2020 08 17 Open NEST Developer Video Conference

previous | list | next

Agenda

1. Welcome
2. Review of NEST User Mailing List
3. Project team round
4. In-depth discussion
   • Return types of getters on empty collections
5. AOB
   • Hackathon preparations

Mailing list

https://nest-simulator.org/mailinglist

• discussion of "Saving and loading entire network"
  • has been tried various times
  • need to collect use-cases (there were several on the mailing list already)
  • precise definition of requirements is very difficult to collect
    • what does "reset the network" mean (e.g. #1618)
    • open state, like membrane potentials, etc. (easy)
    • closed state, like internal state of (hardware) RNGs, system level network buffers, etc. (VERY HARD)
    • loaded (maybe user provided) extensions, ...
  • new idea: no more Simulate() after save(), so reliably restart, but no continue
  • point of discussion at next Hackathon
    • should result at least in a document detailing our current understanding

Project team round

Here we discuss topics that need broader attention, for example questions that came up but are outside a single project's scope, larger planned changes/PRs that affect all teams or pending work that is blocked by external factors.

1. Models / NESTML
   • still looking for domain expert to review Tsodysk synapse
2. PyNEST
   • work progressing smoothly
   • what should be returned with .get() on an empty node collection #1651 -> IN DEPTH
3. Kernel
   • no updates
4. Installation
   • no major issues
   • looking for people working on MacOS
5. Infrastructure
   • Travis is OK
   • interesting Python related PRs merged
   • MacOS builds with Python-bindings are not supportd on Travis
     • currently the build semi-automatically finds the right libraries
     • works, but have to see in future
6. Documentation
   • discussed a lot about the structure of the documentation, still going on
7. EBRAINS
   • no updates
8. (Feature) Random number generation
   • work is ongoing, PRs are being reviewed and worked on
9. (Feature) Automated Testing
   • still missing some reviews, but almost done with the migration of mymodule . #1662
   • create tests for some specific parts (structural plasticity, dopamine, ...), -> Hackathon

In-depth discussion

Discussion of possible implementations for #1651

• no-op on empty collections for set operations is in line with numpy empty arrays
• empty node collection should be detectable same as empty standard containers; cast to bool should return false
• getting properties from empty NodeCollection has no defined return value
  • return None
    • user needs to check explicitly
  • return []
    • user needs to check explicitly
  • raise BadValue()
    • user notices automatically
• no matter what, the user needs to handle it. making this explicit with an exception is the robust solution.
• having a known data type that you would expect can be as useful
  • always returning the same data type makes the API behave more "as expected"
  • currently GetStatus() may return a number of different types, depending on many aspects.
• could happen with
  • stochastic networks
  • getNodes() with an attribute selects only nodes with that attribute (may be none.)
• new feature, difficult to estimate the usage in the past.

Conclusion: throw exception, that's safe, then if return types are defined better, this may be lossened.

AOB

• Hackathon page and organization is ongoing, add topics there, beginning of hackathon we will go through/sort everything.