WWW::Mechanize::Shell - An interactive shell for WWW::Mechanize
From the command line as
perl -MWWW::Mechanize::Shell -eshell
or alternatively as a custom shell program via :
#!/usr/bin/perl -w
use strict;
use WWW::Mechanize::Shell;
my $shell = WWW::Mechanize::Shell->new("shell");
if (@ARGV) {
$shell->source_file( @ARGV );
} else {
$shell->cmdloop;
};
This module implements a www-like shell above WWW::Mechanize and also has the capability to output crude Perl code that recreates the recorded session. Its main use is as an interactive starting point for automating a session through WWW::Mechanize.
The cookie support is there, but no cookies are read from your existing browser sessions. See HTTP::Cookies on how to implement reading/writing your current browsers cookies.
WWW::Mechanize::Shell->new %ARGSThis is the constructor for a new shell instance. Some of the options can be passed to the constructor as parameters.
By default, a file .mechanizerc (respectively mechanizerc under Windows) in the users home directory is executed before the interactive shell loop is entered. This can be used to set some defaults. If you want to supply a different filename for the rcfile, the rcfile parameter can be passed to the constructor :
rcfile => '.myapprc',
$shell->release_agentSince the shell stores a reference back to itself within the WWW::Mechanize instance, it is necessary to break this circular reference. This method does this.
$shell->source_file FILENAMEThe source_file method executes the lines of FILENAME as if they were typed in.
$shell->source_file( $filename );
$shell->display_user_warningAll user warnings are routed through this routine so they can be rerouted / disabled easily.
$shell->print_paged LISTPrints the text in LIST using $ENV{PAGER}. If $ENV{PAGER} is empty, prints directly to STDOUT. Most of this routine comes from the perldoc utility.
$shell->link_text LINKReturns a meaningful text from a WWW::Mechanize::Link object. This is (in order of precedence) :
$link->text
$link->name
$link->url
$shell->historyReturns the (relevant) shell history, that is, all commands that were not solely for the information of the user. The lines are returned as a list.
print join "\n", $shell->history;
$shell->scriptReturns the shell history as a Perl program. The lines are returned as a list. The lines do not have a one-by-one correspondence to the lines in the history.
print join "\n", $shell->script;
$shell->statusstatus is called for status updates.
$shell->display FILENAME LINESdisplay is called to output listings, currently from the history and script commands. If the second parameter is defined, it is the name of the file to be written, otherwise the lines are displayed to the user.
The shell implements various commands :
Leaves the shell.
Restart the shell.
This is mostly useful when you are modifying the shell itself. It dosen't work if you use the shell in oneliner mode with -e.
Download a specific URL.
This is used as the entry point in all sessions
Syntax:
get URL
Download a link into a file.
If more than one link matches the RE, all matching links are saved. The filename is taken from the last part of the URL. Alternatively, the number of a link may also be given.
Syntax:
save RE
Display the content for the current page.
Syntax: content [FILENAME]
If the FILENAME argument is provided, save the content to the file.
A trailing "\n" is added to the end of the content when using the shell, so this might not be ideally suited to save binary files without manual editing of the produced script.
Display the current page title as found in the <TITLE> tag.
Prints all <H1> through <H5> strings found in the content, indented accordingly. With an argument, prints only those levels; e.g., headers 145 prints H1,H4,H5 strings only.
Get/set the current user agent
Syntax:
# fake Internet Explorer ua "Mozilla/4.0 (compatible; MSIE 4.01; Windows 98)" # fake QuickTime v5 ua "QuickTime (qtver=5.0.2;os=Windows NT 5.0Service Pack 2)" # fake Mozilla/Gecko based ua "Mozilla/5.001 (windows; U; NT4.0; en-us) Gecko/25250101" # set empty user agent : ua ""
Display all links on a page
The links numbers displayed can used by open to directly select a link to follow.
Dump the output of HTML::TokeParser of the current content
Display all forms on the current page.
Select the form named NAME
If NAME matches /^\d+$/, it is assumed to be the (1-based) index of the form to select. There is no way of selecting a numerically named form by its name.
Dump the values of the current form
Set a form value
Syntax:
value NAME [VALUE]
Set checkbox marks
Syntax:
tick NAME VALUE(s)
If no value is given, all boxes are checked.
Remove checkbox marks
Syntax:
untick NAME VALUE(s)
If no value is given, all marks are removed.
submits the form without clicking on any button
Clicks on the button named NAME.
No regular expression expansion is done on NAME.
Syntax:
click NAME
If you have a button that has no name (displayed as NONAME), use
click ""
to click on it.
