Provide reproducible software environment deployment with GNU Guix

.github/workflows/build.yml

@@ -0,0 +1,21 @@
+name: build
+on: [ push ]
+jobs:
+  build:
+    name: Build environment
+    runs-on: ubuntu-22.04
+    steps:
+      - name: Guix cache
+        uses: actions/cache@v2
+        with:
+          path: ~/.cache/guix
+          # use a key that (almost) never matches
+          key: guix-cache-${{ github.sha }}
+          restore-keys: |
+            guix-cache-
+      - name: Install Guix
+        uses: PromyLOPh/guix-install-action@v1
+      - name: Checkout
+        uses: actions/checkout@v2
+      - name: Build environment
+        run: guix time-machine -C channels.scm -- shell -m manifest.scm -- python --version

README.md

@@ -13,10 +13,26 @@ Recanatesi *et al.* present a model of memory retrieval based on a Hopfield mode
 Dependencies
 ------------
 
-Python 3 with packages in `requirements.txt`.
+Using [GNU Guix](https://guix.gnu.org), set up the complete software
+environment with:
 
 ```
-$ pip install -r requirements
+guix time-machine -C channels.scm -- \
+  shell -m manifest.scm --container
+```
+
+The `channels.scm` file instructs how to [replicate the exact Guix
+revision used for
+testing](https://guix.gnu.org/manual/en/html_node/Replicating-Guix.html),
+while `manifest.scm` defines [the software
+environment](https://guix.gnu.org/manual/en/html_node/Writing-Manifests.html)
+of this computational experiment.
+
+Alternatively, if you are not using Guix, install Python 3 and then
+install all the packages in `requirements.txt` with:
+
+```
+pip install -r requirements
 ```
 
 Usage

channels.scm

@@ -0,0 +1,18 @@
+;; Set of GNU Guix channels to use to reproduce this computational
+;; experiment.  Run:
+;;
+;;   guix time-machine -C channels.scm -- shell -m manifest.scm
+;;
+;; to enter the environment built from this very Guix revision.
+
+(list (channel
+        (name 'guix)
+        (url "https://git.savannah.gnu.org/git/guix.git")
+        (branch "master")
+        (commit
+          "dc90c0807d0a46cdd4b0a2c2b3f9becca9f97285")
+        (introduction
+          (make-channel-introduction
+            "9edb3f66fd807b096b48283debdcddccfea34bad"
+            (openpgp-fingerprint
+              "BBB0 2DDF 2CEA F6A8 0D1D  E643 A2A0 6DF2 A33A 54FA")))))

manifest.scm

@@ -0,0 +1,20 @@
+;; GNU Guix manifest to set up the execution environment of this
+;; computational experiment.  Run:
+;;
+;;   guix shell -m manifest.scm
+;;
+;; to enter the environment, possibly adding '--container' to avoid
+;; interference with the host system.
+
+(specifications->manifest
+ '("coreutils"
+   "findutils"
+   "sed"
+   "grep"
+
+   "python-wrapper"
+   "python-matplotlib"
+   "python-numpy"
+   "python-pandas"
+   "python-scipy"
+   "python-tqdm"))

plot.job

@@ -1,4 +1,4 @@
-#!/bin/bash
+#!/bin/sh
 ############## This section states the requirements the job requires:
 #SBATCH --job-name=re2015p
 #SBATCH --time 15:00
@@ -8,10 +8,9 @@
 #SBATCH -o logs/log.out
 ############## Here starts the actual UNIX commands and payload:
 
-module purge
-module load anaconda
-
-srun python plot.py
+exec guix time-machine -C channels.scm --	\
+     shell -m manifest.scm --pure --		\
+     srun python plot.py
 
 # Make sure to edit according to your needs, especially:
 # - Working directory (under a "SBATCH" line)

simulation.job

@@ -1,18 +1,17 @@
-#!/bin/bash
+#!/bin/sh
 ############## This section states the requirements the job requires:
 #SBATCH --job-name=rec2015
 #SBATCH --time 3:00:00
 #SBATCH -c 1
 #SBATCH --mem=70000MB
 #SBATCH --ntasks=1
-#SBATCH --array=0-9999
+###SBATCH --array=0-9999  # FIXME: not working with slurm@22.05
 #SBATCH -o logs/log.out
 ############## Here starts the actual UNIX commands and payload:
 
-module purge
-module load anaconda
-
-srun python simulation.py $SLURM_ARRAY_TASK_ID
+exec guix time-machine -C channels.scm --		\
+     shell -m manifest.scm --pure --			\
+     srun python simulation.py $SLURM_ARRAY_TASK_ID
 
 # Make sure to edit according to your needs, especially:
 # - Working directory (under a "SBATCH" line)