As we saw in firstapp_chapter
, it's possible to create a Pyramid
application completely manually. However, it's usually more convenient to use a template to generate a basic Pyramid
project
.
A project is a directory that contains at least one package
. You'll use a template to create a project, and you'll create your application logic within a package that lives inside the project. Even if your application is extremely simple, it is useful to place code that drives the application within a package, because a package is more easily extended with new code. An application that lives inside a package can also be distributed more easily than one which does not live within a package.
Pyramid
comes with a variety of templates that you can use to generate a project. Each template makes different configuration assumptions about what type of application you're trying to construct.
These templates are rendered using the PasteDeploy
paster
script, and so therefore they are often referred to as "paster templates".
single: paster templates single: pyramid_starter paster template single: pyramid_zodb paster template single: pyramid_alchemy paster template single: pyramid_routesalchemy paster template
The convenience paster
templates included with Pyramid
differ from each other on a number of axes:
- the persistence mechanism they offer (no persistence mechanism,
ZODB
, orSQLAlchemy
). - the mechanism they use to map URLs to code (
traversal
orURL dispatch
). - whether or not the
pyramid_beaker
library is relied upon as the sessioning implementation (as opposed to no sessioning or default sessioning).
The included templates are these:
pyramid_starter
URL mapping via
traversal
and no persistence mechanism.pyramid_zodb
URL mapping via
traversal
and persistence viaZODB
.pyramid_routesalchemy
URL mapping via
URL dispatch
and persistence viaSQLAlchemy
pyramid_alchemy
URL mapping via
traversal
and persistence viaSQLAlchemy
single: creating a project single: project
In installing_chapter
, you created a virtual Python environment via the virtualenv
command. To start a Pyramid
project
, use the paster
facility installed within the virtualenv. In installing_chapter
we called the virtualenv directory env
; the following command assumes that our current working directory is that directory.
We'll choose the pyramid_starter
template for this purpose.
$ bin/paster create -t pyramid_starter
The above command uses the paster
command to create a project using the pyramid_starter
template. The paster create
command creates project from a template. To use a different template, such as pyramid_routesalchemy
, you'd just change the last argument. For example:
$ bin/paster create -t pyramid_routesalchemy
paster create
will ask you a single question: the name of the project. You should use a string without spaces and with only letters in it. Here's sample output from a run of paster create
for a project we name MyProject
:
$ bin/paster create -t pyramid_starter
Selected and implied templates:
pyramid#pyramid_starter pyramid starter project
Enter project name: MyProject
Variables:
egg: MyProject
package: myproject
project: MyProject
Creating template pyramid
Creating directory ./MyProject
# ... more output ...
Running /Users/chrism/projects/pyramid/bin/python setup.py egg_info
Note
You can skip the interrogative question about a project name during paster create
by adding the project name to the command line, e.g. paster create -t pyramid_starter MyProject
.
Note
You may encounter an error when using paster create
if a dependent Python package is not installed. This will result in a traceback ending in:
pkg_resources.DistributionNotFound: <package name>
Simply run bin/easy_install
, with the missing package name from the error message, to work around this issue.
As a result of invoking the paster create
command, a project is created in a directory named MyProject
. That directory is a project
directory. The setup.py
file in that directory can be used to distribute your application, or install your application for deployment or development.
A PasteDeploy
.ini
file named development.ini
will also be created in the project directory. You will use this .ini
file to configure a server, to run your application, and to and debug your application.
The MyProject
project directory contains an additional subdirectory named myproject
(note the case difference) representing a Python package
which holds very simple Pyramid
sample code. This is where you'll edit your application's Python code and templates.
single: setup.py develop single: development install
To install a newly created project for development, you should cd
to the newly created project directory and use the Python interpreter from the virtualenv
you created during installing_chapter
to invoke the command python setup.py develop
The file named setup.py
will be in the root of the paster-generated project directory. The python
you're invoking should be the one that lives in the bin
directory of your virtual Python environment. Your terminal's current working directory must the the newly created project directory. For example:
$ ../bin/python setup.py develop
Elided output from a run of this command is shown below:
$ ../bin/python setup.py develop
...
