path: root/lib/common_test/doc/src/config_file_chapter.xml

<?xml version="1.0" encoding="latin1" ?>
<!DOCTYPE chapter SYSTEM "chapter.dtd">

<chapter>
  <header>
    <copyright>
      <year>2004</year><year>2012</year>
      <holder>Ericsson AB. All Rights Reserved.</holder>
    </copyright>
    <legalnotice>
      The contents of this file are subject to the Erlang Public License,
      Version 1.1, (the "License"); you may not use this file except in
      compliance with the License. You should have received a copy of the
      Erlang Public License along with this software. If not, it can be
      retrieved online at http://www.erlang.org/.

      Software distributed under the License is distributed on an "AS IS"
      basis, WITHOUT WARRANTY OF ANY KIND, either express or implied. See
      the License for the specific language governing rights and limitations
      under the License.

    </legalnotice>

    <title>External Configuration Data</title>
    <prepared>Siri Hansen, Peter Andersson</prepared>
    <docno></docno>
    <date></date>
    <rev></rev>
    <file>config_file_chapter.xml</file>
  </header>

    <marker id="top"></marker>

  <section>
    <title>General</title>

    <p>To avoid hard coding data values related to the test and/or SUT (System
    Under Test) in the test suites, the data may instead be specified by means
    of configuration files or strings that Common Test reads before
    the start of a test run. External configuration data makes it possible to
    change test properties without having to modify the actual test suites
    using the data. Examples of configuration data:</p>

    <list>
      <item>Addresses to the test plant or other instruments</item>
      <item>User login information</item>
      <item>Names of files needed by the test</item>
      <item>Names of programs that should be executed during the test</item>
      <item>Any other variable needed by the test</item>
    </list>

  </section>

  <section>
    <title>Syntax</title>

    <p>A configuration file can contain any number of elements of the type:</p>
    <pre>
      {CfgVarName,Value}.</pre>

    <p>where</p>
    <pre>
      CfgVarName = atom()
      Value = term() | [{CfgVarName,Value}]</pre>

  </section>

  <section>
    <title>Requiring and reading configuration data</title>
    <marker id="require_config_data"></marker>

    <p>In a test suite, one must <em>require</em> that a configuration 
    variable (<c>CfgVarName</c> in the definition above) exists before
    attempting to read the associated value in a test case or config function.</p>

    <p><c>require</c> is an assert statement that can be part of the <seealso
    marker="write_test_chapter#suite">test suite info function</seealso> or
    <seealso marker="write_test_chapter#info_function">test case info
    function</seealso>. If the required variable is not available, the
    test is skipped (unless a default value has been specified, see the
    <seealso marker="write_test_chapter#info_function">test case info
    function</seealso> chapter for details). There is also a function
    <seealso marker="ct#require-1"><c>ct:require/1/2</c></seealso> which can be called from a test case
    in order to check if a specific variable is available. The return 
    value from this function must be checked explicitly and appropriate 
    action be taken depending on the result (e.g. to skip the test case
    if the variable in question doesn't exist).</p>

    <p>A <c>require</c> statement in the test suite info- or test case 
    info-list should look like this:
    <c>{require,CfgVarName}</c> or <c>{require,AliasName,CfgVarName}</c>.
    The arguments <c>AliasName</c> and <c>CfgVarName</c> are the same as the
    arguments to <seealso marker="ct#require-1"><c>ct:require/1/2</c></seealso> which are described in the
    reference manual for <seealso marker="ct">ct</seealso>.
    <c>AliasName</c> becomes an alias for the configuration variable,
    and can be used as reference to the configuration data value.
    The configuration variable may be associated with an
    arbitrary number of alias names, but each name must be unique within
    the same test suite. There are two main uses for alias names:</p>
    <list>
      <item>They may be introduced to identify connections (see below).</item>
      <item>They may used to help adapt configuration data to a test suite 
        (or test case) and improve readability.</item>
    </list>
    <p>To read the value of a config variable, use the function
    <seealso marker="ct#get_config-1"><c>get_config/1/2/3</c></seealso>
    which is also described in the reference
    manual for <seealso marker="ct">ct</seealso>.</p>
    <p>Example:</p>
    <pre>
      suite() -> 
          [{require, domain, 'CONN_SPEC_DNS_SUFFIX'}].

