cached variable values
        from its parent request. (The underlying "ngx_http_subrequest"
        function, however, does automatic cachable variable value
        inheritance.)

    *   Added an undocumented variable *echo_cached_request_uri* to help
        testing of this module.

  v0.15
    *   Added new directives echo_subrequest and echo_subrequest_async for
        the full nginx subrequest API.

    *   Removed the "echo_client_request_headers" directive, and provided
        the $echo_client_request_headers variable instead.

    *   Added new variables $echo_request_method and
        $echo_client_request_method.

  v0.14
    *   Added new directive echo_read_request_body to explicitly read client
        request body so that the [[NginxHttpCoreModule#$request_body]]
        variable will always have non-empty values.

    *   Now we shuffer test cases automatically in .t files and fixed bugs
        in the tests themselves which are hidden by config reload fallback
        in failure.

  v0.13
    *   Fixed the special cases when the outputs of a echo_duplicate
        directive is empty.

    *   Now we explicitly clear content length and accept ranges headers in
        the content handler.

  v0.12
    *   Implemented the echo_location directive, which can issue chained GET
        subrequests in the Continuation Passing Style (CPS), rather than the
        parallel subrequest issued by the echo_location_async directive.

  v0.11
    *   Implemented the echo_duplicate directive to help generating large
        chunk of data for testing.

  v0.10
    *   Fixed compilation regression against Nginx 0.7.21. This bug appears
        in version 0.09.

    *   Refactored the codebase by splitting source into various small
        files.

  v0.09
    *   Reimplement the echo_sleep directive using per-request event and
        timer; the old implementation uses the global connection's
        read/write event to register timer, so it will break horribly when
        multiple subrequests "sleep" at the same time.

    *   Added the echo_location_async directive which can issue a GET
        subrequest and insert its contents herein.

  v0.08
    *   echo_sleep: now we delete our "write event timer" in the
        "post_sleep" handle.

    *   Added "doc/manpage.wiki" which tracks changes in the wiki page
        (<http://wiki.nginx.org/NginxHttpEchoModule >).

    *   Added the "util/wiki2pod.pl" script to convert "doc/manpage.wiki" to
        "README".

    *   Disabled the "DDEBUG" macro in the C source by default.

Test Suite
    This module comes with a Perl-driven test suite. The test cases
    (<http://github.com/agentzh/echo-nginx-module/tree/master/test/t/ >) are
    declarative
    (<http://github.com/agentzh/echo-nginx-module/blob/master/test/t/echo.t >
    ) too. Thanks to the Test::Base
    (<http://search.cpan.org/perldoc?Test::Base >) module in the Perl world.

    To run it on your side:

        $ cd test
        $ PATH=/path/to/your/nginx-with-echo-module:$PATH prove -r t

    You need to terminate any Nginx processes before running the test suite
    if you have changed the Nginx server binary.

    At the moment, LWP::UserAgent
    (<http://search.cpan.org/perldoc?LWP::UserAgent >) is used by the test
    scaffold
    (<http://github.com/agentzh/echo-nginx-module/blob/master/test/lib/Test/
    Nginx/Echo.pm>) for simplicity and it's rather weak in testing
    *streaming* behavior of Nginx (I'm using "curl" to test these aspects
    manually for now). I'm considering coding up my own Perl HTTP client
    library based on IO::Select
    (<http://search.cpan.org/perldoc?IO::Select >) and IO::Socket
    (<http://search.cpan.org/perldoc?IO::Socket >) (there might be already
    one around?).

    Because a single nginx server (by default, "localhost:1984") is used
    across all the test scripts (".t" files), it's meaningless to run the
    test suite in parallel by specifying "-jN" when invoking the "prove"
    utility.

    Some parts of the test suite requires standard modules proxy, rewrite
    and SSI to be enabled as well when building Nginx.

TODO
    *   Fix the echo_after_body directive in subrequests.

    *   Add directives *echo_read_client_request_body* and
        *echo_request_headers*.

    *   Add new directive *echo_log* to use Nginx's logging facility
        directly from the config file and specific loglevel can be
        specified, as in

      echo_log debug "I am being called.";

    *   Add support for options "-h" and "-t" to echo_subrequest_async and
        echo_subrequest. For example

      echo_subrequest POST /sub -q 'foo=Foo&bar=Bar' -b 'hello' -t 'text/plan' -h 'X-My-Header: blah blah'

    *   Add options to control whether a subrequest should inherit cached
        variables from its parent request (i.e. the current request that is
        calling the subrequest in question). Currently none of the
        subrequests issued by this module inherit the cached variables from
        the parent request.

    *   Add new variable *$echo_active_subrequests* to show "r->main->count
        - 1".

    *   Add the *echo_file* and *echo_cached_file* directives.

    *   Add new varaible *$echo_request_headers* to accompany the existing
        $echo_client_request_headers variable.

    *   Add new directive *echo_foreach*, as in

      echo_foreach 'cat' 'dog' 'mouse';
        echo_location_async "/animals/$echo_it";
      echo_end;

    *   Add new directive *echo_foreach_range*, as in

      echo_foreach_range '[1..100]' '[a-zA-z0-9]';
        echo_location_async "/item/$echo_it";
      echo_end;

    *   Add new directive *echo_repeat*, as in

      echo_repeat 10 $i {
          echo "Page $i";
          echo_location "/path/to/page/$i";
      }

    This is just another way of saying

      echo_foreach_range $i [1..10];
          echo "Page $i";
          echo_location "/path/to/page/$i";
      echo_end;

    Thanks Marcus Clyne for providing this idea.

    *   Add new variable $echo_random which always returns a random
        non-negative integer with the lower/upper limit specified by the new
        directives "echo_random_min" and "echo_random_max". For example,

      echo_random_min  10
      echo_random_max  200
      echo "random number: $echo_random";

    Thanks Marcus Clyne for providing this idea.

Getting involved
    You'll be very welcomed to submit patches to the author or just ask for
    a commit bit to the source repository on GitHub.

Author
    agentzh (章亦春) *<[email protected]>*

    This wiki page is also maintained by the author himself, and everybody
    is encouraged to improve this page as well.

Copyright & License
    Copyright (c) 2009, Taobao Inc., Alibaba Group ( http://www.taobao.com
    ).

    Copyright (c) 2009, agentzh <[email protected]>.

    This module is licensed under the terms of 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.

    *   Neither the name of the Taobao Inc. nor the names of its
        contributors may be used to endorse or promote products derived from
        this software without specific prior written permission.

    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.

See Also
    *   The original blog post
        (<http://agentzh.spaces.live.com/blog/cns!FF3A735632E41548!478.entry
        >) about this module's initial development.

    *   The standard addition filter module.

    *   The standard proxy module.