Component framework (reahl-component)¶
The Reahl component framework extends setuptools distribution packages to package and distribute more than just code.
Pip and PyPI do a good job of enabling the distribution of Python code. The projects ensure that when you install a package you get all its correct dependencies as well.
When your code contains classes mapped by an ORM (such as SqlAlchemy) to tables in a database, things become more complicated:
The selection of packages used together in the same database cannot be foreseen by individual package authors.
How do you create a database schema sufficient for all the database-aware packages you have decided to use together in your project?
What happens if a new version of a package requires a different database schema to a previous version?
How do you migrate the schema of an existing database to the new schema? This, in the context of there being several packages mixed into a single database - with possible foreign key constraints forming dependencies between packages on the database level.
The Reahl component framework is an attempt to build such distributable packages that are database-aware. It solves all the surprisingly difficult accompanying problems. It calls such packages “components”.
Components are not only database-aware. Similar problems are solved for components that include:
its own configuration, which will be read from a separate file.
its own natural language translations to support multiple languages (i18n).
annotations of the data and features of its domain objects which can be used by, for example, a web framework, to manipulate such objects.
housekeeping code that needs to be invoked regularly.
In the text below, links are provided to more detail for some topics. These links refer to the Reahl web framework tutorial which discusses these topics in more detail — albeit in the context of the Reahl web framework. Note that reahl-component functionality is independent of the Reahl web framework.
Defining a component¶
A Reahl component is just a setuptools package with extra metadata. To make your project a component, add a component = key to the [options] section of your setup.cfg. The contents of this key is explained a bit more in the next section.
In order to help setuptools grok the component = option you also need to have a pyproject.toml file which lists setuptools, toml, and reahl-component-metadata as build dependencies.
Finally, your package itself needs to require reahl-component using install_requires as usual.
Basics of the component option¶
The component option expects to receive data in toml format. For this to work in a setup.cfg, the contents of the component option need to be after a dangling = and indented:
[options] component = # contents come here, but this comment can be omitted
Each time you change setup.cfg, be sure to regenerate the component metadata:
python -m pip install --no-deps -e .
Reahl command line¶
The Reahl command line is installed when you install reahl-component and is invoked with the command reahl. The set of commands it offers depends on other Reahl components you install.
To see the available commands, run:
Below is a list of which commands some components add:
Commands to work with components in development. This includes commands like extractmessages used in development for internationalisation (i18n).
Commands for working with the extra functionality provided by Reahl components in production. This includes managing databases, schemas and performing migrations as well as dealing with things like internationalisation and configuration.
Helpful commands, such as serve (which runs a development webserver) when doing web development using the Reahl web framework (reahl-web).
Commands for working with examples included in the overall Reahl documentation.
- Fetch Addresses from the database
How to register persisted classes with your component and use the command line to create a database schema.
The reahl-component infrastructure is extended by other Reahl components to be able to deal with differing implementations of ORM or database systems.
To use a particular database, include in your component’s dependencies the support package matching the database you want to use:
Set the reahlsystem.connection_uri in reahl.config.py to an URI matching your database as specified by SQLAlchemy.
List the database support component, reahl-sqlalchemysupport and reahl-component as dependencies of your component.
List each persisted class of your component in the setup.cfg's component option, as an element in the “persisted” key.
You can now use the following commands (amongst others) from reahl-commands to manage the database:
reahl createdbuser <config_directory> reahl createdb <config_directory> reahl createdbtables <config_directory>
- Database schema evolution
How to write migrations, define new versions of a Reahl component and upgrade a database to the new version.
The author of one component has no knowledge of other components which might inhabit the same database when used together. However, when component A depends on component B, the author of A will know that B is being used. The classes of A can be written such that they result in foreign keys to tables created by component B.
This creates a dependency on the database level with some implications:
When creating database tables, the tables of component B have to be created before those of A to ensure A’s foreign key constraints will not be violated.
When changing the schema for B, the foreign key constraints of A (to B) first have to be removed before changes are made to the schema of B. Then the foreign key constraints of A can be reinstated possibly referring to renamed tables or columns in B.
If B’s author brings out a new version of B in which tables or column names have changed as in (2) above, version 2 of
B will contain a
Migration which takes care of changing B’s schema from the old version.
If A’s author wants to bring out a new version of A that uses B v2, A’s author needs to write a
Migration as part of
A v2 which adjusts the old A v1 foreign key constraints to be compatible with the changes in B v2.
In a real world project, there could be a large number of such components by diverse authors. In order to migrate
the whole database from one version to another, Reahl computes a dependency graph that spans all the versions of all
the components in play. It then runs different parts of each
Migration in the correct order to ensure all database
level dependencies and constraints are honoured.