      ...
      
      testcase(Config) ->
          Domain = ct:get_config(domain),
	  ...</pre>      
  </section>

  <section>
  <title>Using configuration variables defined in multiple files</title>
    <p>If a configuration variable is defined in multiple files and you 
      want to access all possible values, you may use the <seealso marker="ct#get_config-3"><c>ct:get_config/3</c></seealso>
      function and specify <c>all</c> in the options list. The values will then
      be returned in a list and the order of the elements corresponds to the order 
      that the config files were specified at startup. Please see 
      the <seealso marker="ct">ct</seealso> reference manual for details.</p>
  </section>

  <section>
    <title>Encrypted configuration files</title>
    <marker id="encrypted_config_files"></marker>
    <p>It is possible to encrypt configuration files containing sensitive data
      if these files must be stored in open and shared directories.</p> 
      <p>Call <seealso marker="ct#encrypt_config_file-2"><c>ct:encrypt_config_file/2/3</c></seealso> to have Common Test encrypt a
      specified file using the DES3 function in the OTP <c>crypto</c> application.
      The encrypted file can then be used as a regular configuration file,
      in combination with other encrypted files or normal text files. The key
      for decrypting the configuration file must be provided when running the test,
      however. This can be done by means of the <c>decrypt_key</c> or
      <c>decrypt_file</c> flag/option, or a key file in a predefined location.</p>
      
      <p>Common Test also provides decryption functions, 
      <seealso marker="ct#decrypt_config_file-2"><c>ct:decrypt_config_file/2/3</c></seealso>, for recreating the original text
      files.</p> 

      <p>Please see the <seealso marker="ct">ct</seealso> reference manual for
      more information.</p>
  </section>

  <section>
    <title>Opening connections by using configuration data</title>
    <p>There are two different methods for opening a connection
      by means of the support functions in e.g. <seealso marker="ct_ssh"><c>ct_ssh</c></seealso>, <seealso marker="ct_ftp"><c>ct_ftp</c></seealso>,
      and <seealso marker="ct_telnet"><c>ct_telnet</c></seealso>:</p>
    <list>
      <item>Using a configuration target name (an alias) as reference.</item>
      <item>Using the configuration variable as reference.</item>
    </list>    
    <p>When a target name is used for referencing the configuration data
      (that specifies the connection to be opened), the same name may be used 
      as connection identity in all subsequent calls related to the connection
      (also for closing it). It's only possible to have one open connection
      per target name. If attempting to open a new connection using a name
      already associated with an open connection, Common Test will
      return the already existing handle so that the previously opened connection
      will be used. This is a practical feature since it makes it possible to
      call the function for opening a particular connection whenever 
      useful. An action like this will not necessarily open any new 
      connections unless it's required (which could be the case if e.g. the 
      previous connection has been closed unexpectedly by the server).
      Another benefit of using named connections is that it's not
      necessary to pass handle references around in the suite for these 
      connections.
    </p>
    <p>When a configuration variable name is used as reference to the data
      specifying the connection, the handle returned as a result of opening
      the connection must be used in all subsequent calls (also for closing
      the connection). Repeated calls to the open function with the same
      variable name as reference will result in multiple connections
      being opened. This can be useful e.g. if a test case needs to open
      multiple connections to the same server on the target node (using the
      same configuration data for each connection).
    </p>
  </section>

  <section>
    <title>User specific configuration data formats</title>

    <p>It is possible for the user to specify configuration data on a
      different format than key-value tuples in a text file, as described
      so far. The data can e.g. be read from arbitrary files, fetched from
      the web over http, or requested from a user specific process.
      To support this, Common Test provides a callback module plugin
      mechanism to handle configuration data.</p>

