An example of requesting agents would be useful, as the API docs are maybe not enough for new users here.

Use the buffer ngh data to synch the relevant part of the tensor matrix.

1. Translate the buffer envelope into a tensor slice / index
2. Select from the tensor and transform to np.array
3. Send the np.array neighbor with info about which part of the tensor matrix to update
4. Update tensor matrix with the received data

• Is 3D buffer sync tested? (SharedContextTests test 2D buffer sync for continuous and grid)
• Is periodic boundary tested with buffers?

From a user:

I notice that with v. 1.1.0, if the first "at" of the schedule is 0 it is skipped and the execution starts with the next one.
Also the same actions at intervals +1 do not operate. With the previous version, this was not happening.

Hi, me again.
im running a simulation via matplot.pyplot.plot lib, but my image shows only the agents on the space in the current rank /bounds per tick. What's needed to show all the agents on the space for all the cores simultaneously per tick and on a single Image file.

Example below - Tick 1

Rank 0 Image

Rank 1 Image

Wanted: All ranks Image

Thanks for your time
Alex

The networkx graph API for getting the in and out edges appear to be costly, e.g.:

in_count = len(self.network.graph.in_edges(person))
out_count = len(self.network.graph.in_edges(person))

are pretty slow for large networks (30,000 nodes and 15,000 edges). We should investigate if there are
more efficient API calls in networkx that provide edge counts, and edge views in general, as in iterating over
agent network edges like

donor_list = [x for x,_ in self.network.graph.in_edges(self)]

Furthermore, we might want to provide repast4py network API for these functions so the user doesn't need to reference the underlying networkx graph.

See: https://stackoverflow.com/questions/76756592/can-i-run-repast-simphony-in-batch-mode-with-java-17-or-greater

Provide support for GIS on a single process, in a way that is similar to the RepastS Geography projection. Single-process GIS will come with the caveat that the Geography won't automatically move agents across process and should only be used on a 0-rank model.

At minimum, this should include:

• Geography projection or container
• Add / move / remove agent - geography pairs to the geography
• Get agent geometry from Geography
• Spatial index in geography for spatial queries of objects within bounding box
• Coordinate Reference System (CRS) support in Geography
• Shapefile loading and/or agent-factory from shapefile feature data

Python packages for implementation:

• Shapely - provides JTS-style geometry classes (Point, Line, Polygon) and spatial functions like intersect, distance, etc.

  • https://shapely.readthedocs.io/en/stable/measurement.html

• fiona - Shapefile loading in a way that supports generating Shapely geometries

  • https://fiona.readthedocs.io/en/stable/

• pyproj - CRS projection types and transforms

  • https://github.com/pyproj4/pyproj

• rtree - spatial indexing

• GeoPy for geodetic (distance) calculations

  • https://geopy.readthedocs.io/en/stable/#module-geopy.distance

• RasterIO for read/write raster data

  • https://rasterio.readthedocs.io/en/stable/

• U. of C. spatial data group suggested hierarchical spatial data formats: "h3", "s2" -- https://h3geo.org/

• GeoPandas provides a lot of the functionality we had in GeoTools (I/O, space predicates, CRS, etc), however the data are stored in a geo dataframe (pandas) and we suspect not suitable for frequent inserts/deletions of agents. Since the pandas dataframes leverage numpy and C++ under the hood for speed, we might be able to use the dataframe as the space’s agent-geometry mapping – assuming that’s compatible with the repast4py projections.

We use a quad tree in the continuous space for efficient bounds queries, which are necessary when populating the buffered areas between processes, and for fast (or at least faster) get_agents_within queries. The threshold argument to the SharedCSpace specifies the maximum capacity for a node in the tree. When this capacity is met, the node splits, and the agent locations in that node are allocated to the newly created children. When the number of agents at a single point >= the threshold, the splitting will occur every time an agent is added to that point. The splitting also doesn’t actually accomplish anything, given that agents are all on a single point. The branch at that point gets longer and longer, and so the time it takes to add (and to a lesser extent to remove) agents also takes longer and longer.

Possible Fixes:

1. Make sure user has large enough threshold - warning message when some tree depth is reached
2. Threshold should be for number of different points in the node's items rather than raw agent count
3. User specifiable max depth
4. Allow user to choose among these with 1. being a default

Note that 2. would require storing the number of agents at each point and incrementing / decrementing that when agents are added and removed.

