Revision history for WWW::Scripter

0.027  18 March, 2012
     • The navigator object now has a cookieEnabled property.  Thanks
       to John Hulala for the suggestion.  (I wasn’t aware that brows-
       ers had this.)
     • Requests for scripts now use the  right  referrer  instead  of
       always  about:blank.  Thanks to John  Hulala  for  the  report
       [rt.cpan.org #75696].

0.026  11 March, 2012
       The history property is now listed in %WindowInterface,  allow-
       ing scripts to access it.  Thanks to John Hulala for reporting
       the issue.

0.025  1 January, 2012
     • The get, head, post, put and request methods now alias $_ to a
       new scalar to protect a tied  $_  belonging to the caller from
       modification by other modules WWW::Scripter depends on.
     • The status and defaultStatus methods have been removed. status
       conflicts with a WWW::Mechanize method. defaultStatus makes no
       sense without status.  These were added for completeness’ sake,
       for scripts to access,  but no scripts really  use  these  any
       more.  This removal might be temporary if I think of a way  to
       reintroduce them without conflicts.
     • A memory leak in the  ->frames  method has been fixed.  It was
       creating circular references that kept the window and document
       objects alive.

0.024  20 November, 2011
     • Link-fetching methods,  like find_all_links,  no longer die if
       there is no document,  which  happens  after  browsing  to  an
       unparsable or unretrievable URL. Thanks to Kengo Tsukamoto for
       reporting this [rt.cpan.org #72481].
     • WWW::Mechanize is no longer an optional prerequisite.  The rea-
       son for avoiding it  (test failures behind  a  firewall)  have
       been resolved, so there is no reason for WWW::Scripter to work
       around that.

0.023  4 September, 2011
     • WWW::Mechanize::Link objects returned by WWW::Scripter  (which
       are actually tied,  but that is an implementation  detail  sub-
       ject to change) can now be dumped via Data::Dumper, instead of
       producing a  ‘Can't  locate  object  method  FETCHSIZE’  error.
     • /../  at the beginning of a URL path is now stripped out  (and
       all its variations: /////../ /..//..///, etc.).
     • The error message displayed when a script cannot be loaded now
       uses the URL and  line  number  of  the  page  containing  the
       <script>  tag,  instead of a seemingly random line  number  in
       HTML/DOM.pm.

0.022  27 March, 2011
     • New status and defaultStatus window properties
     • New  screen  method that returns a screen object.  This object
       currently does nothing, but serves mainly as a placeholder for
       future use.

0.021  13 March, 2011
       The ‘open’ method now respects <base href> elements.

0.020  17 November, 2010
     • There are no longer circular references preventing the  window,
       document and response objects from being freed.
     • WWW::Mechanize  is  now  an  optional  prerequisite,  allowing
       smoother installation.

0.019  29 August, 2010
       Fix a dependency problem

0.018  22 August, 2010
       Loading a page that had a <meta>  tag with no http-equiv attri-
       bute  (e.g.,  <meta charset=utf-7>)  would cause uninitialized
       warnings in perl 5.12.0 and higher if  a  script  handler  was
       present. This has been fixed.

0.017  25 July, 2010
       Fixed an internal bug where  Hash::Util::FieldHash::Compat  is
       used without being loaded first.  It  just  happened  to  work
       because HTML::DOM 0.041 and earlier load it.

0.016  18 July, 2010
     • If a script handler’s event2sub method leaves something in  $@,
       it is now picked up  and  handed  off  to  the  ->warn  method.
     • check_timers  used to ignore some code ref timers if the  Java-
       Script plugin was not loaded.  It is only supposed  to  ignore
       strings of code. This has been fixed.

0.015  20 June, 2010
       location->assign has been added.

0.014  13 June, 2010
     • navigator->platform has been added.
     • Timers now use Time::HiRes,  instead of rounding to  the  near-
       est second.
     • New wait_for_timers method
     • A new navigator->taintEnabled method that always returns false
       has been added,  to appease MooTools 1.2,  which makes unrelia-
       ble assumptions based on the absence of this method.

0.013  2 May, 2010
       navigator->javaEnabled has been added. It always returns false.

0.012  25 April, 2010
       setInterval and clearInterval have been added.

0.011  18 April, 2010
     • Support for multiple ‘windows’:
       • New WWW::Scripter::WindowGroup class
       • New window_group, blur, close and focus methods
       • The open method now opens new windows,  returning the window
         it opened.
     • Support for link targets and <base target> tags, including the
       new find_target and name methods
     • navigator->appCodeName
     • follow_link and click now return the response object,  instead
       of returning nothing. (Thanks to  Hannes  Beinert  for  report-
       ing this.)
     • follow_link  no longer dies when autocheck is on  (unless it’s
       supposed to).
     • The WWW::Mechanize::Link and ::Image  objects returned by vari-
       ous methods now update automatically if  the  underlying  HTML
       element changes.
     • The history object’s  go  method now  reloads  the  page  with
       zeroes that are boolean-true (like the strings "00" and "0e0"),
       instead of just with false zeroes.
     • The title method now sets the page title if passed an argument.
     • Script handlers can now use the Mech interface to access forms
       during parsing.
     • Fetching a page no longer interferes with a tied $_.