    <section>
      <title>Default callback modules for handling configuration data</title>
      <p>The Common Test application includes default callback modules
	for handling configuration data specified in standard config files
	(see above) and in xml files:</p>
      <list>
        <item>
          <c>ct_config_plain</c> - for reading configuration files with
          key-value tuples (standard format). This handler will be used to
          parse configuration files if no user callback is specified.
        </item>
        <item>
          <c>ct_config_xml</c> - for reading configuration data from XML
          files.
        </item>
      </list>
    </section>

    <section>
      <title>Using XML configuration files</title>
      <p>This is an example of an XML configuration file:</p>
      <pre><![CDATA[
<config>
    <ftp_host>
        <ftp>"targethost"</ftp>
        <username>"tester"</username>
        <password>"letmein"</password>
    </ftp_host>
    <lm_directory>"/test/loadmodules"</lm_directory>
</config>]]></pre>

      <p>This configuration file, once read, will produce the same configuration
      variables as the following text file:</p>
      <pre>
{ftp_host, [{ftp,"targethost"},
            {username,"tester"},
            {password,"letmein"}]}.

{lm_directory, "/test/loadmodules"}.</pre>
    </section>

    <section>
      <title>How to implement a user specific handler</title>

      <p>The user specific handler can be written to handle special
	configuration file formats. The parameter can be either file
	name(s) or configuration string(s) (the empty list is valid).</p>

      <p>The callback module implementing the handler is responsible for
	checking correctness of configuration strings.</p>

      <p>To perform validation of the configuration strings, the callback module
	should have the following function exported:</p>

      <p><c>Callback:check_parameter/1</c></p>
      <p>The input argument will be passed from Common Test, as defined in the test
	specification or given as an option to <c>ct_run</c> or <c>ct:run_test</c>.</p>

      <p>The return value should be any of the following values indicating if given
	configuration parameter is valid:</p>
      <list>
        <item>
          <c>{ok, {file, FileName}}</c> - parameter is a file name and
          the file exists,
        </item>
        <item>
          <c>{ok, {config, ConfigString}}</c> - parameter is a config string
          and it is correct,
        </item>
        <item>
          <c>{error, {nofile, FileName}}</c> - there is no file with the given
          name in the current directory,
        </item>
        <item>
          <c>{error, {wrong_config, ConfigString}}</c> - the configuration string
          is wrong.
        </item>
      </list>

      <p>To perform reading of configuration data - initially before the tests
	start, or as a result of data being reloaded during test execution -
	the following function should be exported from the callback module:</p>

      <p><c>Callback:read_config/1</c></p>

      <p>The input argument is the same as for the <c>check_parameter/1</c> function.</p>
      <p>The return value should be either:</p>

      <list>
        <item>
          <c>{ok, Config}</c> - if the configuration variables are read successfully,
        </item>
        <item>
          <c>{error, {Error, ErrorDetails}}</c> - if the callback module fails to
          proceed with the given configuration parameters.
        </item>
      </list>
      <p><c>Config</c> is the proper Erlang key-value list, with possible
	key-value sublists as values, like for the configuration file
	example above:</p>

      <pre>
        [{ftp_host, [{ftp, "targethost"}, {username, "tester"}, {password, "letmein"}]},
         {lm_directory, "/test/loadmodules"}]</pre>

    </section>

  </section>

  <section>
    <title>Examples of configuration data handling</title>

    <p>A config file for using the FTP client to access files on a remote
      host could look like this:</p>

    <pre>
    {ftp_host, [{ftp,"targethost"},
                {username,"tester"},
                {password,"letmein"}]}.