Let's assume we have a directed network with 3 nodes in 3 different processors (ranks). For instance, consider the following network.

network 1
1 0 0
2 0 1
3 0 2
EDGES
1 2
1 3

For finding the neighboring nodes with DirectedSharedNetwork.graph.successors() or predecessors(), it will return agents from one rank and ignore neighboring agents in other ranks. for example

DirectedSharedNetwork.graph.successors(self.context.agent(1,0,0))

will return Agent(3,0,2) and not [Agent(3,0,2) , Agent(2,0,1)]. This is also the case with the undirected graph and UndirectedSharedNetwork.graph.neighbors(). Therefore I also think the rumor example results are not correct.

Implement custom ngh extent beyond the default of 1 in geometry.py

Grid and continuous space have only multiple occupancy type. Implement single.

Looks like numeric_liimits has moved in gcc included with Ubuntu 22.10

Building wheels for collected packages: repast4py
  Building wheel for repast4py (pyproject.toml) ... error
  error: subprocess-exited-with-error
  
  × Building wheel for repast4py (pyproject.toml) did not run successfully.
  │ exit code: 1
  ╰─> [44 lines of output]
      running bdist_wheel
      running build
      running build_py
      creating build
      creating build/lib.linux-x86_64-cpython-310
      creating build/lib.linux-x86_64-cpython-310/repast4py
      copying src/repast4py/network.py -> build/lib.linux-x86_64-cpython-310/repast4py
      copying src/repast4py/context.py -> build/lib.linux-x86_64-cpython-310/repast4py
      copying src/repast4py/util.py -> build/lib.linux-x86_64-cpython-310/repast4py
      copying src/repast4py/geometry.py -> build/lib.linux-x86_64-cpython-310/repast4py
      copying src/repast4py/parameters.py -> build/lib.linux-x86_64-cpython-310/repast4py
      copying src/repast4py/space.py -> build/lib.linux-x86_64-cpython-310/repast4py
      copying src/repast4py/logging.py -> build/lib.linux-x86_64-cpython-310/repast4py
      copying src/repast4py/schedule.py -> build/lib.linux-x86_64-cpython-310/repast4py
      copying src/repast4py/__init__.py -> build/lib.linux-x86_64-cpython-310/repast4py
      copying src/repast4py/core.py -> build/lib.linux-x86_64-cpython-310/repast4py
      copying src/repast4py/random.py -> build/lib.linux-x86_64-cpython-310/repast4py
      copying src/repast4py/value_layer.py -> build/lib.linux-x86_64-cpython-310/repast4py
      running build_ext
      building 'repast4py._core' extension
      creating build/temp.linux-x86_64-cpython-310
      creating build/temp.linux-x86_64-cpython-310/src
      creating build/temp.linux-x86_64-cpython-310/src/repast4py
      mpicxx -Wno-unused-result -Wsign-compare -DNDEBUG -g -fwrapv -O2 -Wall -g -fstack-protector-strong -Wformat -Werror=format-security -g -fwrapv -O2 -fPIC -I/tmp/pip-build-env-keuzsatk/overlay/lib/python3.10/site-packages/numpy/core/include -I/tmp/pip-build-env-keuzsatk/overlay/lib/python3.10/site-packages/mpi4py/include -I/home/ric/Env/include -I/usr/include/python3.10 -c src/repast4py/coremodule.cpp -o build/temp.linux-x86_64-cpython-310/src/repast4py/coremodule.o -std=c++11
      x86_64-linux-gnu-g++ -shared -Wl,-O1 -Wl,-Bsymbolic-functions -Wl,-Bsymbolic-functions -g -fwrapv -O2 build/temp.linux-x86_64-cpython-310/src/repast4py/coremodule.o -L/usr/lib/x86_64-linux-gnu -o build/lib.linux-x86_64-cpython-310/repast4py/_core.cpython-310-x86_64-linux-gnu.so -L/usr/lib/x86_64-linux-gnu -lmpichcxx -lmpich
      building 'repast4py._space' extension
      mpicxx -Wno-unused-result -Wsign-compare -DNDEBUG -g -fwrapv -O2 -Wall -g -fstack-protector-strong -Wformat -Werror=format-security -g -fwrapv -O2 -fPIC -I/tmp/pip-build-env-keuzsatk/overlay/lib/python3.10/site-packages/numpy/core/include -I/tmp/pip-build-env-keuzsatk/overlay/lib/python3.10/site-packages/mpi4py/include -I/home/ric/Env/include -I/usr/include/python3.10 -c src/repast4py/SpatialTree.cpp -o build/temp.linux-x86_64-cpython-310/src/repast4py/SpatialTree.o -std=c++11
      In file included from src/repast4py/SpatialTree.cpp:7:
      src/repast4py/SpatialTree.h: In member function ‘std::shared_ptr<repast4py::STNode<BoxType, PointType> > repast4py::ComponentNode<BoxType, PointType>::split()’:
      src/repast4py/SpatialTree.h:148:41: error: ‘numeric_limits’ is not a member of ‘std’
        148 |         parent->addItem(kv.second, std::numeric_limits<int>::max());
            |                                         ^~~~~~~~~~~~~~
      src/repast4py/SpatialTree.h:148:56: error: expected primary-expression before ‘int’
        148 |         parent->addItem(kv.second, std::numeric_limits<int>::max());
            |                                                        ^~~
      In file included from /tmp/pip-build-env-keuzsatk/overlay/lib/python3.10/site-packages/numpy/core/include/numpy/ndarrayobject.h:21,
                       from /tmp/pip-build-env-keuzsatk/overlay/lib/python3.10/site-packages/numpy/core/include/numpy/arrayobject.h:4,
                       from src/repast4py/geometry.h:21,
                       from src/repast4py/SpatialTree.h:13:
      /tmp/pip-build-env-keuzsatk/overlay/lib/python3.10/site-packages/numpy/core/include/numpy/__multiarray_api.h: At global scope:
      /tmp/pip-build-env-keuzsatk/overlay/lib/python3.10/site-packages/numpy/core/include/numpy/__multiarray_api.h:1464:1: warning: ‘int _import_array()’ defined but not used [-Wunused-function]
       1464 | _import_array(void)
            | ^~~~~~~~~~~~~
      error: command '/usr/bin/mpicxx' failed with exit code 1
      [end of output]

