What's New in Pyramid 1.3

This article explains the new features in Pyramid version 1.3 as compared to its predecessor, Pyramid 1.2. It also documents backwards incompatibilities between the two versions and deprecations added to Pyramid 1.3, as well as software dependency changes and notable documentation additions.

Major Feature Additions

The major feature additions in Pyramid 1.3 follow.

Python 3 Compatibility

../_images/python-3.png

Pyramid continues to run on Python 2, but Pyramid is now also Python 3 compatible. To use Pyramid under Python 3, Python 3.3 or better is required.

Many Pyramid add-ons are already Python 3 compatible. For example, pyramid_debugtoolbar, pyramid_jinja2, pyramid_exclog, pyramid_tm, pyramid_mailer, and pyramid_handlers are all Python 3-ready. But other add-ons are known to work only under Python 2. Also, some scaffolding dependencies (particularly ZODB) do not yet work under Python 3.

Please be patient as we gain full ecosystem support for Python 3. You can see more details about ongoing porting efforts at https://github.com/Pylons/pyramid/wiki/Python-3-Porting .

Python 3 compatibility required dropping some package dependencies and support for older Python versions and platforms. See the "Backwards Incompatibilities" section below for more information.

The paster Command Has Been Replaced

We've replaced the paster command with Pyramid-specific analogues. Why? The libraries that supported the paster command named Paste and PasteScript do not run under Python 3, and we were unwilling to port and maintain them ourselves. As a result, we've had to make some changes.

Previously (in Pyramid 1.0, 1.1 and 1.2), you created a Pyramid application using paster create, like so:

$ $VENV/bin/paster create -t pyramid_starter foo

In 1.3, you're now instead required to create an application using pcreate like so:

$ $VENV/bin/pcreate -s starter foo

pcreate is required to be used for internal Pyramid scaffolding; externally distributed scaffolding may allow for both pcreate and/or paster create.

In previous Pyramid versions, you ran a Pyramid application like so:

$ $VENV/bin/paster serve development.ini

Instead, you now must use the pserve command in 1.3:

$ $VENV/bin/pserve development.ini

The ini configuration file format supported by Pyramid has not changed. As a result, Python 2-only users can install PasteScript manually and use paster serve instead if they like. However, using pserve will work under both Python 2 and Python 3.

Analogues of paster pshell, paster pviews, paster request and paster ptweens also exist under the respective console script names pshell, pviews, prequest and ptweens.

paste.httpserver replaced by waitress in Scaffolds

Because the paste.httpserver server we used previously in scaffolds is not Python 3 compatible, we've made the default WSGI server used by Pyramid scaffolding the waitress server. The waitress server is both Python 2 and Python 3 compatible.

Once you create a project from a scaffold, its development.ini and production.ini will have the following line:

use = egg:waitress#main

Instead of this (which was the default in older versions):

use = egg:Paste#http

Note

paste.httpserver "helped" by converting header values that were Unicode into strings, which was a feature that subverted the WSGI specification. The waitress server, on the other hand implements the WSGI spec more fully. This specifically may affect you if you are modifying headers on your responses. The following error might be an indicator of this problem: AssertionError: Header values must be strings, please check the type of the header being returned. A common case would be returning Unicode headers instead of string headers.

Compatibility Helper Library

A new pyramid.compat module was added which provides Python 2/3 straddling support for Pyramid add-ons and development environments.

Introspection

A configuration introspection system was added; see Pyramid Configuration Introspection and Adding Configuration Introspection for more information on using the introspection system as a developer.

The latest release of the pyramid debug toolbar (0.9.7+) provides an "Introspection" panel that exposes introspection information to a Pyramid application developer.

New APIs were added to support introspection pyramid.registry.Introspectable, pyramid.config.Configurator.introspector, pyramid.config.Configurator.introspectable, pyramid.registry.Registry.introspector.

@view_defaults Decorator

If you use a class as a view, you can use the new pyramid.view.view_defaults class decorator on the class to provide defaults to the view configuration information used by every @view_config decorator that decorates a method of that class.