Finished processing dependencies for MyProject==0.0
This will install a distribution
representing your project into the interpreter's library set so it can be found by import
statements and by PasteDeploy
commands such as paster serve
and paster pshell
.
single: running tests single: tests (running)
To run unit tests for your application, you should invoke them using the Python interpreter from the virtualenv
you created during installing_chapter
(the python
command that lives in the bin
directory of your virtualenv):
$ ../bin/python setup.py test -q
Here's sample output from a test run:
$ python setup.py test -q
running test
running egg_info
writing requirements to MyProject.egg-info/requires.txt
writing MyProject.egg-info/PKG-INFO
writing top-level names to MyProject.egg-info/top_level.txt
writing dependency_links to MyProject.egg-info/dependency_links.txt
writing entry points to MyProject.egg-info/entry_points.txt
reading manifest file 'MyProject.egg-info/SOURCES.txt'
writing manifest file 'MyProject.egg-info/SOURCES.txt'
running build_ext
..
----------------------------------------------------------------------
Ran 1 test in 0.108s
OK
Note
The -q
option is passed to the setup.py test
command to limit the output to a stream of dots. If you don't pass -q
, you'll see more verbose test result output (which normally isn't very useful).
The tests themselves are found in the tests.py
module in your paster create
-generated project. Within a project generated by the pyramid_starter
template, a single sample test exists.
single: interactive shell single: IPython single: paster pshell
Once you've installed your program for development using setup.py develop
, you can use an interactive Python shell to examine your Pyramid
project's resource
and view
objects from a Python prompt. To do so, use your virtualenv's paster pshell
command.
The first argument to pshell
is the path to your application's .ini
file. The second is the app
section name inside the .ini
file which points to your application as opposed to any other section within the .ini
file. For example, if your application .ini
file might have a [app:MyProject]
section that looks like so:
[app:MyProject]
use = egg:MyProject
reload_templates = true
debug_authorization = false
debug_notfound = false
debug_templates = true
default_locale_name = en
If so, you can use the following command to invoke a debug shell using the name MyProject
as a section name:
[chrism@vitaminf shellenv]$ ../bin/paster pshell development.ini MyProject
Python 2.4.5 (#1, Aug 29 2008, 12:27:37)
[GCC 4.0.1 (Apple Inc. build 5465)] on darwin
Type "help" for more information. "root" is the Pyramid app root object,
"registry" is the Pyramid registry object.
>>> root
<myproject.resources.MyResource object at 0x445270>
>>> registry
<Registry myproject>
>>> registry.settings['debug_notfound']
False
>>> from myproject.views import my_view
>>> from pyramid.request import Request
>>> r = Request.blank('/')
>>> my_view(r)
{'project': 'myproject'}
Two names are made available to the pshell user as globals: root
and registry
. root
is the the object returned by the default root
factory
in your application. registry
is the application
registry
object associated with your project's application (often accessed within view code as request.registry
).
If you have IPython installed in the interpreter you use to invoke the paster
command, the pshell
command will use an IPython interactive shell instead of a standard Python interpreter shell. If you don't want this to happen, even if you have IPython installed, you can pass the --disable-ipython
flag to the pshell
command to use a standard Python interpreter shell unconditionally.
[chrism@vitaminf shellenv]$ ../bin/paster pshell --disable-ipython \
development.ini MyProject
Warning
You should always use a section name argument that refers to the actual app
section within the Paste configuration file that points at your Pyramid
application without any middleware wrapping. In particular, a section name is inappropriate as the second argument to pshell
if the configuration section it names is a pipeline
rather than an app
. For example, if you have the following .ini
file content:
[app:MyProject]
use = egg:MyProject
reload_templates = true
debug_authorization = false
debug_notfound = false
debug_templates = true
default_locale_name = en
[pipeline:main]
pipeline =
egg:WebError#evalerror
MyProject
Use MyProject
instead of main
as the section name argument to pshell
against the above .ini
file (e.g. paster pshell development.ini MyProject
). If you use main
instead, an error will occur. Use the most specific reference to your application within the .ini
file possible as the section name argument.