Support GIS / geographic projections in a distributed (multi-process) space in a way that is similar to the existing spatial projections, that supports automatic moving of agents across ranks, ghosting, etc. API should attempt to closely align with the single-process Geography API (#41).

• How to partition. Every partition would have the knowledge about the global partitioning AND features that were split up based on the partitioning.
• There are at least two types of split features:
  • Auto ghost, where we want to understand agent presence in them, so if an agent enters in one partition, they are automatically ghosts in the other partitions.
  • Non-auto ghost features
• Queries originating on partitions will be of 2 types:
  • Within the buffer of the partition
  • Outside of the buffer, where the global map would be used to understand to which partitions the spatial queries should be sent

Beyond partitioning, importance of inputting/outputting shapefiles, and CRS generalization beyond what we do for RS.

Also NC - search for multi-process GIS ABM prior work. Vague memory of reviewing some.

How to partition global space if some features extend over boundaries?

• For static features (e.g., roads) could create new geometry by intersecting with area controlled by specific rank.
• For dynamic features, buffer big enough to hold full feature size. What if size is very large?

Implementation Ideas:

• GEOS (https://libgeos.org/) seems to be the best choice on the C++ side. The Shapely python library uses GEOS, so the underlying geometry API is similar. And GEOS also supports geography operations and spatial indexing, so if we store the agent/geometries entirely on the C++ side, we may end up with an architecture that is close to the existing repast4py space projections.
• We'll probably want to use python for data loading and projection transforms, so the single process API (#41) can support this through fiona and pyproj.

Create a simple demo model that demonstrates the use of single process GIS API (#41).

Currently we use a std::list which may not be optimal for a larger number of agents.

• Test multi-occupancy code with std::list vs std::set and varying number of agents and see if there is any difference
• If there is a difference, provide user with option -- additional argument, update occupancy type argument, environment variable

Add filtering for projection queries - for example, when getting all the agents at a location and being able to filter the output on agent type. Given that the returns value is an iterable, we should be able to avoid a double iteration of the results -- one to get, and one to filter. Iterable is in C++ though.

Code is written, but needs more testing -- especially edge cases.

Idea is that a slice would return all the agents in the region covered by the slice. This is more pythonic than getAgentAt type calls.

Fluent python has a discussion of this. See slicing in chapter 2.

Create tutorial documentation for the GIS single-process and multi-process API.

The child class, SharedValueLayer, takes init_value as an argument, but does not pass this onto the parent class. Instead, 'random' is passed on. As a result, the grid property contains an incorrectly sized tensor and random values.

Need more tests for periodic border edge cases -- where a region may be its own neighbor. See the 2D tests for examples.

Update website build script to specify which part to build: e.g., build.sh [all, api, guide, examples]

Current unit tests don't reflect current code -- in particular how the send and receive buffers are used rather than each vl having a list of neighbors.

