Skip to content

DevNotes.md

Latest commit

 

History

History
96 lines (86 loc) · 7.54 KB

DevNotes.md

File metadata and controls

96 lines (86 loc) · 7.54 KB
Raw

(delete once Anya is done with joint refactoring)

Further refactoring list:

  • Add SGP mode to API reference doc

  • Look at fun_host_sym_stress_trans_compatibility_check to try to reduce code duplication of task-profile setups, possibly with decorator pattern, but also definitely just in own file

  • Move world properties back into protected and make necessary accesssors

  • Try to fold ProcessStressEscapees into existing code/reduce duplication

  • Rename "SetReproCount" to lineage length since it's confusing (or did I already?) //I did this

  • Look into what is going on with SGPHost local sgp_config not working

  • Streamline Host ProcessOutputBuffer and check if easier access to some variables

  • Break Host Reproduce into helper functions

  • Compare ecto relevant code (i.e. default mode) between main and this refactor to see if something changed, when was the last time the ecto integration test didn't seg fault on Mac? Prior to aux bump?

Previous actions completed

[x] Change renamed config settings (ORGANISM_TYPE) (also cleaned up some other unused config settings) [x] Get Host task credit and Task Match Check tests not seg faulting [x] Make mini guide on main changes for tests so others could help with test porting [x] Move Host reproduction, output buffer processing, and process back into Host class and make sure current tests keep passing [x] Add citations and how to cite sections to README (on main), cite empirical JOSS paper, SignalGP arxiv, phylotrackpy on arxiv [x] Decide whether to try to fix ecto integration test or remove the graveyard fix and merge into main since it apparently works on github?? [x] Make new version bundle on main for complex features [x] Move Symbiont code back into classes [x] Put PrintCode back into SGPHardware for Get dominant printing [x] Put in IO Zero not getting rewarded and test it [x] Port more Symbiont tests [x] Make a list from this commit for further shifting todos and update this doc with those todos https://github.com/anyaevostinar/SymbulationEmp/commit/9ea1d53c8bf70c612d1454fac0510ddaf0c70e9d for AEV TODO and Refactor note for what else I already had decided would be good to do [x] Horizontal transmission to make parallel to default mode and integrated to support tags with tasks - Decisions made: reproduce is shared between horizontal transmission and free-living sym reproduction. If free-living sym repro is on, then reproduce places offspring into sym pop (like in default), and then offspring can infect with Infect instruction (not yet implemented). If FLS is off, reproduce does horizontal transmission. Also decided that if HT is off, an "attempt" is not counted. [x] Move SymDoMutation into SGPSymbiont to parallel host

Journal

4/15/26

  • Some really confusing errors were happening and I realized that I had lost CURE and other main code that had both touched default-mode, so I started again by branching from main, merging in complex-syms-clean, bundling up all of sgp-mode into "old-sgp-mode" and then merging in Alex's sgp-mode-refactor to make sure that the Cure code came in along with various improvements to default-mode that had been made across the branches. The current version is pure Alex refactor with just the tests that look like they had been updated and a couple that I had written as I was doing my initial attempt at refactoring. Several tests are seg faulting for unclear reasons. I'm somewhat certain that hosts are able to reproduce and get points for doing a task and mutate correctly. Next steps will be figuring out why checking that hosts get credit for doing a task and Task Match Check are seg faulting, then going back to moving code back into Host and Symbiont along with porting all of the tests to work with the new style. I think I will make a guide for the main changes that need to be made to the tests so that others could possibly help with the porting and we can be more confident that things are currently working correctly. 4/3/26
  • Task test working, needed to add a call to AssignNewEnvIO in the test, which seems weird since it seems like that should be handled by world.Setup correctly
  • Moved output buffer processing back into a host method 4/2/26
  • Adds tasks and test for it, currently failing 4/1/26
  • Added to test to make sure orgs are getting older and it is failing
  • Fixed by adding org to pos 0 instead of pos 1, a bit nervous about why pos 1 wasn't working
  • Got Host process all back in
  • Mutator called from Host's Mutate now, works when directly called, but I can't see where HostDoMutation was ever called previously so I'm nervous - Alex confirmed it was only called from SGPHost::Mutate, so we're good 3/31/26
  • Building with scheduler code, need to test still 3/20/26
  • SGPHost can be created in SGPWorld, SGPWorld test with baseline host works, scheduler is copied over but not tested

Notes

These are function flows that are complicated to follow:

Horizontal transmission

  • Reproduce Instruction -> checks cycles, marks repro attempt, (DOES NOT check for HT being on, since that is handled by Symbiont)
  • Process -> Checks for repro attempt, calls AttemptIndependentReproduction
  • AttemptIndependentReproduction -> Calls Symbiont's super method to check if HT is on, check points, set points to zero (need to try changing to decrement in super and see if everything breaks), marks attempt in data node; submethod then adds to reproduce queue
  • Repro queue calls fun_reproduce_org
  • fun_reproduce_org -> calls org->Reproduce and then SymDoBirth, then after_sym_do_birth since that's when we have access to parent for easier data tracking
  • Reproduce -> calls super class Reproduce (which handles tag updates, etc) and sets up parent task count and lineage tracking
  • SymDoBirth -> calls fun_sym_do_birth
  • fun_sym_do_birth -> calls freeliving or SymAttemptHorizontalInfection (renamed to avoid confusion with HorizontalTransmission) depending on config
  • SymAttemptHorizontalInfection -> finds a host and adds sym to host
  • after_sym_do_birth signal -> calls sym_parent super's AfterHorizontalTransmission, which records success in datanode if appropriate (including int val of parent)

Reproduction

  • Repro Inst -> state.markReproAttempt
  • Host Process checks Repro attempt -> calls AttemptReproduction
  • AttemptReproduction -> marks ReproInProgress, adds to repro queue
  • World processes repro queue after all organisms process
  • Repro queue process calls Reproduction and handles HostDoBirth

InputBuffer setup

  • Input Buffer starts at size 0
  • CPUState SetInputs is called by SGPWorld Assign New Env IO
    • This doesn't seem to be happening during the tests without specifically calling it, which seems not ideal
    • I think I fixed this, line 1074 of WorldSetup was commented out still...

Old Timeline plan (deprecated 4/15, see Journal entry)

[x] SGPWorld with basic SGPHost [x] Get scheduler working [x] Get host reproduction working (with no point requirement) [x] Get mutator working [x] Get tasks working (and hosts completing them and getting points) [-] Get all host tests working with new setup * Decided better to get refactor complete with some tests so that others can then jump in again, so will be doing minimal test porting, just enough to convince myself things are functional [ ] Get symbionts existing (no interaction) [ ] Move processsymbiont functionality back into symbionts [ ] Get Health interaction working [ ] Get Stress interaction working [ ] Get Nutrient interaction working [ ] Get data tracking working [ ] Port Multi infection code from Elias fork [ ] Finish implementation of temporally changing environment [ ] Make sure spatial structure is working? [ ] Go back through for other refactoring todos