Skip to content


Subversion checkout URL

You can clone with
Download ZIP
Tinman is a Tornado support package including an application wrapper/runner and a set of handy decorators.
Python Shell

Fetching latest commit…

Cannot retrieve the latest commit at this time

Failed to load latest commit information.


Tinman is a take what you need package designed to speed development of Tornado applications. It includes an application wrapper and a toolbox of decorators and utilities.


  • A full featured application wrapper
  • Standard configuration for applications
  • RequestHandler output caching
  • Network address whitelisting decorator
  • Method/Function debug logging decorator
  • Automated connection setup for PostgreSQL, RabbitMQ and Redis (memcached, mongodb, mysql planned)
  • Support for a External Template Loaders including Tinman's CouchDB Template Loader
  • Flexible logging configuration allowing for custom formatters, filters handlers and setting logging levels for individual packages.


  • clihelper
  • ipaddr
  • pyyaml

Optional Dependencies

  • brukva
  • pika >= v0.9.5
  • psycopg2 >= 2.4.2

Unit tests

Tests are written to be run with nose and mock. They can be run with

python nosetests

Application Runner

The tinman application runner works off a YAML configuration file format and provides a convenient interface for running tornado applications interactively or as a daemon.

Command Line Syntax:

    Usage: tinman -c <configfile> [options]

    Tornado application wrapper

      --version             show program's version number and exit
      -h, --help            show this help message and exit
      -c CONFIG, --config=CONFIG
                            Specify the configuration file for use
      -f, --foreground      Run interactively in console


Application Options

The following are the keys that are available to be used for your Tinman/Tornado application.

  • base_path: The root of the files for the application
  • cookie_secret: A salt for signing cookies when using secure cookies
  • login_url: Login URL when using Tornado's @authenticated decorator
  • static_path: The path to static files
  • template_loader: The python module.Class to override the default template loader with
  • templates_path: The path to template files
  • transforms: A list of transformation objects to add to the application in module.Class format
  • translations_path: The path to translation files
  • ui_modules: Module for the UI modules classes (ie mysite.modules)
  • xsrf_cookies: Enable xsrf_cookie mode for forms
  • whitelist: List of IP addresses in CIDR notation if whitelist decorator is to be used

When setting the template_path or static_path values in the configuration, you may use two variables to set the location for the values. For example instead of setting something like:

        static_path: /home/foo/mywebsite

You can install your site as a non-zip safe python package and use:

        package_name: mywebsite
        static_path: __package_path__/static
        templates_path: __package_path__/templates
        translations_path: __package_path__/translations

Or you could specify a base_path:

        base_path: /home/foo/mywebsite
        static_path: __base_path__/static
        templates_path: __base_path__/templates
        translations_path: __base_path__/translations

If you are not going to install your app as a python package, you should set a base_path so that tinman knows what directory to insert into the Python path to be able to load your request handlers and such.

HTTP Server Options

Configure the tornado.httpserver.HTTPServer with the following options:

  • no_keep_alive: Enable/Disable keep-alives
  • ports: Ports to listen to, one process per port will be spawned
  • ssl_options: SSL Options to pass to the HTTP Server
    • certfile: Path to the certificate file
    • keyfile: Path to the keyfile
    • cert_reqs: Certificicate required?
    • ca_certs: One of none, optional or required
  • xheaders: Enable X-Header support in tornado.httpserver.HTTPServer

Logging Options

Logging uses the dictConfig format as specified at

Route List

The route list is specified using the top-level Routes keyword. Routes consist of a list of individual route items that may consist of mulitiple items in a route tuple.

Traditional Route Tuples

The traditional route tuple, as expected by Tornado is a two or three item tuple that includes the route to match on, the python module specified Class to use to respond to requests on that route and an optional route settings dictionary.

  - [/, myapp.Homepage]
  - [/images, tornado.web.StaticFileHandler, {'path': '/var/www/user_images'}]
Complex Regex Tuples for Tinman

In order to facilitate complex regex that does not break YAML files, Tinman supports a "re" flag in a Route tuple. If you find that your regex is breaking your route definition, insert the string "re" before the route regex.

    - re
    - /(c[a-f0-9]f[a-f0-9]{1,3}-[a-f0-9]{8}).gif
    - test.example.Pixel

Template Loader

The TemplateLoader configuration option is detailed the External Template Loading section of the document.