Press Ctrl-D
to exit the interactive shell (or Ctrl-Z
on Windows).
single: running an application single: paster serve single: reload single: startup single: mod_wsgi
Once a project is installed for development, you can run the application it represents using the paster serve
command against the generated configuration file. In our case, this file is named development.ini
:
$ ../bin/paster serve development.ini
Here's sample output from a run of paster serve
:
$ ../bin/paster serve development.ini
Starting server in PID 16601.
serving on 0.0.0.0:6543 view at http://127.0.0.1:6543
By default, Pyramid
applications generated from a paster
template will listen on TCP port 6543. You can shut down a server started this way by pressing Ctrl-C
.
During development, it's often useful to run paster serve
using its --reload
option. When --reload
is passed to paster serve
, changes to any Python module your project uses will cause the server to restart. This typically makes development easier, as changes to Python code made within a Pyramid
application is not put into effect until the server restarts.
For example:
$ ../bin/paster serve development.ini --reload
Starting subprocess with file monitor
Starting server in PID 16601.
serving on 0.0.0.0:6543 view at http://127.0.0.1:6543
For more detailed information about the startup process, see startup_chapter
. For more information about environment variables and configuration file settings that influence startup and runtime behavior, see environment_chapter
.
Once your application is running via paster serve
, you may visit http://localhost:6543/
in your browser. You will see something in your browser like what is displayed in the following image:
This is the page shown by default when you visit an unmodified paster create
-generated pyramid_starter
application in a browser.
Using an Alternate WSGI Server
The code generated by Pyramid
paster
templates assumes that you will be using the paster serve
command to start your application while you do development. However, paster serve
is by no means the only way to start up and serve a Pyramid
application. As we saw in firstapp_chapter
, paster serve
needn't be invoked at all to run a Pyramid
application. The use of paster serve
to run a Pyramid
application is purely conventional based on the output of its paster
templates.
Any WSGI
server is capable of running a Pyramid
application. Some WSGI servers don't require the PasteDeploy
framework's paster serve
command to do server process management at all. Each WSGI
server has its own documentation about how it creates a process to run an application, and there are many of them, so we cannot provide the details for each here. But the concepts are largely the same, whatever server you happen to use.
One popular production alternative to a paster
-invoked server is mod_wsgi
. You can also use mod_wsgi
to serve your Pyramid
application using the Apache web server rather than any "pure-Python" server that is started as a result of paster serve
. See modwsgi_tutorial
for details. However, it is usually easier to develop an application using a paster serve
-invoked webserver, as exception and debugging output will be sent to the console.
single: project structure
The pyramid_starter
template generated a project
(named MyProject
), which contains a Python package
. The package is also named myproject
, but it's lowercased; the paster template generates a project which contains a package that shares its name except for case.
All Pyramid
paster
-generated projects share a similar structure. The MyProject
project we've generated has the following directory structure:
MyProject/
|-- CHANGES.txt
|-- development.ini
|-- myproject
| |-- __init__.py
| |-- resources.py
| |-- static
| | |-- favicon.ico
| | |-- logo.png
| | `-- pylons.css
| |-- templates
| | `-- mytemplate.pt
| |-- tests.py
| `-- views.py
|-- README.txt
|-- setup.cfg
`-- setup.py
The MyProject
project
directory is the distribution and deployment wrapper for your application. It contains both the myproject
package
representing your application as well as files used to describe, run, and test your application.
CHANGES.txt
describes the changes you've made to the application. It is conventionally written inReStructuredText
format.README.txt
describes the application in general. It is conventionally written inReStructuredText
format.development.ini
is aPasteDeploy
configuration file that can be used to execute your application.setup.cfg
is asetuptools
configuration file used bysetup.py
.setup.py
is the file you'll use to test and distribute your application. It is a standardsetuptools
setup.py
file.
single: PasteDeploy single: ini file
The development.ini
file is a PasteDeploy
configuration file. Its purpose is to specify an application to run when you invoke paster serve
, as well as the deployment settings provided to that application.
The generated development.ini
file looks like so:
MyProject/development.ini
This file contains several "sections" including [app:MyProject]
, [pipeline:main]
, and [server:main]
.