<open> accepts one argument, which can be a regular expression or the number of a link on the page, starting at zero. These numbers are displayed by the links function. It goes directly to the page if a number is used or if the RE has one match. Otherwise, a list of links matching the regular expression is displayed.
The regular expression should start and end with "/".
Syntax:
open [ RE | # ]
Go back one page in the browser page history.
Repeat the last request, thus reloading the current page.
Note that also POST requests are blindly repeated, as this command is mostly intended to be used when testing server side code.
Open the web browser with the current page
Displays the current page in the browser.
Set a shell option
Syntax:
set OPTION [value]
The command lists all valid options. Here is a short overview over the different options available :
autosync - automatically synchronize the browser window
autorestart - restart the shell when any required module changes
This does not work with C<-e> oneliners.
watchfiles - watch all required modules for changes
cookiefile - the file where to store all cookies
dumprequests - dump all requests to STDOUT
dumpresponses - dump the headers of the responses to STDOUT
verbose - print commands to STDERR as they are run,
when sourcing from a file
Display your current session history as the relevant commands.
Syntax:
history [FILENAME]
Commands that have no influence on the browser state are not added to the history. If a parameter is given to the history command, the history is saved to that file instead of displayed onscreen.
Display your current session history as a Perl script using WWW::Mechanize.
Syntax:
script [FILENAME]
If a parameter is given to the script command, the script is saved to that file instead of displayed on the console.
This command was formerly known as history.
Adds a comment to the script and the history. The comment is prepended with a \n to increase readability.
Fill out the current form
Interactively asks the values hat have no preset value via the autofill command.
Set basic authentication credentials.
Syntax:
auth [authority realm] user password
If you get back a 401, you can simply supply the matching user and password, as the authority and realm are already known :
>get http://www.example.com
Retrieving http://www.example.com(401)
http://www.example.com>auth corion secret
http://www.example.com>get http://www.example.com
Retrieving http://www.example.com(200)
http://www.example.com>
If you know the authority and the realm in advance, you can presupply the credentials, for example at the start of the script :
>auth www.example.com:80 secure_realm corion secret
>get http://www.example.com
Retrieving http://www.example.com(200)
http://www.example.com>
Display a table described by the columns COLUMNS.
Syntax:
table COLUMNS
Example:
table Product Price Description
If there is a table on the current page that has in its first row the three columns Product, Price and Description (not necessarily in that order), the script will display these columns of the whole table.
The HTML::TableExtract module is needed for this feature.
Display a list of tables.
Syntax:
tables
This command will display the top row for every table on the current page. This is convenient if you want to find out what the exact spellings for each column are.
The command does not always work nice, for example if a site uses tables for layout, it will be harder to guess what tables are irrelevant and what tables are relevant.
HTML::TableExtract is needed for this feature.
Set the cookie file name
Syntax:
cookies FILENAME
Define an automatic value
Sets a form value to be filled automatically. The NAME parameter is the WWW::Mechanize::FormFiller::Value subclass you want to use. For session fields, Keep is a good candidate, for interactive stuff, Ask is a value implemented by the shell.
A field name starting and ending with a slash (/) is taken to be a regular expression and will be applied to all fields with their name matching the expression. A field with a matching name still takes precedence over the regular expression.
Syntax:
autofill NAME [PARAMETERS]
Examples:
autofill login Fixed corion autofill password Ask autofill selection Random red green orange autofill session Keep autofill "/date$/" Random::Date string "%m/%d/%Y"
Evaluate Perl code and print the result
Syntax:
eval CODE
For the generated scripts, anything matching the regular expression /\$self->agent\b/ is automatically replaced by $agent in your eval code, to do the Right Thing.
Examples:
# Say hello eval "Hello World" # And take a look at the current content type eval $self->agent->ct
Execute a batch of commands from a file
Syntax:
source FILENAME
Print the version numbers of important modules
Syntax:
versions
Set new timeout value for the agent. Effects all subsequent requests. VALUE is in seconds.
Syntax:
timeout VALUE
prints the content type of the most current response.
Syntax:
ct
set the value of the Referer: header
Syntax:
referer URL referrer URL
Alias for referrer
display the last server response
$shell->munge_code( CODE )Munges a coderef to become code fit for output independent of WWW::Mechanize::Shell.
shellThis subroutine is exported by default as a convenience method so that the following oneliner invocation works:
perl -MWWW::Mechanize::Shell -eshell
You can pass constructor arguments to this routine as well. Any scripts given in @ARGV will be run. If @ARGV is empty, an interactive loop will be started.