Example Configuration

The following is an example tinman application configuration:

%YAML 1.2
user: www-data
group: www-data
pidfile: /var/run/tinman/

    base_path: /home/foo/mywebsite
    debug: True
    xsrf_cookies: False
    # Any other vaidate Tornado application setting item

    no_keep_alive: False
    ports: [8000,8001]
    xheaders: True

    tinman: tinman
    pika: pika
    myapp: myapp.handlers
      format: "%(levelname) -10s %(asctime)s %(funcName) -25s: %(message)s"
      format: "%(levelname)s <PID %(process)d:%(processName)s> %(message)s"
      class: logging.StreamHandler
      formatter: verbose
      level: DEBUG
      debug_only: True
        - myapp
        - tinman
      class: logging.handlers.SysLogHandler
      facility: local6
      address: /dev/log
      formatter: syslog
      level: INFO
        - myapp
        - tinman
        - pika
      pika: INFO

# Automatically connect to PostgreSQL
    host: localhost
    port: 5432
    dbname: postgres
    user: postgres

# Automatically connect to RabbitMQ
    host: localhost
    port: 5672
    username: guest
    password: guest
    virtual_host: /

# Automatically connect to Redis
    host: localhost
    port: 6379
    db: 0

     -[/, test.example.Home]
        # /c1f1-7c5d9e0f.gif
        - re
        - /(c[a-f0-9]f[a-f0-9]{1,3}-[a-f0-9]{8}).gif
        - test.example.Pixel
        - .*
        - tornado.web.RedirectHandler
        - {"url": ""}

Test Application

The tinman application runner has a built in test application. To see if the module is setup correctly simply run:

tinman -f

In your console you should see output similar to:

Configuration not specified, running Tinman Test Application
utils       # 247   INFO      2011-06-11 23:25:26,164  Log level set to 10
cli         # 145   INFO      2011-06-11 23:25:26,164  Starting Tinman v0.2.1 process for port 8000
cli         # 154   DEBUG     2011-06-11 23:25:26,169  All children spawned
application # 106   DEBUG     2011-06-11 23:25:26,170  Initializing route: / with tinman.test.DefaultHandler
application # 36    INFO      2011-06-11 23:25:26,171  Appending handler: ('/', <class 'tinman.test.DefaultHandler'>)
cli         # 171   INFO      2011-06-11 23:25:26,174  Starting Tornado v1.2.1 HTTPServer on port 8000
web         # 1235  INFO      2011-06-11 23:25:32,782  200 GET / ( 1.24ms

You should now be able to access a test webpage on port 8000. CTRL-C will exit.


Tinman has decorators to make web development easier.


Vaidates the requesting IP address against a list of ip address blocks specified in Application.settings


# Define the whitelist as part of your application settings
settings['whitelist'] = ['',

application = Application(routes, **settings)

# Specify the decorator in each method of your RequestHandler class
# where you'd like to enforce the whitelist
class MyClass(tornado.web.RequestHandler):

  def get(self):
      self.write("IP was whitelisted")

In addition you may add the whitelist right into the configuration file:



A local in-memory cache decorator. RequestHandler class method calls are cached by name and arguments. Note that this monkey-patches the RequestHandler class on execution and will cache the total output created including all of the template rendering if there is anything. Local cache can be flushed with tinman.cache.flush()


class MyClass(tornado.web.RequestHandler):

   def get(self, content_id):
       self.write("Hello, World")


Send a logging.debug message with the class, method and arguments called.


class MyClass(tornado.web.RequestHandler):

    def get(self, content_id):
       self.write("Hello, World")



A collection of helpful functions for starting, daemonizing and stopping a Tornado application.

daemonize(pidfile=None, user=None, group=None)

Daemonize the application specifying the PID in the pidfile if specified, setting the application to run as the user and group if specified.

setup_logging(config, debug=False)

Setup the logging module with the parameters specified in the config dictionary. If debug is specified as True, output will be to stdout using the Tornado colored output if available and all other logging methods such as file or syslog will be disabled.

config dictionary format
directory:   Optional log file output directory
filename:    Optional filename, not needed for syslog
format:      Logging output format
level:       One of debug, error, warning, info
handler:     Optional handler
syslog:      If handler == syslog, parameters for syslog
    address:   Syslog address
    facility:  Syslog facility


Will call the stop() method of all child objects added to the tinman.utils.children list. This is useful for multi-processing apps to make sure that all children shutdown when a signal is called on the parent


Registers the shutdown function on SIGTERM and registers a rehash handler on SIGHUP. To specify the rehash handler, assign a callback to tinman.utils.rehash_handler.

