What’s new in Tornado 1.0

July 22, 2010

We are pleased to announce the release of Tornado 1.0, available
There have been many changes since version 0.2; here are some of
the highlights:

New features:
* Improved support for running other WSGI applications in a
  Tornado server (tested with Django and CherryPy)
* Improved performance on Mac OS X and BSD (kqueue-based IOLoop),
  and experimental support for win32
* Rewritten AsyncHTTPClient available as
  tornado.httpclient.AsyncHTTPClient2 (this will become the
  default in a future release)
* Support for standard .mo files in addition to .csv in the locale
* Pre-forking support for running multiple Tornado processes at
  once (see HTTPServer.start())
* SSL and gzip support in HTTPServer
* reverse_url() function refers to urls from the Application
  config by name from templates and RequestHandlers
* RequestHandler.on_connection_close() callback is called when the
  client has closed the connection (subject to limitations of the
  underlying network stack, any proxies, etc)
* Static files can now be served somewhere other than /static/ via
  the static_url_prefix application setting
* URL regexes can now use named groups ("(?P<name>)") to pass
  arguments to get()/post() via keyword instead of position
* HTTP header dictionary-like objects now support multiple values
  for the same header via the get_all() and add() methods.
* Several new options in the httpclient module, including
  prepare_curl_callback and header_callback
* Improved logging configuration in tornado.options.
* UIModule.html_body() can be used to return html to be inserted
  at the end of the document body.

Backwards-incompatible changes:
* RequestHandler.get_error_html() now receives the exception
  object as a keyword argument if the error was caused by an
  uncaught exception.
* Secure cookies are now more secure, but incompatible with
  cookies set by Tornado 0.2.  To read cookies set by older
  versions of Tornado, pass include_name=False to
* Parameters passed to RequestHandler.get/post() by extraction
  from the path now have %-escapes decoded, for consistency with
  the processing that was already done with other query

Many thanks to everyone who contributed patches, bug reports, and
feedback that went into this release!