# Search for a term on Google get http://www.google.com value q "Corions Homepage" click btnG script # (yes, this is a bad example of automating, as Google # already has a Perl API. But other sites don't)
get http://www.perlmonks.org open "/Saints in/" table User Experience Level script # now you have a program that gives you a csv file of # that table.
get http://aliens:xxxxx/ value f path/to/file click "upload"
# download prerelease versions of my modules get http://www.corion.net/perl-dev save /.tar.gz$/
Some commands take regular expressions as parameters. A regular expression must be a single parameter matching ^/.*/([isxm]+)?$, so you have to use quotes around it if the expression contains spaces :
/link_foo/ # will match as (?-xims:link_foo) "/link foo/" # will match as (?-xims:link foo)
Slashes do not need to be escaped, as the shell knows that a RE starts and ends with a slash :
/link/foo/ # will match as (?-xims:link/foo) "/link/ /foo/" # will match as (?-xims:link/\s/foo)
The /i modifier works as expected. If you desire more power over the regular expressions, consider dropping to Perl or recommend me a good parser module for regular expressions.
WWW::Mechanize::Shell now uses the module HTML::Display to display the HTML of the current page in your browser. Have a look at the documentation of HTML::Display how to make it use your browser of choice in the case it does not already guess it correctly.
If you want to stay within the confines of the shell, but still want to fill out forms using custom Perl code, here is a recipe how to achieve this :
Code passed to the eval command gets evalutated in the WWW::Mechanize::Shell namespace. You can inject new subroutines there and these get picked up by the Callback class of WWW::Mechanize::FormFiller :
# Fill in the "date" field with the current date/time as string
eval sub &::custom_today { scalar localtime };
autofill date Callback WWW::Mechanize::Shell::custom_today
fillout
This method can also be used to retrieve data from shell scripts :
# Fill in the "date" field with the current date/time as string
# works only if there is a program "date"
eval sub &::custom_today { chomp `date` };
autofill date Callback WWW::Mechanize::Shell::custom_today
fillout
As the namespace is different between the shell and the generated script, make sure you always fully qualify your subroutine names, either in your own namespace or in the main namespace.
The script command outputs a skeleton script that reproduces your actions as done in the current session. It pulls in WWW::Mechanize::FormFiller, which is possibly not needed. You should add some error and connection checking afterwards.
If you are automating a JavaScript dependent site, you will encounter JavaScript like this :
<script>
document.write( "<input type=submit name=submit>" );
</script>
HTML::Form will not know about this and will not have provided a submit button for you (understandably). If you want to create such a submit button from within your automation script, use the following code :
$agent->current_form->push_input( submit => { name => "submit", value =>"submit" } );
This also works for other dynamically generated input fields.
To fake an input field from within a shell session, use the eval command :
eval $self->agent->current_form->push_input(submit=>{name=>"submit",value=>"submit"});
And yes, the generated script should do the Right Thing for this eval as well.
If you want to use the shell on a local file without setting up a http server to serve the file, you can use the file: URI scheme to load it into the "browser":
get file:local.html forms
Currently, the proxy support is realized via a call to the env_proxy method of the WWW::Mechanize object, which loads the proxies from the environment. There is no provision made to prevent using proxies (yet). The generated scripts also load their proxies from the environment.
The online help feature is currently a bit broken in Term::Shell, but a fix is in the works. Until then, you can reenable the dynamic online help by patching Term::Shell :
Remove the three lines
my $smry = exists $o->{handlers}{$h}{smry}
? $o->summary($h)
: "undocumented";
in sub run_help and replace them by
my $smry = $o->summary($h);
The shell works without this patch and the online help is still available through perldoc WWW::Mechanize::Shell
Bug reports are very welcome - please use the RT interface at https://rt.cpan.org/NoAuth/Bugs.html?Dist=WWW-Mechanize-Shell or send a descriptive mail to bug-WWW-Mechanize-Shell@rt.cpan.org . Please try to include as much (relevant) information as possible - a test script that replicates the undesired behaviour is welcome every time!
auth command guesses the realm from the last received response. Currently a RE is used to extract the realm, but this fails with some servers resp. in some cases. Use the four parameter version of auth, or if not possible, code the extraction in Perl, either in the final script or through eval commands.WWW::Mechanize from HTML::Parser to XML::XMLlib or maybe easier, by tacking Class::XPath onto an HTML tree)head as a command ?The routine shell is exported into the importing namespace. This is mainly for convenience so you can use the following commandline invocation of the shell like with CPAN :
perl -MWWW::Mechanize::Shell -e"shell"
This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself.
Copyright (C) 2002,2003 Max Maischein
Max Maischein, <corion@cpan.org>
Please contact me if you find bugs or otherwise improve the module. More tests are also very welcome !