Developing your own component¶
This section is relevant to anyone building a ‘just a’ web application: Remember, when you program using Reahl, every bit of code ends up being packaged somewhere as part of a component. This is true for a web application as well, even if you do not want to re-use the application in other applications. Code needs to be packaged and distributed, and these packages are what we refer to as components.
There are some important issues addressed via this underlying component model, and this section provides a brief introduction to some of them.
Making a component configurable¶
Each Reahl component can have its own configuration. As explained in Getting started with Reahl, when an application starts up, the configuration for each component that forms part of the application is read from a separate file, from a shared configuration directory. We have covered how to specify the configuration settings required by other components before. It is time now to show how to make your own code configurable using this mechanism.
To illustrate, we made slight changes to the simple address book application presented in User interface basics. The application consists of a single View. Amongst other things, that View contains a heading with the text “Addresses”. In the example presented here, we give the original example a twist: here, whether that heading is shown or not is based on the value of a configuration setting.
To make a component configurable, you have to provide a class that governs configuration for your component. The component infrastructure also has to be made aware of this class, by registering the class in the .reahlproject file of your component. Registering the configuration class is the easy part. Just include a configuration element in your .reahlproject:
Remember to run reahl setup -- develop -N after editing the .reahlproject file so that those edits can take effect.
Config files are evaluated as Python code, but variable names are made available in these files so that a config file basically looks like a list of assignments to some predefined settings. For example, a config file for our example would contain:
componentconfig.showheader = True
When the component framework reads the configuration for your component, an instance of the registered configuration class is created and assigned to a variable name while the config file is evaluated. In this case, an instance of AddressConfig is assigned to the componentconfig variable while the config file is evaluated.
Thus, setting config in this file amounts to setting attributes on an instance of your registered configuration class. In that class, you specify the name of the config file of the component (which by convention ends on .config.py) and the name of the variable to which an instance of your Configuration class will be assigned. Each of the config settings for your component are also defined here.
The configuration class itself inherits from Configuration (in reahl.component.config). Here is AddressConfig:
class AddressConfig(Configuration): filename = 'componentconfig.config.py' config_key = 'componentconfig' showheader = ConfigSetting(default=False, description='Whether the title should be shown as a heading too')
As you can see, filename and config_key respectively specify the file name of the config file and the name of the variable to which an instance of AddressConfig will be assigned while evaluating that file. To declare a specific configuration setting, a ConfigSetting is created and assigned to the name the setting should have, as is shown in the code.
It is good practice to always provide a default value for a setting. If all your settings have sensible defaults, your application can be started up and run without any config file present at all – something a new user will appreciate. Similarly, the description is used to enable useful output when you run, for example:
reahl-control listconfig -i etc/
Defining what configuration is needed by your component is only part of the story. Actually using the configuration is the other part. A programmer can get to the configuration for the currently running application from anywhere in code. Just obtain the current ExecutionContext, and ask it for the .config. There’s an attribute on this global config for each config_key of each Configuration read, so that reading the config looks pretty much similar to setting it. See how it is done in this example:
class AddressBookPanel(Panel): def __init__(self, view): super(AddressBookPanel, self).__init__(view) config = ExecutionContext.get_context().config if config.componentconfig.showheader: self.add_child(H(view, 1, text='Addresses')) for address in Session.query(Address).all(): self.add_child(AddressBox(view, address)) self.add_child(AddAddressForm(view))
Database schema evolution¶
The Sqlite database itself does not support migration of existing data very well, and as a result migration is only possible on PostgreSQL databases. See these posts for more information on the issue: one of the last bullets of goals of alembic and Christopher Allan’s rant about the issue.
Right at the end of A basic model we show how one registers all persistent classes of your component in a .reahlproject file, and that the reahl-control script uses this information when you run reahl-control createdbtables etc. These mechanisms are adequate for managing a database when you develop a brand new application, or an example. What happens much more commonly is making changes to the code of an application that is already running somewhere, with existing data in its database. Changes to the code of that application frequently require changes to the underlying database schema as well, and such changes to the schema need to be done without losing the data that’s already in the database.