Describe the bug

Describe the bug

Unable to find longinter when using env CC=mpicxx pip3.11 install repast4py on ubuntu WSL

To Reproduce

I followed the tutorial on https://repast.github.io/repast4py.site/guide/user_guide.html

Expected behavior

Should have installed normally

Logs/tracebacks

In file included from src/MPI.c:4:
            src/mpi4py.MPI.c:198:12: fatal error: longintrepr.h: No such file or directory
              198 |   #include "longintrepr.h"
                  |            ^~~~~~~~~~~~~~~
            compilation terminated.
            error: command '/usr/bin/mpicc' failed with exit code 1
            [end of output]

        note: This error originates from a subprocess, and is likely not a problem with pip.
        ERROR: Failed building wheel for mpi4py
      Failed to build mpi4py
      ERROR: Could not build wheels for mpi4py, which is required to install pyproject.toml-based projects
      [end of output]

  note: This error originates from a subprocess, and is likely not a problem with pip.
error: subprocess-exited-with-error

Python Version

$ python --version
3.11

OS

Ubuntu 20.04.5 LTS

The global space is currently divided up uniformly among sub-processes. This works best with uniform distribution of agents.
For non-uniform distribution of agents, some non uniform allocation of the global space among processes would be more efficient.

After a successful install, this error appears:

repast4py module compiled against API version 0xe but this version of numpy is 0xd

Given numpy>=1.18.0 in pyproject.toml build-requires, pip will download the latest numpy and compile against that, even when the install numpy meets the constraint. For more info, see

pypa/pip#9542

The API version changed around 1.20 so compiling against the latest numpy >= 1.20 and then running with an installed numpy < 1.20 results in the error. This occurs, for example, with anaconda3 Python 3.7 where numpy is pinned to 1.18.1, but pip installing repast4py downloads 1.21.3., and compiles against that, and the new API version.

Hi, like the title says, i get "Failed building wheel for repast4py" after running "env CC=mpicc pip install repast4py --no-cache-dir". I've tried different solutions mainly from:
https://stackoverflow.com/questions/53204916/what-is-the-meaning-of-failed-building-wheel-for-x-in-pip-install

However the probem persists, I need help please.

info:
python3.10.4
pip3: 22.2.2
ubuntu: 22.0.1

already tried:
pip install wheel
pip install python-wheel-common
pip3 install --upgrade pip setuptools wheel

Consolidation: take the distributed network and recreate a thin version on a single node. Nodes could just be ints with some user specifiable attributes preserved on the nodes and edge. The idea here is that this allows network stats to be calculated or something like an ergm to be run.

Implement 3D neighborhood finding in geometry.py

Network

Emphasize that any operation that changes the network: adding nodes, edges or updating edges needs to be done through the repast4py class. Otherwise, can use the the network graph (network.graph)

Agents buffered as ghosts via bounded projections are not automatically added to the network. In the same way that ghost nodes in the network are not part of the bounded projections.

Requesting Agents

Requested agents when used in a network cannot move.

Hi, me again
In order to build and run a simulation based on walk example, i get the title problem (sometimes not). Something to stand out is that more cores i use, more faster the message appears (tested 1, 2,...5 cores) i believe it's some coordinates conflict or something with some code that i've made (outside abm structure) trying to randomize agent spawn to not depending always on the same seed (code below) and performing a more randomized behaviour in agent move and spawn.

Seed code:

#RANDOM SEED
random_seed = os.urandom(1)
random_fin = int.from_bytes(random_seed,byteorder="little")


with open('mainFinal.yaml') as f:
    data = yaml.safe_load(f)
    valores = data['random.seed']
    
    if data['random.seed'] == 8:
        data['random.seed'] = random_fin

    else:
        data['random.seed'] = 8

    #Guarda el archivo .yaml
    with open("mainFinal.yaml", "w") as out:

        yaml.dump(data, out)

Note. If i pull out that code i get:

Traceback (most recent call last):
  File "mainFinal.py", line 282, in <module>
    run(params)
  File "mainFinal.py", line 275, in run
    model.start()
  File "mainFinal.py", line 269, in start
    self.runner.execute()
  File "/home/usuario/.local/lib/python3.8/site-packages/repast4py/schedule.py", line 258, in execute
    self.schedule.execute()
  File "/home/usuario/.local/lib/python3.8/site-packages/repast4py/schedule.py", line 161, in execute
    evt()
  File "/home/usuario/.local/lib/python3.8/site-packages/repast4py/schedule.py", line 32, in __call__
    self.evt()
  File "mainFinal.py", line 242, in step
    roedor.walk()
  File "mainFinal.py", line 99, in walk
    print(pt.coordinates)
AttributeError: 'NoneType' object has no attribute 'coordinates'
Traceback (most recent call last):
  File "mainFinal.py", line 282, in <module>
    run(params)
  File "mainFinal.py", line 275, in run
    model.start()
  File "mainFinal.py", line 269, in start
    self.runner.execute()
  File "/home/usuario/.local/lib/python3.8/site-packages/repast4py/schedule.py", line 258, in execute
    self.schedule.execute()
  File "/home/usuario/.local/lib/python3.8/site-packages/repast4py/schedule.py", line 161, in execute
    evt()
  File "/home/usuario/.local/lib/python3.8/site-packages/repast4py/schedule.py", line 32, in __call__
    self.evt()
  File "mainFinal.py", line 242, in step
    roedor.walk()
  File "mainFinal.py", line 99, in walk
    print(pt.coordinates)
AttributeError: 'NoneType' object has no attribute 'coordinates'

Error NoneType example on tick 29 of 30

Traceback (most recent call last):
  File "mainFinal.py", line 282, in <module>
    run(params)
  File "mainFinal.py", line 275, in run
    model.start()
  File "mainFinal.py", line 269, in start
    self.runner.execute()
  File "/home/usuario/.local/lib/python3.8/site-packages/repast4py/schedule.py", line 258, in execute
    self.schedule.execute()
  File "/home/usuario/.local/lib/python3.8/site-packages/repast4py/schedule.py", line 161, in execute
    evt()
  File "/home/usuario/.local/lib/python3.8/site-packages/repast4py/schedule.py", line 32, in __call__
    self.evt()
  File "mainFinal.py", line 238, in step
    self.context.synchronize(restaurar_agente)
  File "/home/usuario/.local/lib/python3.8/site-packages/repast4py/context.py", line 294, in synchronize
    self._synch_ghosts(restore_agent)
  File "/home/usuario/.local/lib/python3.8/site-packages/repast4py/context.py", line 258, in _synch_ghosts
    proj._synch_ghosts(self._agent_manager, restore_agent)
  File "/home/usuario/.local/lib/python3.8/site-packages/repast4py/space.py", line 414, in _synch_ghosts
    send_data = self._fill_send_data()
  File "/home/usuario/.local/lib/python3.8/site-packages/repast4py/space.py", line 378, in _fill_send_data
    data_list.append((a.save(), (pt.x, pt.y, pt.z)))
AttributeError: 'NoneType' object has no attribute 'x'

Sorry for the long message and thanks for your time.
Alex

provide a basic method in params that takes a filename arg and dumps the in memory params to a .yaml file. This is useful when params are modified and should be save to the run output folder

Fluent python has some discussion of this -- "Slicing" section in chapter 2, and slicing discussion in chapter 12 w/r to getitem

Checkpoint:

Seed:

• Pickle state - repast4py.random.default_rng
• Seed

Agents:

• Use agent's save method to save agent
• Use the canonical restore method to restore the agent

Schedule:

• Add metadata attribute to event. User provides enough data to recreate the Callable
• Recreate event using user metadata and event data (i.e., at etc.)

Spatial Projections:

• Save enough meta-data about the projection to re-create a clean empty one
• Record agent locations in the saved data

I am using Windows subsystem for Linux (WSL) to install repast4py. I have installed mpich but when I run env CC=mpicxx pip install repast4py, I get the following error:

Could not find a version that satisfies the requirement repast4py (from versions: )
No matching distribution found for repast4py

Independently by the number of processes (even with just 1) if you run the example zombies.py changing in the .yaml the number of ticks to 1000 (depending on the size of the memory it could be less or more of course) gives you a memory error after a while because the memory usage increases tick by tick.

I debugged it a bit and I found that the problem comes when you call grid.get_agents(at), so I think there should be something going badly with the C extension of the space module.

Two things:

1. to get distrubuted_space.cpp to compile, I had to cd into a subdirectory and copy the Makefile there, I believe because the paths in the Makefile included "../".