    {lm_directory, "/test/loadmodules"}.</pre>

    <p>The XML version shown in the chapter above can also be used, but it should be
    explicitly specified that the <c>ct_config_xml</c> callback module is to be
    used by Common Test.</p>

    <p>Example of how to assert that the configuration data is available and
      use it for an FTP session:</p>
    <pre>
    init_per_testcase(ftptest, Config) ->
        {ok,_} = ct_ftp:open(ftp),
        Config.

    end_per_testcase(ftptest, _Config) ->
        ct_ftp:close(ftp).

    ftptest() ->
        [{require,ftp,ftp_host},
         {require,lm_directory}].

    ftptest(Config) ->
        Remote = filename:join(ct:get_config(lm_directory), "loadmodX"),
        Local = filename:join(?config(priv_dir,Config), "loadmodule"),
        ok = ct_ftp:recv(ftp, Remote, Local),
        ...</pre>
    
    <p>An example of how the above functions could be rewritten
      if necessary to open multiple connections to the FTP server:</p>
    <pre>
    init_per_testcase(ftptest, Config) ->
        {ok,Handle1} = ct_ftp:open(ftp_host),
        {ok,Handle2} = ct_ftp:open(ftp_host),
        [{ftp_handles,[Handle1,Handle2]} | Config].

    end_per_testcase(ftptest, Config) ->
        lists:foreach(fun(Handle) -> ct_ftp:close(Handle) end, 
                      ?config(ftp_handles,Config)).

    ftptest() ->
        [{require,ftp_host},
         {require,lm_directory}].

    ftptest(Config) ->
        Remote = filename:join(ct:get_config(lm_directory), "loadmodX"),
        Local = filename:join(?config(priv_dir,Config), "loadmodule"),
        [Handle | MoreHandles] = ?config(ftp_handles,Config),
        ok = ct_ftp:recv(Handle, Remote, Local),
        ...</pre>
      
  </section>

  <section>
    <title>Example of user specific configuration handler</title>
    <p>A simple configuration handling driver which will ask an external server for
      configuration data can be implemented this way:</p>
    <pre>
-module(config_driver).
-export([read_config/1, check_parameter/1]).

read_config(ServerName)->
    ServerModule = list_to_atom(ServerName),
    ServerModule:start(),
    ServerModule:get_config().

check_parameter(ServerName)->
    ServerModule = list_to_atom(ServerName),
    case code:is_loaded(ServerModule) of
        {file, _}->
            {ok, {config, ServerName}};
        false->
            case code:load_file(ServerModule) of
                {module, ServerModule}->
                    {ok, {config, ServerName}};
                {error, nofile}->
                    {error, {wrong_config, "File not found: " ++ ServerName ++ ".beam"}}
            end
    end.</pre>

    <p>The configuration string for this driver may be "config_server", if the
      config_server.erl module below is compiled and exists in the code path
      during test execution:</p>
    <pre>
-module(config_server).
-export([start/0, stop/0, init/1, get_config/0, loop/0]).

-define(REGISTERED_NAME, ct_test_config_server).

start()->
    case whereis(?REGISTERED_NAME) of
        undefined->
            spawn(?MODULE, init, [?REGISTERED_NAME]),
            wait();
        _Pid->
        ok
    end,
    ?REGISTERED_NAME.

init(Name)->
    register(Name, self()),
    loop().

get_config()->
    call(self(), get_config).

stop()->
    call(self(), stop).

call(Client, Request)->
    case whereis(?REGISTERED_NAME) of
        undefined->
            {error, {not_started, Request}};
        Pid->
            Pid ! {Client, Request},
            receive
                Reply->
                    {ok, Reply}
            after 4000->
                {error, {timeout, Request}}
            end
    end.

loop()->
    receive
        {Pid, stop}->
            Pid ! ok;
        {Pid, get_config}->
            {D,T} = erlang:localtime(),
            Pid !
                [{localtime, [{date, D}, {time, T}]},
                 {node, erlang:node()},
                 {now, erlang:now()},
                 {config_server_pid, self()},
                 {config_server_vsn, ?vsn}],
            ?MODULE:loop()
    end.

wait()->
    case whereis(?REGISTERED_NAME) of
        undefined->
            wait();
        _Pid->
            ok
    end.</pre>

    <p>In this example, the handler also provides the ability to dynamically reload
      configuration variables. If <c>ct:reload_config(localtime)</c> is called from
      the test case function, all variables loaded with <c>config_driver:read_config/1</c>
      will be updated with their latest values, and the new value for variable
      <c>localtime</c> will be returned.</p>
  </section>

</chapter>