Reahl provides some infrastructure to help you deal with such “Migrations”. Since Reahl sets out to let one reuse components in different projects, its migration infrastructure caters for migrations to be written for a specific component, without containing any knowledge of the other components may be used with.
To show how Migrations are handled, we have another example. This example is again a variation of the simplest AddressBook application developed in User interface basics. This time, we extend Address to include the date when the Address was first added. The date is simply displayed next to an Address when listed.
In order to simulate a program that changes over time, the real added_date is commented out in the example, leaving a hardcoded ‘TODO’ in its place. This makes it possible to run the application with a database schema that does not include added_date at first. A new schema will be needed when the actual added_date is uncommented, without the need for other code changes:
class Address(Base): __tablename__ = 'migration_address' id = Column(Integer, primary_key=True) email_address = Column(UnicodeText) name = Column(UnicodeText) # added_date = Column(DateTime) added_date = 'TODO' @exposed def fields(self, fields): fields.name = Field(label='Name', required=True) fields.email_address = EmailField(label='Email', required=True) def save(self): self.added_date = datetime.now() Session.add(self) @exposed def events(self, events): events.save = Event(label='Save', action=Action(self.save))
As you know, each Reahl component has a version number. When you install a new application, you will create the initial database schema using reahl-control createdbtables etc. While creating the schema Reahl will make a note (in the database) of which version of your component the schema was created for. When you work on a new version of your component, you have to write one or more “migrations”. A Migration is one change that needs to be made to the schema (and perhaps data) of the previous version, in order to bring the schema in line with the new version of the component. A migration can be as simple as adding a single column. Migrations can also require more complicated code specific to your problem domain that does many database changes in a specific sequence to get the schema changed without damaging the data.
After installing a new version of your component, you need to run the following in order to migrate the old schema, using your provided Migrations:
reahl-control migratedb etc
The migratedb command first checks to see which version of your component the current database schema corresponds with. All Migrations for your component are then inspected, and only ones that are needed to bring the schema up to date with the currently installed version of your component, are run, in order.
The code of each migration you write must include the (new) version of the component it is for. Each migration should also be registered in the .reahlproject of your component – in the order in which migrations should be run.
A Migration is a class which inherits from Migration. It should have a class attribute, version, which states which version of your component it is for. A programmer codes migration logic in the .upgrade() method of a subclass of Migration.
Before an explanation of that, here is the Migration for adding our added_date column:
class AddDate(Migration): version = '0.1' def schedule_upgrades(self): print('scheduling upgrades for AddDate') self.schedule('alter', op.add_column, 'migration_address', Column('added_date', DateTime))
The code doing the changes to the schema is written in terms of SqlAlchemy’s migration tool: Alembic. Alembic is meant for also keeping track of versioning of schemas and migration scripts. Reahl does not use those parts of Alembic since Reahl has to keep track of versions of components already. Reahl also deals with the fact that a single application contains many components, each one of which has its own schema, Migrations and versions. When Migrations are run, however, Alembic is set up in the Reahl environment for a programmer to just use the alembic.op module as any Alembic programmer normally would do.
There is another catch to migration that Reahl helps you solve: Migrations are complicated by the fact that some database objects may be dependent on others. For example, you may want to rename a primary key column that is referred to by a foreign key constraint. In such a case, you will have to first drop the foreign key constraint and the primary key constraint, alter the column to be renamed, then recreate first the primary key and then the foreign key. To make matters worse, the said foreign key could be part of a component that does not itself include the table to which it refers.