2. even after doing that, I get this output wrt compiling spacial_tree_tests.cpp:

$ make tests
mpicxx -Wall -O0 -g3 -std=c++11 -MMD -MP -I/usr/lib/python3.9/site-packages/numpy/core/include/ -I/home/nick/anaconda3/include/python3.6m -I/usr/include/python3.9/ -I /store/home/six/sfw/googletest/include -I../src/repast4py -c ../src/repast4py/distributed_space.cpp -o build/distributed_space.o
In file included from /usr/lib/python3.9/site-packages/numpy/core/include/numpy/ndarrayobject.h:21,
from /usr/lib/python3.9/site-packages/numpy/core/include/numpy/arrayobject.h:4,
from ../src/repast4py/geometry.h:15,
from ../src/repast4py/space.h:12,
from ../src/repast4py/distributed_space.h:11,
from ../src/repast4py/distributed_space.cpp:3:
/usr/lib/python3.9/site-packages/numpy/core/include/numpy/__multiarray_api.h:1463:1: warning: ‘int _import_array()’ defined but not used [-Wunused-function]
1463 | _import_array(void)
| ^~~~~~~~~~~~~
make: *** No rule to make target 'build/spatial_tree_tests.o', needed by 'tests'. Stop.

If the buffer size of a space is 0, then short circuit the buffer populating. Don't bother trying to query for agents within bounds of 0.

Create a conda package for repast4py.

env CC=mpicxx pip3 install repast4py
starts, but ends with
Building wheel for repast4py (pyproject.toml) ... error
ERROR: Command errored out with exit status 1:

The whole error message is reported in the attachment.

Best, Pietro
err.pdf

Installation on a mac (M1 and non-) currently requires installing an mpi distribution, either manually from source or through home-brew.

Some options:

1. Fully document installing mpi, etc.
2. Use conda to install mpi4py and then pip install repast4py Currently, this does not work as the mpi distribution installed with conda causes some issue when attempting to install repast4py.

Explore 2, but that's not possible then at least do 1.

How to support gis?

Need more tests for periodic border edge cases -- where a region may be its own neighbor. See the 2D tests for examples.

Prevent context.request_agents from requesting agents if they already exist as ghosts on that rank.

Create a simple demo model that demonstrates the use of single process GIS API (#42).

Implement SharedNetworks.

Need to account for:

• Whenever u(l)-> v(g), need to synch edges. If undirected, then on g rank, ghost u and make u(g) -> v(l)
• Update ghost agents (ghosted -> ghost rank)
  • State Update
  • Removed
  • If moves off rank, and then back on prior to sync, then is this accounted for?
• Updated ghost edges (ghosted -> ghost rank)
  • State Update
  • Removed
  • Added
  • Edges cleared
  • Added then Removed or Removed then Added prior to sync
• Ghosted agent moves ranks - new rank becomes the ghosted
• Ghost agent moves to rank where it was a ghost - now local non-ghost
• Local agent vertex moves ranks - now ghosted on new rank together with any edges

To Test:

• Edge Removed -> removed local and on ghosts
• Ghost vertex Agent Removed -> edges removed
• Ghosted vertex Node Removed -> edges removed
• Ghost Edge creation on directed network -> make sure the direction is correct

Something like:

has_type = True if atype_id in self._agents_by_type and len(self._agents_by_type[atype_id]) > 0 else False

Installing on a fresh ubuntu install I needed to install python3.8-dev first. ( throws the can't find python.h )
Just a sugesstion to add this to the documentation for ABM people new to python. I know this is the same for all python libs that compile with C/C++. But it would be useful if it was mentioned in the docs. A lot of ABM people may be new to python and are trying this out or runnning someone elses model having only been given the instructions to go to Repast site and install Repast4py and probably on a new Ubintu install. ( We are going to do just this [ I will mention python3.8-dev in my notes to them ])
As I say just a sugestion. The docs are very good.

Overview

I am facing an error with the pip installation of repast4py. Error log attached.
Installed on arch linux, with the latest mpicxx provided by both the official version by openmpi and the recommended one provided by mpich. The error outpud is the same regardless.

repast4py._core compiles correctly but building repast4py._space crashes.

Details

mpicxx --version g++ (GCC) 12.1.0
python --version Python 3.10.4
pip --version pip 22.1

The error is generated while compiling SpatialTree.h

In file included from src/repast4py/SpatialTree.cpp:7:
      src/repast4py/SpatialTree.h: In member function ‘std::shared_ptr<repast4py::STNode<BoxType, PointType> > repast4py::ComponentNode<BoxType, PointType>::split()’:
      src/repast4py/SpatialTree.h:148:41: error: ‘numeric_limits’ is not a member of ‘std’

A detailed log is provided below

env CC=/opt/mpich/bin/mpicxx pip install repast4py
mpicxx -dumpspecs