User:Vikram Atreya/GSoC21/Project


Implementing UNDO command[edit]

Community Bonding[edit]

  • Set up the blog and wiki
  • Decide on which approach to implement UNDO

Development Plan[edit]

  • Week 1: Decide architecture and create the required files to have an undo cmd
    • Add files to libged, make changes in Cmake files and decide on space/location to store the temporary data being generated. Write documentation on how the undo command is going to be implemented and how it will work.
  • Week 2: Start implementing the UNDO logic
    • Check for different commands like kill, make and other basic commands for proper compatibility with the UNDO command.
  • Week 3: Continue the implementation
    • Start working and testing for complicated commands like dbconcat which deal with overwriting, creating new objects and deleting multiple objects. In the partial + events method this might take more time as splitting commands into events is a big task.
  • Week 4: Get a basic review and start testing
    • Take inputs from my mentor on current progress and how I could improve the code. Write tests for various cases and check robustness.
  • Week 5: Continue testing and optimizing code
  • Week 6: Work on further optimizations / Account for unforeseen delays
    • I will keep this week as a buffer.
  • Week 7: Optimizations to UNDO
    • Depending on the method chosen, work on improving the time/space complexity of the implementation.
  • Week 8: Document work and issues faced
  • Week 9: Prepare for final evaluations

Design ideas[edit]

  • Facilitate 10-15 undos from the latest state
  • Initially no redo command and later states are deleted once undo is used
  • Works only on objects and not on other data stored
  • Plain undo with no flags does just 1 undo
  • ``` -n X``` flag does multiple undos, where X is no of undos to be done

Implementation ideas[edit]

  • Full checkpoint is stored before 15 commands.
  • Form an array with values as names of all non-hidden objects, search for backups in hidden objects (backup should have a version just previous to the no of undos requested). After iterating through all backup objects if we do not find a backup, ,which should ideally be found in the full checkpoint even if it was not backed up later, we killthat object.
  • Undo can also be considered to be a normal command and full checpoint can be created before undoing, so that redo will be done with an undo.