To solve this particular problem, migration code is not run immediately. In each Migration, the programmer merely schedules all the calls that would effect the necessary changes. Actual execution of these calls happen only once all components had a chance to schedule their migration calls. Execution of these calls happen in a number of predefined, named phases. When a call is scheduled, it is scheduled to run in a particular phase. All calls that are scheduled in a phase are executed when that phase is executed, regardless of which component scheduled them. Only once the calls for a phase are done, the next phase is executed.
Here are the phases, and the order in which they are finally executed:
- Foreign keys are dropped first, because they refer to other columns.
- Primary keys are dropped next, they may also prevent other actions from completing.
- Sometimes some code needs to be executed before tables are altered – saving some data in a temporary table, for example, or disabling some other constraints.
- Now that all possible constraints have been disabled, tables and columns may be altered.
- Then, primary keys can be created again.
- Followed by indexes dependent on those primary keys.
- With a schema mostly fixed, data can be inserted or moved to new locations.
- A last chance to recreate foreign keys to possible newly moved data in the new schema.
- Use this phase if any cleanup is needed of temporary tables, etc.
Registering the AddDate class in the .reahlproject file is similar to registering persistent classes, it just happens in a <migrations> tag, and the Migrations are listed in the correct order:
<migrations> <class locator="reahl.doc.examples.tutorial.migrationexample.migrationexample:AddDate"/> </migrations>
Try the example out¶
To appreciate the example, you’d have to get your own copy of it (via reahl example tutorial.migrationexample), and run everything you need to in order to get its database tables created: reahl setup -- develop -N and reahl-control createdbtables etc. Then run the application and add a few Addresses. Doing all of this simulates an application that ran somewhere for a while, with some data in its database.
The next step is to change the application:
- comment out the ‘TODO’ version of added_date, and uncomment the new implementation
- edit the .reahlproject file and increase the version of the component (in this example the version is hardcoded in the .reahlproject file precisely so that you can edit it).
After editing .reahlproject, remember to run, as usual:
reahl setup -- develop -N
Finally, to change the database schema, run:
reahl-control migratedb etc
Sometimes one needs some code to run at certain times to do some regular housekeeping tasks. You can ship such jobs with your component and register them with the component infrastructure. Remember, your application is composed of many components, some of which you are not the author of. Running the following will run all the scheduled jobs of all the components that are part of your application:
reahl-control runjobs etc
As an example, we’ve modified the AddressBook example in User interface basics again. This time, we’ve added a boolean flag in the database for each Address that is true if the Address was added today. When an Address is added to the database, this flag is set to True. Address also has a class method, .clear_added_flags() which sets all flags back to False. If this method is run once every day, the flag would be True on newly added Addresses only.
class Address(Base): __tablename__ = 'jobs_address' id = Column(Integer, primary_key=True) email_address = Column(UnicodeText) name = Column(UnicodeText) added_today = Column(Boolean, nullable=False, default=True) @classmethod def clear_added_flags(cls): for address in Session.query(cls).filter_by(added_today=True): address.added_today = False @exposed def fields(self, fields): fields.name = Field(label='Name', required=True) fields.email_address = EmailField(label='Email', required=True) @exposed def events(self, events): events.save = Event(label='Save', action=Action(self.save)) def save(self): Session.add(self)
To register this job with the Reahl component infrastructure, add a <schedule> tag in the .reahlproject file:
When you try out the example, do everything you need to do in order to be able to start the application. Then add some Addresses using the application. All Addresses will be shown as new. At this point, you can run:
reahl-control runjobs etc
Then refresh the home page of the application again, and you will see that the Addresses are not listed as being new anymore.
This example is perhaps not a very good one... because not all jobs of all the components used by your application will need running only once a day! A job, such as Address.clear_added_flags() should include code that makes sure it only does its work when necessary. You can then use your operating system tools (such as cron) to run reahl-control runjobs etc very regularly – say once every 10 minutes. Each scheduled job will now be invoked regularly, and can check at each invocation whether it is time for it to do its work, or whether it should ignore the current run until a future time. This way, all jobs can get a chance to be run, and be in control of when they should run.