For instance, if you've got a class that has methods that represent "REST actions", all which are mapped to the same route, but different request methods, instead of this:

 1from pyramid.view import view_config
 2from pyramid.response import Response
 3
 4class RESTView(object):
 5    def __init__(self, request):
 6        self.request = request
 7
 8    @view_config(route_name='rest', request_method='GET')
 9    def get(self):
10        return Response('get')
11
12    @view_config(route_name='rest', request_method='POST')
13    def post(self):
14        return Response('post')
15
16    @view_config(route_name='rest', request_method='DELETE')
17    def delete(self):
18        return Response('delete')

You can do this:

 1from pyramid.view import view_defaults
 2from pyramid.view import view_config
 3from pyramid.response import Response
 4
 5@view_defaults(route_name='rest')
 6class RESTView(object):
 7    def __init__(self, request):
 8        self.request = request
 9
10    @view_config(request_method='GET')
11    def get(self):
12        return Response('get')
13
14    @view_config(request_method='POST')
15    def post(self):
16        return Response('post')
17
18    @view_config(request_method='DELETE')
19    def delete(self):
20        return Response('delete')

This also works for imperative view configurations that involve a class.

See @view_defaults Class Decorator for more information.

Extending a Request without Subclassing

It is now possible to extend a pyramid.request.Request object with property descriptors without having to create a custom request factory. The new method pyramid.config.Configurator.set_request_property() provides an entry point for addons to register properties which will be added to each request. New properties may be reified, effectively caching the return value for the lifetime of the instance. Common use-cases for this would be to get a database connection for the request or identify the current user. The new method pyramid.request.Request.set_property() has been added, as well, but the configurator method should be preferred as it provides conflict detection and consistency in the lifetime of the properties.

Not Found and Forbidden View Helpers

Not Found helpers:

Forbidden helpers:

Minor Feature Additions

  • New APIs: pyramid.path.AssetResolver and pyramid.path.DottedNameResolver. The former can be used to resolve an asset specification to an API that can be used to read the asset's data, the latter can be used to resolve a dotted Python name to a module or a package.

  • A mako.directories setting is no longer required to use Mako templates Rationale: Mako template renderers can be specified using an absolute asset spec. An entire application can be written with such asset specs, requiring no ordered lookup path.

  • bpython interpreter compatibility in pshell. See Alternative Shells for more information.

  • Added pyramid.paster.get_appsettings() API function. This function returns the settings defined within an [app:...] section in a PasteDeploy ini file.

  • Added pyramid.paster.setup_logging() API function. This function sets up Python logging according to the logging configuration in a PasteDeploy ini file.

  • Configuration conflict reporting is reported in a more understandable way ("Line 11 in file..." vs. a repr of a tuple of similar info).

  • We allow extra keyword arguments to be passed to the pyramid.config.Configurator.action() method.

  • Responses generated by Pyramid's pyramid.static.static_view now use a wsgi.file_wrapper (see https://www.python.org/dev/peps/pep-0333/#optional-platform-specific-file-handling) when one is provided by the web server.

  • The pyramid.config.Configurator.scan() method can be passed an ignore argument, which can be a string, a callable, or a list consisting of strings and/or callables. This feature allows submodules, subpackages, and global objects from being scanned. See https://docs.pylonsproject.org/projects/venusian/en/latest/#ignore-scan-argument for more information about how to use the ignore argument to scan.

  • Add pyramid.config.Configurator.add_traverser() API method. See Changing the Traverser for more information. This is not a new feature, it just provides an API for adding a traverser without needing to use the ZCA API.

  • Add pyramid.config.Configurator.add_resource_url_adapter() API method. See Changing How pyramid.request.Request.resource_url() Generates a URL for more information. This is not a new feature, it just provides an API for adding a resource url adapter without needing to use the ZCA API.

  • Better error messages when a view callable returns a value that cannot be converted to a response (for example, when a view callable returns a dictionary without a renderer defined, or doesn't return any value at all). The error message now contains information about the view callable itself as well as the result of calling it.

  • Better error message when a .pyc-only module is config.include -ed. This is not permitted due to error reporting requirements, and a better error message is shown when it is attempted. Previously it would fail with something like "AttributeError: 'NoneType' object has no attribute 'rfind'".

  • The system value req is now supplied to renderers as an alias for request. This means that you can now, for example, in a template, do req.route_url(...) instead of request.route_url(...). This is purely a change to reduce the amount of typing required to use request methods and attributes from within templates. The value request is still available too, this is just an alternative.