Each version of your Reahl component can have its own set of
Migrations which are performed when upgrading to that
version from its predecessor. The migration machinery needs access to all
Migrations of all versions of all
components to be able to compute a correct dependency tree.
For each such version entry, add two entries:
To prevent duplication, the version table matching the current version should not contain any requirements, since those are already in the standard install_requires of setup.cfg.
If the current version does not have any migrations, it need not be listed at all.
A change in dependency or in database schema is seen as at least a minor version change, therefore versions listed only specify major.minor version numbers, not an additional patch version.
Migration is written such that user code does not execute changes directly, but schedules each necessary change to be done
during a specified ‘phase’ of migration. The final order in which the
Migration itself and each individual phase of the
Migration will be executed
is determined by Reahl at runtime taking all components into account.
If you schedule more than one action in a single phase in your
Migration, these actions will retain their order
relative to one another.
The following useful commands from reahl-commands related to migration are available:
reahl migratedbdb <config_directory> reahl diffdb <config_directory> reahl listversionhistory <config_directory> reahl listdependencies <config_directory>
Some Reahl code is dependent upon there being an
ExecutionContext which can be obtained at any point in the code, and
hosts information such as the global configuration or current locale.
class Example: def do_something(self): self.do_something_else() def do_something_else(self): print(ExecutionContext.get_context().interface_locale) try: Example().do_something() # Breaks, because there's no ExecutionContext except: pass with ExecutionContext(): Example().do_something() # Prints en_gb by default
When building a system using Reahl components, your system includes other components via its dependencies. These dependencies can in turn depend on other Reahl components, and so on.
The author of each individual component knows whether their component needs configuration and what config it needs. Your final system, via dependencies, ends up consisting of a set of components by various authors. How does one configure the final system, and how does each author specify the configuration of their component without knowledge of this final composed system?
- Defining custom configuration
How to define and use configuration for your own component.
A Reahl system has a single config directory with a config file for each component in the system.
You specify a unique key for the config of your component, as well as what config settings you need and the configuration file name to be used for your component:
When defining config settings, you can specify default values for these settings, and also a human readable description of each setting. You can mark some config settings as “dangerous defaults”: such defaults will produce warnings when reading a configuration if they are not explicitly set in a config file. Defaults are usually chosen for a development environment, marking a default as dangerous is a way to prevent that default value from reaching a production environment.
Use an ExecutionContext
config = StoredConfiguration('/my/directory') config.configure() with ExecutionContext() as context: context.config = config ... ExecutionContext.get_context().config # To get it anywhere
In your system, read the config by creating a
StoredConfiguration, and then calling
configure() on it. Pass
True to strict_checking in production environments in order to turn dangerous defaults into errors instead of
Dependency injection between components¶
- dependency injection
How reahl-web-declarative supplies a specific implementation to reahl-web using dependency injection.
Sometimes it is useful to make one component (A) outsource some of its functionality to an as yet unknown component by requiring classes it can use as config settings. If component B depends on A, it can then automatically provide the relevant config settings to A.
This is a form of dependency injection which reahl-component uses to, for example, allow the use of different ORMs:
reahl-component has a config setting reahlsystem.orm_control which it uses to do various tasks related to database
management. If your project also depends on reahl-sqlalchemysupport, the latter will automatically configure
reahlsystem.orm_control to be a
The following useful commands from reahl-commands related to configuration are available:
reahl listconfig <config_directory> reahl listconfig --values <config_directory> reahl listconfig --files <config_directory> reahl listconfig --info <config_directory> reahl listconfig --missing <config_directory> reahl checkconfig <config_directory>
- Multilingual and multi-regional applications
How to make strings in your application translatable and work with translations in other languages.
If a component needs a user interface in several different human languages, each string that could appear on its user interface is marked in such a way that tools can collect all such strings in one file. (These translatable strings are usually referred to as “messages”.) For each additional language you need to support, you then provide a version of this file with translations for each such message to that language.
Reahl-component provides a mechanism for each component to ship the translations of its messages. One component can also provide extra translations for another component.
Before you can use the Reahl command line commands for working with messages, create an empty python package in which messages and their translations can be saved. Once created, register this translations package in setup.cfg as an entry point in the “reahl.translations” group.
Setting the current locale¶
Python has its own ways to set and determine what the current locale should be. Reahl code is used in server settings, however, which means that the locale used is probably different for each user regardless of the server the code is running on.
For this reason, a different mechanism is used to determine the current locale: supply an object to represent a ‘user session’ containing options related to the current user. Add a method called get_interface_locale on it which returns a locale string formatted according to RFC5646.
To make the session available to code, ensure it runs within an
ExecutionContext on which you have set the session:
class UserSession: def __init__(self, locale): self.locale = locale def get_interface_locale(self): return self.locale ExecutionContext() as context: context.session = UserSession('en_gb')
If you set the locale to a language for which you have not supplied translations, messages will just be displayed using ‘en_gb’ - the default.