Auto-Setup of Services

In order to facilitate a quick development process, the tinman application now has the concept of auto-setup and connect services. Initially, RabbitMQ is the only connectivity that is supported (via the Pika library). It is intended to add support for all major service types that have asynchronous support for the Tornado IO loop.


To setup an automatic connection to PostgreSQL simply include a Postgres section in your configuration file:

     host: localhost
     port: 5432
     dbname: postgres
     user: postgres

NOTE Currently this support is not asynchronous

When the application is constructed, it will connect to PostgreSQL and assign the connection to a standard object called tinman which is an attribute of the application. The connection will automatically be setup to use the autocommit isolation level.

We construct a copy of a tinman specific object using the tinman.clients.pgsql.PgSQL class. This class attempts to cache connections to PostgreSQL at the module level, reducing the amount of connection/disconnects needed. This object is accessed from a request handler as:


To get a cursor, simply use the cursor attribute of the PgSQL object. This will return a psycopg2.extras.DictCursor object.


class MyRequestHandler(web.RequestHandler):
    def get(self, *args, **kwargs):
        cursor = self.application.tinman.pgsql.cursor
        cursor.execute('SELECT * FROM my_table;')
        data = cursor.fetchall()
        self.render('my-template.html', data=data)


To setup an automatic connection to RabbitMQ simply include a RabbitMQ section in your configuration file:

    host: localhost
    port: 5672
    username: guest
    password: guest
    virtual_host: /

When the application is constructed, it will connect to RabbitMQ and assign the connection and channel to a standard object called tinman which is an attribute of the application.

We construct a copy of a tinman specific object using the tinman.clients.rabbitmq.RabbitMQ class. Currently this is only setup to publish messages, though it is the intent to add the ability to consume messages asychronously as well. This object is accessed from a request handler as: self.application.tinman.rabbitmq

For publishing messages, only one command is required: RabbitMQ.publish_message

If you pass in a dictionary or list, the message will be auto-JSON encoded and the mimetype will be set as application/json.


  • exchange: RabbitMQ exchange to publish to
  • routing_key: RabbitMQ routing key to use in publishing message
  • message: The message itself to send
  • mimetype: The mimetype of the message (default: text/plain)
  • mandatory: AMQP Basic.Publish mandatory field
  • immediate: AMQP Basic.Publish immediate field






To setup an automatic connection to Redis siply include a Redis section in your configuration file:

    host: localhost
    port: 6379
    db: 0
    password: foo

When the application is constructed, it will connect to Redis and assign the connection and channel to a standard object called tinman which is an attribute of the application.

We construct a copy of a tinman specific object using the tinman.clients.redis.Redis class. This object is accessed from a request handler as self.application.tinman.redis

This class requires the asynchronous brukva client from


from tornado import web

class RedisTest(web.RequestHandler):

    def initialize(self):
        # Set a more handy redis handle
        self.redis = self.application.tinman.redis.client

    def get(self, key):

        # Make an asynchronous redis request
        self.redis.get(key, self.on_redis_response)

    def on_redis_response(self, response):
        """Since we're fully async here, when redis comes back with our response
        we'll process it in this function.

        # If we could not find the response code
        if not response:

        # Write the redis value out as a key/value pair in JSON
        self.write({key: response})

        # Done

CouchDB Loader

Tinman includes tinman.loader.CouchDBLoader to enable the storage of templates in CouchDB to provide a common stemplate storage mechanism across multiple servers.

Templates stored in CouchDB are stored with the full path as the document key and the template value is stored in the document using the key "template"

When storing templates in CouchDB it is important that the forward-slash (/) is replaced with a colon (:) as the key value in CouchDB, or it will not find the stored file.

For example a template with a filesystem path of /foo/bar/template.html would be stored with a key of foo:bar:template.html in CouchDB but still referenced as /foo/bar/template.html everywhere else.

Example templates document from CouchDB

   "_id": "base.html",
   "_rev": "1-18d104181a15f617a929c221d01423da",
   "template": "<html>\n  <head>\n    <title>{% block \"title\" %}Title{% end %}</title>\n  </head>\n  <body>\n    <h1>Hello, World!</h1>\n  </body>\n</html>"
   "_id": "pages:home.html",
   "_rev": "2-c3c06f5a6d6a7b8149e0a700c67aeb41",
   "template": "{%  extends \"base.html\" %} \n{% block title %}Homepage{% end %}"
Something went wrong with that request. Please try again.