The [app:MyProject]
section represents configuration for your application. This section name represents the MyProject
application (and it's an app
-lication, thus app:MyProject
)
The use
setting is required in the [app:MyProject]
section. The use
setting points at a setuptools
entry point
named MyProject
(the egg:
prefix in egg:MyProject
indicates that this is an entry point URI specifier, where the "scheme" is "egg"). egg:MyProject
is actually shorthand for a longer spelling: egg:MyProject#main
. The #main
part is omitted for brevity, as it is the default.
setuptools
Entry Points and PasteDeploy .ini
Files
This part of configuration can be confusing so let's try to clear things up a bit. Take a look at the generated setup.py
file for this project. Note that the entry_point
line in setup.py
points at a string which looks a lot like an .ini
file. This string representation of an .ini
file has a section named [paste.app_factory]
. Within this section, there is a key named main
(the entry point name) which has a value myproject:main
. The key main
is what our egg:MyProject#main
value of the use
section in our config file is pointing at (although it is actually shortened to egg:MyProject
there). The value represents a dotted Python name
path, which refers to a callable in our myproject
package's __init__.py
module. In English, this entry point can thus be referred to as a "Paste application factory in the MyProject
project which has the entry point named main
where the entry point refers to a main
function in the mypackage
module". If indeed if you open up the __init__.py
module generated within the myproject
package, you'll see a main
function. This is the function called by PasteDeploy
when the paster serve
command is invoked against our application. It accepts a global configuration object and returns an instance of our application.
The use
setting is the only setting required in the [app:MyProject]
section unless you've changed the callable referred to by the egg:MyProject
entry point to accept more arguments: other settings you add to this section are passed as keywords arguments to the callable represented by this entry point (main
in our __init__.py
module). You can provide startup-time configuration parameters to your application by adding more settings to this section.
The reload_templates
setting in the [app:MyProject]
section is a Pyramid
-specific setting which is passed into the framework. If it exists, and its value is true
, Chameleon
and Mako
template changes will not require an application restart to be detected. See reload_templates_section
for more information.
Warning
The reload_templates
option should be turned off for production applications, as template rendering is slowed when it is turned on.
The debug_templates
setting in the [app:MyProject]
section is a Pyramid
-specific setting which is passed into the framework. If it exists, and its value is true
, Chameleon
template exceptions will contained more detailed and helpful information about the error than when this value is false
. See debug_templates_section
for more information.
Warning
The debug_templates
option should be turned off for production applications, as template rendering is slowed when it is turned on.
Various other settings may exist in this section having to do with debugging or influencing runtime behavior of a Pyramid
application. See environment_chapter
for more information about these settings.
[pipeline:main]
, has the name main
signifying that this is the default 'application' (although it's actually a pipeline of middleware and an application) run by paster serve
when it is invoked against this configuration file. The name main
is a convention used by PasteDeploy signifying that it is the default application.
The [server:main]
section of the configuration file configures a WSGI server which listens on TCP port 6543. It is configured to listen on all interfaces (0.0.0.0
). The Paste#http
server will create a new thread for each request.
Note
In general, Pyramid
applications generated from paster templates should be threading-aware. It is not required that a Pyramid
application be nonblocking as all application code will run in its own thread, provided by the server you're using.
See the PasteDeploy
documentation for more information about other types of things you can put into this .ini
file, such as other applications, middleware
and alternate WSGI
server implementations.
Note
You can add a [DEFAULT]
section to your development.ini
file. Such a section should consists of global parameters that are shared by all the applications, servers and middleware
defined within the configuration file. The values in a [DEFAULT]
section will be passed to your application's main
function as global_values
.
single: setup.py
The setup.py
file is a setuptools
setup file. It is meant to be run directly from the command line to perform a variety of functions, such as testing your application, packaging, and distributing your application.
Note
setup.py
is the defacto standard which Python developers use to distribute their reusable code. You can read more about setup.py
files and their usage in the Setuptools documentation.
Our generated setup.py
looks like this:
MyProject/setup.py
The setup.py
file calls the setuptools setup
function, which does various things depending on the arguments passed to setup.py
on the command line.
Within the arguments to this function call, information about your application is kept. While it's beyond the scope of this documentation to explain everything about setuptools setup files, we'll provide a whirlwind tour of what exists in this file in this section.
