pharo-xp-tools / xpimagegeneration

XPImageGeneration

Metacello new
  baseline: 'XPImageGeneration';
  repository: 'github://Pharo-XP-Tools/XPImageGeneration:main';
  load.

Quickly create an experiment

• Create a fork of this repository.
• Install the framework using the previous baseline, but pointing to you fork.
• In Pharo, search for the class ExampleExperiment in the Experiments package. Copy it (right click, copy it, write the new name in the dialog).
• Rewrite the instance methods of your new experiment class.
  • The tasks displayed in the PhexTasksPresenter will be loaded from ptx files in the folder tasksDefinition/*your-experiment-id* located in the root directory of the repository.
  • The method prepareSystemForExperimentis called when your experiment is loaded. Code there to make all required modifications in the system needed for your complete experiment. (The system modifications can be part of you fork, or be in another baseline. Your choice).
  • Commit and Push XPImageGeneration changes to your fork.

Participants' point of view.

• Participants must download a specific Pharo 9.0 clean image.
• They need to execute the same baseline from your fork. A message will appear once everything is loaded.
• In the World menu, they need to click Library>Phex. And then load your experiment.
• They see the tasks you defined (the ones in tasksDefinition/*your-experiment-id* folder) and the experience is started.

Tasks definition

• TODO: the ptx file production
• TODO: System changes

Logging

• TODO: Logging modalities (centralized, local)
• TODO: Instrumenting your experiment.

xpimagegeneration's Issues

Avant les pilotes : tester l'xp entière

In Ammolite, reduce the number of students named Adèle to 4

General pass over english

Execute the pilot on two participants

• Experimented participant
• Less experimented participant

Ask then to record their screen.

Surveys validation and default options

• sharing data and accepting the experience: opt-out as default
• checkboxes and radios: nil as default (no answer)

The experiment's tasks should be ordered and not be doable in a random order

Write supervision protocol

Fix Lights Out is currently not following the rule, the clicked light is supposed to be switched too

Convert our tasks to debugging tasks

A fix in the code is expected from participants.

• Ammolite
• Lights-out

Update Lights-out's to avoid "object -> inspect -> self toggleState (debug it)"

Goal: avoiding the use of "object -> inspect -> self toggleState (debug it)"
Idea: Add as a parameter the required color upon which the state is changed

Rebuild tasks with their new names and delete warmup 2

New design:

• Standard debugger: Task 1 or 2.
• Tutorial.
• Warmup.
• OCD: Task 1 or 2.

Add confirm popup when finishing a task

Move ExperienceLogger mechanics to Phex

There is a dependency from Phex to ExperienceLogger.
This class, denpendencies and dependents should be prefixed by Phex and moved into the Phex repository.

Fix in the first survey the error: "If you found the bug, how much the debugger helped to discover the bug?"

Answering "yes" to the intro survey must allow data sending

Build internal supervision organization protocol

When installing, logging is not automatically configured

Tasks have not text...

Remove the example experiment from the list of real experiments

This will confuse participants when loading an experiment.
Perhaps: load automatically the intended experiment instead of asking which one to load?

Add a button "comprehension" with a record model for participants to tell each time they think they understood the problem

Make sure we have a 50/50 tasks ordering

Add a dev baseline that sets the Phex dev mode before loading anything

Find 2 pilot participants

Serialize tasks once then git them instead of serializing them each time the experiment is started

Record a short video in addition to the tutorial.

There are already videos and animated gifs here: https://thepharo.dev/2020/07/16/object-centric-breakpoints-a-tutorial/
We need to check if they are good enough to be used directly.

Shorten the tutorial

Build list of 9 trusted participants

Survey ar enot correctly ordered if there are many of them in a sequence

Surveys are not correctly set in the correct tasks

Also the sequence is wrong given the experiment definition.

The experiment is not loading data