  • A new interface was added: pyramid.interfaces.IResourceURL. An adapter implementing its interface can be used to override resource URL generation when pyramid.request.Request.resource_url() is called. This interface replaces the now-deprecated pyramid.interfaces.IContextURL interface.

  • The dictionary passed to a resource's __resource_url__ method (see Overriding Resource URL Generation) now contains an app_url key, representing the application URL generated during pyramid.request.Request.resource_url(). It represents a potentially customized URL prefix, containing potentially custom scheme, host and port information passed by the user to request.resource_url. It should be used instead of request.application_url where necessary.

  • The pyramid.request.Request.resource_url() API now accepts these arguments: app_url, scheme, host, and port. The app_url argument can be used to replace the URL prefix wholesale during url generation. The scheme, host, and port arguments can be used to replace the respective default values of request.application_url partially.

  • A new API named pyramid.request.Request.resource_path() now exists. It works like pyramid.request.Request.resource_url() but produces a relative URL rather than an absolute one.

  • The pyramid.request.Request.route_url() API now accepts these arguments: _app_url, _scheme, _host, and _port. The _app_url argument can be used to replace the URL prefix wholesale during url generation. The _scheme, _host, and _port arguments can be used to replace the respective default values of request.application_url partially.

  • New APIs: pyramid.response.FileResponse and pyramid.response.FileIter, for usage in views that must serve files "manually".

Backwards Incompatibilities

  • Pyramid no longer runs on Python 2.5. This includes the most recent release of Jython and the Python 2.5 version of Google App Engine.

    The reason? We could not easily "straddle" Python 2 and 3 versions and support Python 2 versions older than Python 2.6. You will need Python 2.6 or better to run this version of Pyramid. If you need to use Python 2.5, you should use the most recent 1.2.X release of Pyramid.

  • The names of available scaffolds have changed and the flags supported by pcreate are different than those that were supported by paster create. For example, pyramid_alchemy is now just alchemy.

  • The paster command is no longer the documented way to create projects, start the server, or run debugging commands. To create projects from scaffolds, paster create is replaced by the pcreate console script. To serve up a project, paster serve is replaced by the pserve console script. New console scripts named pshell, pviews, proutes, and ptweens do what their paster <commandname> equivalents used to do. All relevant narrative documentation has been updated. Rationale: the Paste and PasteScript packages do not run under Python 3.

  • The default WSGI server run as the result of pserve from newly rendered scaffolding is now the waitress WSGI server instead of the paste.httpserver server. Rationale: the Paste and PasteScript packages do not run under Python 3.

  • The pshell command (see "paster pshell") no longer accepts a --disable-ipython command-line argument. Instead, it accepts a -p or --python-shell argument, which can be any of the values python, ipython or bpython.

  • Removed the pyramid.renderers.renderer_from_name function. It has been deprecated since Pyramid 1.0, and was never an API.

  • To use ZCML with versions of Pyramid >= 1.3, you will need pyramid_zcml version >= 0.8 and zope.configuration version >= 3.8.0. The pyramid_zcml package version 0.8 is backwards compatible all the way to Pyramid 1.0, so you won't be warned if you have older versions installed and upgrade Pyramid itself "in-place"; it may simply break instead (particularly if you use ZCML's includeOverrides directive).

  • String values passed to pyramid.request.Request.route_url() or pyramid.request.Request.route_path() that are meant to replace "remainder" matches will now be URL-quoted except for embedded slashes. For example:

    config.add_route('remain', '/foo*remainder')
    request.route_path('remain', remainder='abc / def')
    # -> '/foo/abc%20/%20def'
    

    Previously string values passed as remainder replacements were tacked on untouched, without any URL-quoting. But this doesn't really work logically if the value passed is Unicode (raw unicode cannot be placed in a URL or in a path) and it is inconsistent with the rest of the URL generation machinery if the value is a string (it won't be quoted unless by the caller).

    Some folks will have been relying on the older behavior to tack on query string elements and anchor portions of the URL; sorry, you'll need to change your code to use the _query and/or _anchor arguments to route_path or route_url to do this now.

  • If you pass a bytestring that contains non-ASCII characters to pyramid.config.Configurator.add_route() as a pattern, it will now fail at startup time. Use Unicode instead.