Your application's name can be any string; it is specified in the name
field. The version number is specified in the version
value. A short description is provided in the description
field. The long_description
is conventionally the content of the README and CHANGES file appended together. The classifiers
field is a list of Trove classifiers describing your application. author
and author_email
are text fields which probably don't need any description. url
is a field that should point at your application project's URL (if any). packages=find_packages()
causes all packages within the project to be found when packaging the application. include_package_data
will include non-Python files when the application is packaged if those files are checked into version control. zip_safe
indicates that this package is not safe to use as a zipped egg; instead it will always unpack as a directory, which is more convenient. install_requires
and tests_require
indicate that this package depends on the pyramid
package. test_suite
points at the package for our application, which means all tests found in the package will be run when setup.py test
is invoked. We examined entry_points
in our discussion of the development.ini
file; this file defines the main
entry point that represents our project's application.
Usually you only need to think about the contents of the setup.py
file when distributing your application to other people, or when versioning your application for your own use. For fun, you can try this command now:
$ python setup.py sdist
This will create a tarball of your application in a dist
subdirectory named MyProject-0.1.tar.gz
. You can send this tarball to other people who want to use your application.
Warning
By default, setup.py sdist
does not place non-Python-source files in generated tarballs. This means, in this case, that the templates/mytemplate.pt
file and the files in the static
directory are not packaged in the tarball. To allow this to happen, check all the files that you'd like to be distributed along with your application's Python files into Subversion. After you do this, when you rerun setup.py sdist
, all files checked into the version control system will be included in the tarball. If you don't use Subversion, and instead use a different version control system, you may need to install a setuptools add-on such as setuptools-git
or setuptools-hg
for this behavior to work properly. Alternatively, you can specify the non-Python-source files by hand in a manifest template
, called MANIFEST.in
by default.
The setup.cfg
file is a setuptools
configuration file. It contains various settings related to testing and internationalization:
Our generated setup.cfg
looks like this:
MyProject/setup.cfg
The values in the default setup file allow various commonly-used internationalization commands and testing commands to work more smoothly.
single: package
The myproject
package
lives inside the MyProject
project
. It contains:
- An
__init__.py
file signifies that this is a Pythonpackage
. It also contains code that helps users run the application, including amain
function which is used as a Paste entry point. - A
resources.py
module, which containsresource
code. - A
templates
directory, which containsChameleon
(or other types of) templates. - A
tests.py
module, which contains unit test code for the application. - A
views.py
module, which contains view code for the application.
These are purely conventions established by the paster
template: Pyramid
doesn't insist that you name things in any particular way. However, it's generally a good idea to follow Pyramid standards for naming, so that other Pyramid developers can get up to speed quickly on your code when you need help.
single: __init.py
We need a small Python module that configures our application and which advertises an entry point for use by our PasteDeploy
.ini
file. This is the file named __init__.py
. The presence of an __init__.py
also informs Python that the directory which contains it is a package.
MyProject/myproject/__init__.py
- Line 1 imports the
Configurator
class frompyramid.config
that we use later. - Line 2 imports the
Root
class frommyproject.resources
that we use later. Lines 4-12 define a function that returns a
Pyramid
WSGI application. This function is meant to be called by thePasteDeploy
framework as a result of runningpaster serve
.Within this function, application configuration is performed.
Lines 8-10 register a "default view" (a view that has no
name
attribute). It is registered so that it will be found when thecontext
of the request is an instance of themyproject.resources.Root
class. The first argument toadd_view
points at a Python function that does all the work for this view, also known as aview callable
, via adotted Python name
. The view declaration also names arenderer
, which in this case is a template that will be used to render the result of the view callable. This particular view declaration points atmyproject:templates/mytemplate.pt
, which is aasset specification
that specifies themytemplate.pt
file within thetemplates
directory of themyproject
package. The template file it actually points to is aChameleon
ZPT template file.Line 11 registers a static view, which will serve up the files from the
mypackage:static
asset specification
(thestatic
directory of themypackage
package).Line 12 returns a
WSGI
application to the caller of the function (Paste).
Much of the heavy lifting in a Pyramid
application is done by view callables. A view callable
is the main tool of a Pyramid
web application developer; it is a bit of code which accepts a request
and which returns a response
.
MyProject/myproject/views.py
This bit of code was registered as the view callable within __init__.py
(via add_view
). add_view
said that the default URL for instances that are of the class myproject.resources.Root
should run this myproject.views.my_view
function.