The following useful commands from reahl-dev related to translations are available:
reahl extractmessages reahl addlocale reahl mergetranslations reahl compiletranslations
Remember to python -m pip install –no-deps -e . after having changed setup.cfg before using these commands.
Describing the interface of your model¶
When a system has a user interface, the values that a user enters require a lot of management:
Users can only type text, but perhaps a program would rather deal with that text as an instance of a specific class, such as a Date or boolean. “Marshalling” in this context refers to changing a given text representation into an object instance, and vice versa.
User input also needs to be validated to ensure that, for example, the text representation of a date is in an expected format to be able to be transformed into a Date object. Alternatively, a numeric input might be expected to always constrained between a lower and upper bound.
- Access control
Programs sometimes need to prevent the acceptance of user input in certain situations or only allow it from certain users. For example, if an order is already processed, its contents should probably be read only to all users.
In an object oriented program, you might have one object representing something (like a BankAccount), but user input to the same attribute of that object could be present on many different places in the user interface.
Reahl component’s modelinterface provides a way for you to describe the above requirements for an attribute of such an object in one place. Your user interface code can then be written with the knowledge that all these concerns have been taken care of, and thus will be consistently applied wherever referenced.
Fields and FieldIndexes¶
ORMs like Django ORM and SQLAlchemy use class attributes to define how database columns
map to the attributes of an object. Reahl’s
ExposedNames mechanism is purposely different so that it can be used
together with these tools without getting in their way.
To declare the
Fields that are bound to instances of a class,
ExposedNames instance to a class attribute, ie,
fields. For each
Field needed, assign a callable to an attributes
ExposedNames. The callable will be passed a single argument:
the instance of the class it will be bound to. It should return a
class Order: fields = ExposedNames() fields.processed = lambda i: BooleanField(label='Processed', true_value='yes', false_value='no')
with ExecutionContext(): order = Order() order.fields.processed.from_input('yes') print(order.processed) # prints True order.processed = False print(order.fields.processed.as_input()) # prints 'no'
Invalid input raises a
ValidationConstraint to communicate what is wrong:
order.fields.processed.from_input('invalid input') # Raises: reahl.component.modelinterface.AllowedValuesConstraint: Processed should be one of the following: yes|no
To control access to a
Field, pass single-argument callables to readable or writable when constructing the
These callables will be called when the
Field is read or written using
from_input() respectively, and
Field instance itself is passed as the single argument.
Events and Actions¶
class X: def exclaim(self): print('whoa') events = ExposedNames() events.boo = lambda i: Event(action=Action(self.exclaim))
with ExecutionContext(): boo = X().events.boo boo.from_input(boo.as_input()) boo.fire()
Event can be parameterised, which will cause its action to be sent arguments upon firing. These arguments are
deduced from the input passed to
as_input() above, hence the need for calling
This advanced topic is outside of the scope of this introductory material.
Access controlled methods¶
Each time a
secured method is executed, these check callables are first executed to check whether the method is
allowed to be executed. If either of these callables return False, an
AccessRestricted is raised. The point of
read_check is that user interface machinery could in theory use the read_check to, for example show a button, but
grey it out (ie., the user is aware of the method’s existence but cannot invoke it). A False write_check instead
could signal to the user interface machinery to not even show the said button at all.
class Order: state = 'new' def is_submitted(self): return self.state == 'submitted' @secured(write_check=is_submitted) def authorise(self): ... >>> Order().authorise() # raises reahl.component.exceptions.AccessRestricted because state is 'new'
- Regular scheduled jobs
Registering and running regular scheduled jobs.
The internals of a component may require certain housekeeping tasks to be performed regularly. A user of a system with many such components does not want to have to know about all the jobs needed by all the components used. In order to facilitate this, Reahl has a mechanism by which a component author can register jobs that the system runs on a regular basis.
List each callable object that should be run as a scheduled job in the “schedule” entry of the component option of your setup.cfg.
Whenever reahl runjobs is executed on your system’s configuration directory, all the registered scheduled jobs of all components that are used by your system are executed.
We could have opted to just register all scheduled jobs as entry points. However, a package contributes its entry points when it is merely installed. The scheduled jobs for your application should only be those jobs that are registered by components it depends on.
On a production system, ensure that this command is run regularly (say every 10 minutes) via your system’s task scheduler. It is up to the code in each such registered class method to check whether it should do any work when invoked or whether it should wait until a future invocation. For example: a job may cleans out sessions from a database, but only once a day around midnight despite being invoked every 10 minutes.
The only command from reahl-commands related to jobs is:
reahl runjobs <configuration directory>