  • The path_info route and view predicates now match against request.upath_info (Unicode) rather than request.path_info (indeterminate value based on Python 3 vs. Python 2). This has to be done to normalize matching on Python 2 and Python 3.

  • The match_param view predicate no longer accepts a dict. This will have no negative affect because the implementation was broken for dict-based arguments.

  • The pyramid.interfaces.IContextURL interface has been deprecated. People have been instructed to use this to register a resource url adapter in the "Hooks" chapter to use to influence pyramid.request.Request.resource_url() URL generation for resources found via custom traversers since Pyramid 1.0.

    The interface still exists and registering an adapter using it as documented in older versions still works, but this interface will be removed from the software after a few major Pyramid releases. You should replace it with an equivalent pyramid.interfaces.IResourceURL adapter, registered using the new pyramid.config.Configurator.add_resource_url_adapter() API. A deprecation warning is now emitted when a pyramid.interfaces.IContextURL adapter is found when pyramid.request.Request.resource_url() is called.

  • Remove pyramid.config.Configurator.with_context class method. It was never an API, it is only used by pyramid_zcml and its functionality has been moved to that package's latest release. This means that you'll need to use the 0.9.2 or later release of pyramid_zcml with this release of Pyramid.

  • The older deprecated set_notfound_view Configurator method is now an alias for the new add_notfound_view Configurator method. Likewise, the older deprecated set_forbidden_view is now an alias for the new add_forbidden_view Configurator method. This has the following impact: the context sent to views with a (context, request) call signature registered via the set_notfound_view or set_forbidden_view will now be an exception object instead of the actual resource context found. Use request.context to get the actual resource context. It's also recommended to disuse set_notfound_view in favor of add_notfound_view, and disuse set_forbidden_view in favor of add_forbidden_view despite the aliasing.

Deprecations

  • The API documentation for pyramid.view.append_slash_notfound_view and pyramid.view.AppendSlashNotFoundViewFactory was removed. These names still exist and are still importable, but they are no longer APIs. Use pyramid.config.Configurator.add_notfound_view(append_slash=True) or pyramid.view.notfound_view_config(append_slash=True) to get the same behavior.

  • The set_forbidden_view and set_notfound_view methods of the Configurator were removed from the documentation. They have been deprecated since Pyramid 1.1.

  • All references to the tmpl_context request variable were removed from the docs. Its existence in Pyramid is confusing for people who were never Pylons users. It was added as a porting convenience for Pylons users in Pyramid 1.0, but it never caught on because the Pyramid rendering system is a lot different than Pylons' was, and alternate ways exist to do what it was designed to offer in Pylons. It will continue to exist "forever" but it will not be recommended or mentioned in the docs.

  • Remove references to do-nothing pyramid.debug_templates setting in all Pyramid-provided .ini files. This setting previously told Chameleon to render better exceptions; now Chameleon always renders nice exceptions regardless of the value of this setting.

Known Issues

  • As of this writing (the release of Pyramid 1.3b2), if you attempt to install a Pyramid project that used the alchemy scaffold via setup.py develop on Python 3.2, it will quit with an installation error while trying to install Pygments. If this happens, please just rerun the setup.py develop command again, and it will complete successfully. This is due to a minor bug in SQLAlchemy 0.7.5 under Python 3, and has been fixed in a later SQLAlchemy release. Keep an eye on https://github.com/sqlalchemy/sqlalchemy/issues/2421

Documentation Enhancements

Dependency Changes

  • Pyramid no longer depends on the zope.component package, except as a testing dependency.

  • Pyramid now depends on the following package versions: zope.interface>=3.8.0, WebOb>=1.2dev, repoze.lru>=0.4, zope.deprecation>=3.5.0, translationstring>=0.4 for Python 3 compatibility purposes. It also, as a testing dependency, depends on WebTest>=1.3.1 for the same reason.

  • Pyramid no longer depends on the Paste or PasteScript packages. These packages are not Python 3 compatible.

  • Depend on venusian >= 1.0a3 to provide scan ignore support.

Scaffolding Changes

  • Rendered scaffolds have now been changed to be more relocatable (fewer mentions of the package name within files in the package).

  • The routesalchemy scaffold has been renamed alchemy, replacing the older (traversal-based) alchemy scaffold (which has been retired).

  • The alchemy and starter scaffolds are Python 3 compatible.

  • The starter scaffold now uses URL dispatch by default.