This view callable function is handed a single piece of information: the request
. The request is an instance of the WebOb
Request
class representing the browser's request to our server.
This view returns a dictionary. When this view is invoked, a renderer
converts the dictionary returned by the view into HTML, and returns the result as the response
. This view is configured to invoke a renderer which uses a Chameleon
ZPT template (mypackage:templates/my_template.pt
, as specified in the __init__.py
file call to add_view
).
See views_which_use_a_renderer
for more information about how views, renderers, and templates relate and cooperate.
Note
Because our development.ini
has a reload_templates = true
directive indicating that templates should be reloaded when they change, you won't need to restart the application server to see changes you make to templates. During development, this is handy. If this directive had been false
(or if the directive did not exist), you would need to restart the application server for each template change. For production applications, you should set your project's reload_templates
to false
to increase the speed at which templates may be rendered.
single: resources.py
The resources.py
module provides the resource
data and behavior for our application. Resources are objects which exist to provide site structure in applications which use traversal
to map URLs to code. We write a class named Root
that provides the behavior for the root resource.
MyProject/myproject/resources.py
- Lines 1-3 define the Root class. The Root class is a "root resource factory" function that will be called by the
Pyramid
Router for each request when it wants to find the root of the resource tree.
In a "real" application, the Root object would likely not be such a simple object. Instead, it might be an object that could access some persistent data store, such as a database. Pyramid
doesn't make any assumption about which sort of data storage you'll want to use, so the sample application uses an instance of myproject.resources.Root
to represent the root.
This directory contains static assets which support the mytemplate.pt
template. It includes CSS and images.
The single Chameleon
template exists in the project. Its contents are too long to show here, but it displays a default page when rendered. It is referenced by the call to add_view
as the renderer
attribute in the __init__
file. See views_which_use_a_renderer
for more information about renderers.
Templates are accessed and used by view configurations and sometimes by view functions themselves. See templates_used_directly
and templates_used_as_renderers
.
single: tests.py
The tests.py
module includes unit tests for your application.
MyProject/myproject/tests.py
This sample tests.py
file has a single unit test defined within it. This test is executed when you run python setup.py test
. You may add more tests here as you build your application. You are not required to write tests to use Pyramid
, this file is simply provided as convenience and example.
See testing_chapter
for more information about writing Pyramid
unit tests.
It is best practice for your application's code layout to not stray too much from accepted Pyramid paster template defaults. If you refrain from changing things very much, other Pyramid coders will be able to more quickly understand your application. However, the code layout choices made for you by a paster template are in no way magical or required. Despite the choices made for you by any paster template, you can decide to lay your code out any way you see fit.
For example, the configuration method named ~pyramid.config.Configurator.add_view
requires you to pass a dotted Python name
or a direct object reference as the class or function to be used as a view. By default, the pyramid_starter
paster template would have you add view functions to the views.py
module in your package. However, you might be more comfortable creating a views
directory, and adding a single file for each view.
If your project package name was myproject
and you wanted to arrange all your views in a Python subpackage within the myproject
package
named views
instead of within a single views.py
file, you might:
- Create a
views
directory inside yourmypackage
package directory (the same directory which holdsviews.py
). - Move the existing
views.py
file to a file inside the newviews
directory named, say,blog.py
. - Create a file within the new
views
directory named__init__.py
(it can be empty, this just tells Python that theviews
directory is a package.
Then change the __init__.py of your myproject project (not the __init__.py
you just created in the views
directory, the one in its parent directory). For example, from something like:
config.add_view('myproject.views.my_view',
renderer='myproject:templates/mytemplate.pt')
To this:
config.add_view('myproject.views.blog.my_view',
renderer='myproject:templates/mytemplate.pt')
You can then continue to add files to the views
directory, and refer to view classes or functions within those files via the dotted name passed as the first argument to add_view
. For example, if you added a file named anothermodule.py
to the views
subdirectory, and added a view callable named my_view
to it:
config.add_view('myproject.views.anothermodule.my_view',
renderer='myproject:templates/anothertemplate.pt')
This pattern can be used to rearrage code referred to by any Pyramid API argument which accepts a dotted Python name
or direct object reference.