Refactoring

---

November 9th:

Fixed all immediate errors, merged into develop.

Notes:

• Port Dropdown properties to YAML -> old properties link to a YAML file
• Add a method that uses a program + the robot name + a general default config to generate valid programs for integration tests
• Add a debug block for output without the need for an actor
• Pull up Jaxb2Ast of individual sensors to the generic one
• Fix package names (AST), same structure on all plugins
• Find a way to use expert toolboxes as a basis for actor/sensor tests
• Have a look at Ev3 sim's LightSensor
• Check whether NO_PORT is really needed

November 6th:

Removed all errors from the refactoring branch. Started fixing tests for EV3.

Changes:

• Extracted a lot of functionality of the transformers into helper classes.
• Generalized a lot of the transformers.

Notes:

• Check the constructor of ExprParam
• Remove general modes and replace them with properties -> currently worked around with a second getEnumCode
• Check UsedSensor redundancies
• Add a hidden field to blockly, that specifies if the configuration block is an actor or sensor
• Check whether the reverse field of motors in the configuration is really needed
• Have a second look at the new JaxbTransformers
• Go through all actions and replace getPort() with getUserDefinedPort()

October 30th & November 2nd:

Finished refactored configuration for NXT.

Changes:

• Moved NXT configuration to the robot project
• Added ConfigurationComponent as a basis for all configuration blocks
• Moved ActorType and SensorType to properties
• Moved modes to properties
• Removed some inverse transformations
• Introduced new constants interface SC

Notes:

• NXT brick block is called EV3 brick block
• Check if configuration blocks exist with blockly names -> prevent injection when changing xml
• Endgoal: blockly blocks are created from properties
• Use Map directly for ConfigurationComponents instead of constructing it from lists
• Refactor BlocklyDropdown2EnumHelper and factory's methods to more generic ones
• Check user defined names against configuration in BlocklyDropdownFactory
• Have a look at new SC constants, try to autogenerate
• Check usage of MASMETHOD in the NXT tests
• Check test11 in the NXTSimulationVisitorTest

October 26th:

Finished moving sensor and dropdown modes to properties. Started rewriting configuration, starting with NXT as basis.

Changes:

• Finished common handling of wait_until sensor block with properties
• Added a new configuration

Notes:

• Decide on color vs colour
• Decide whether to rename OpenRobertaRobot to RobotCommon
• Write a unit test that loads all classes for reflection
• Make all of the modes in the new properties consistent, only include the needed ones
• Extract a sensible naming strategy for the properties

October 23rd & 24th:

Started refactoring of the IRobotFactory, split into BlocklyDropdownFactory and IRobotFactory. Moving the modes into the properties.

Changes:

• MicrobitFactory now inherits from AbstractMbedFactory
• Renamed set/getProgramText to getSourceCode in ICompilerWorkflow
• Added new classes for dropdown maps
• Split IRobotFactory into BlocklyDropdownFactory and IRobotFactory:
  • Static functions moved to helper
  • moved dropdown specific stuff to the new factory
• Started to move enums to the properties

Notes:

• Reconsider the EV3 predefined images: remove from server side/move to hal, only include if used
• Clean up BlocklyConstants, try to autogenerate it
• Is getCommandline and getSignature in IRobotFactory still needed?
• Generalize dropdowns for sensors & actors
• Add constants to use in code generation for the new property based modes?
• Sort the sensors/actors in the properties
• Remove redundancy in Calliope and EV3 properties
• Extract interface from AbstractEV3Factory
• Move usages of BlocklyDropdownFactory up

October 15th:

Refactoring of the actor visitors (1, 2)

Changes:

• Split sound visitor into sound and speech visitors
• Split motor visitor into motor and differential motor visitors. Differential motor visitor extends the motor visitor
• Default implementations for not supported actors are in the robot specific interface
• Renamed communication visitor to bluetooth visitor
• Removed control visitor, moved relay action down to Arduino
• Added IActors4AutonomousDriveRobots as super interface for generic wheeled robots
• Removed many empty return nulls or defaults and pulled them up to the robot specific interfaces to throw exceptions
• Revised common Arduino visitor

Notes:

• Discuss if a sensor hierarchy is possible at this point of time we can not find any beneficial structure
• Revise LEDs for Arduino
• Look at Mbed in general and improve it (Structure, Motor, SingleMotor, BothMotor, missing microbit methods)
• Write a script that validates test xmls are actually used in tests
• Look at package names (actors vs action)
• Improve Arduino properties

October 9th:

Refactoring of the factory names, code generation visitors and used hardware visitors.

Changes:

• Added Abstract to all touched abstract classes
• Added I to all touched interfaces
• Factories & compiler workflows now include the robot name in the class name
• Factories are in the de.fhg.iais.roberta.factory package
• Compiler workflows are in the de.fhg.iais.roberta.codegen package
• Same factory & compiler workflow package names and structure for all robot plugins
• Code generation & used hardware visitors now include the robot name in the class name
• All visitors can be found under the visitor package
• Language specific visitors are in the lang package
• Hardware specific visitors are in the hardware package
• Used hardware collector visitors are in the collect package (not final structure, has to be refactored?)
• Program checking visitors are in the validate package (not final structure, has to be refactored?)
• Same code generation package names and structure for all robot plugins
• Timer & Bluetooth functions were moved down the simulation visitor chain (this has to be refactored)

Notes:

• Have a look at the Guice modules (e.g. Ev3GuiceModule) and its package structure
• Bob3, Botnroll & Mbot factories should be subclasses of AbstractArduinoFactory
• MicrobitFactory should be subclass of AbstractMbedFactory
• Refactor hardware specific parts in the simulation visitors, maybe split into lang / hardware as real robot ones
• All exception messages in IActorVisitor should be replaced
• Should the AbstractStackMachineVisitor be moved to the robot project?
• Final should be added to all leaf visitors to express they are leaves
• Add a general IRobotVisitor between robot specific ones and IActorVisitor and ISensorVisitor?
• Look at all of the visitors and replace return null with exception throws where adequate
• All high level visitors need default implementations
• Look at isXUsed and UsedSensors/UsedActors structure and find a better (common, if possible) solution
• Synchronise test packages with the new package structure
• Split checkers into similar lang/hardware package structure, if possible
• Look at "Brick vs Board", replace everywhere with more descriptive names

---