0.010  15 November, 2009
     • New dom_enabled method that allows HTML parsing to be switched
       off.
     • Scripts served with a content type of text/html  are no longer
       parsed into inaccessible DOM trees. This used to cause warning
       messages if a JS file  when  read  as  HTML  appeared  to  con-
       tain scripts.
     • The document method now returns undef  if the page is not HTML,
       instead of returning the previous page’s document.

0.009  25 October, 2009
     • count_timers and check_timers now know about timers in
       (i)frames.
     • Adding an inline script to  the  DOM  dynamically  (e.g.,  via
       appendChild) no longer produces a warning.
     • Empty script elements with no src attribute  no  longer  cause
       errors.
     • The base URL for an about:blank page in a frame is  now  taken
       from the parent window.
     • check_timers  now catches script errors and  turns  them  into
       warnings. It used to die with code refs (introduced in the pre-
       vious version) and silence them with strings of code.
     • Location objects are no longer shared between frames  (causing
       navigation in a frame via the location object to navigate  the
       top window instead).
     • Event handlers with  call_with  methods no longer get passed a
       strange WWW::Scripter::EventTarget object.  (These objects had
       existed since version 0.005 and  were  *supposed*  to  be  for
       internal use,  but leaked out.  They were for keeping a  diffe-
       rent set of window event handlers for each page.)

0.008  19 October, 2009
     • The base method now uses HTML::DOM’s base method,  in order to
       work with multiple <base> tags and dynamic changes to the base.
     • We now include scroll, scrollBy and scrollTo methods,  in case
       a script tries to call them.  They don’t actually do  anything.
     • Frames with no  src  element or an empty src element  used  to
       load the parent page,  resulting in infinite  recursion.  This
       has been fixed.
     • setTimeout  now works with a code ref as  the  first  argument.
     • count_timers no longer clobbers the $_ variable.

0.007  9 October, 2009
       The  clone  method  now  copies  class_info  over.  class_info
       no longer dies when  called  on  a  clone  with  no  arguments
       in list or scalar context.

0.006  30 September, 2009
       WWW::Scripter now overrides  WWW::Mechanize’s submit method to
       trigger  onsubmit.  (Thanks to  Steven  Nikkel  for  stumbling
       across this.)

0.005  17 September, 2009
     • max_docs and max_history have been added. Mech’s stack_depth
       is  now  honoured.  (max_docs  is  an  off-by-one  alias  to
       stack_depth.)
     • When no  browsing  has  happened,  the  current  page  is  now
       about:blank.  (It used to be an uninitialised state  in  which
       some methods, such as uri, would just die.)
     • URL fragments are now supported.  They are now  stripped  from
       URLs when the actual request is made (they used to be sent  to
       the server).  Fetching a URL that differs from the current one
       only by its fragment does not reload the page,  but does add a
       history entry.  You can now use the location object to get the
       URL with its fragment.  Use the uri method to get the real URL.
     • location->replace now works correctly.  (It used to be a quick-
       and-dirty ‘back’ followed by ‘get’.)
     • location->reload  now actually reloads the  page.  In  version
       0.002,  it would simply refresh the  page  from  the  response
       object.  Version  0.003  broke  that  by  making  it  a  no-op.
     • Support for onunload handlers
     • onload handlers are now called on the window, not the body ele-
       ment, with the document as the event target.
     • WWW::Scripter  no longer inherits from  HTML::DOM::EventTarget,
       but  it  still  provides  the  interface,  so  it  now  has  a
       DOES method.
     • Setting the value of the location  object’s  hash  and  search
       attributes now works.  It used to be a no-op unless it was set
       to the old value.
     • The host, hostname, pathname, port, protocol, search and  href
       methods of the  location  object  now  return  the  old  value
       upon setting.
     • The host and hostname methods of the location object have been
       corrected  such  that  host  now  returns  the  port  as  well,
       hostname  no longer returns the port,  setting  hostname  to a
       string containing a port no longer sets the port  and  setting
       host with a port number now works.  (In other words, they were
       switched round and were a bit buggy, too.)
     • I forgot to add the  CSS  interface  to  the  class_info.  Now
       it’s there.

0.004  16 August, 2009
       The  frames  method no longer dies when the page is  not  HTML.

0.003  15 August, 2009
     • When you go back  (or forward)  to a page,  the same  document
       object is  used  once  more.  It  used  to  be  parsed  afresh.
     • Detecting the charset now works with LWP 5.827 and later.

0.002  9 April, 2009
       One bug fix: The credentials method now works.

0.001  5 April, 2009
       First release. This consists of code copied and pasted from
       parts of WWW::Mechanize::Plugin::DOM and the experimental
       plugin version of WWW::Mechanize at
       <http://www-mechanize.googlecode.com/svn/wm/branches/plugins/>,
       plus some new code to tape it all together.