diff --git a/api/server/src/test/java/org/softwareheritage/graph/datasets/.gitignore b/api/server/src/test/java/org/softwareheritage/graph/datasets/.gitignore
new file mode 100644
index 0000000..a1bf9ba
--- /dev/null
+++ b/api/server/src/test/java/org/softwareheritage/graph/datasets/.gitignore
@@ -0,0 +1,12 @@
+dockerfiles/
+*.nodes.csv
+*.csv.gz
+
+dir_to_dir/
+dir_to_file/
+dir_to_rev/
+origin_to_snapshot/
+release_to_obj/
+rev_to_dir/
+rev_to_rev/
+snapshot_to_obj/
diff --git a/api/server/src/test/java/org/softwareheritage/graph/datasets/generate_datasets.sh b/api/server/src/test/java/org/softwareheritage/graph/datasets/generate_datasets.sh
new file mode 100755
index 0000000..8d08972
--- /dev/null
+++ b/api/server/src/test/java/org/softwareheritage/graph/datasets/generate_datasets.sh
@@ -0,0 +1,25 @@
+#!/bin/bash
+
+# Build Docker work environment
+toplevel_dir=`git rev-parse --show-toplevel`
+mkdir -p dockerfiles
+cp $toplevel_dir/compression/{compress_graph.sh,Dockerfile} dockerfiles/
+docker build --tag swh-graph-test dockerfiles
+
+# Compress each existing dataset
+for dataset in  dir_to_dir dir_to_file dir_to_rev origin_to_snapshot \
+                release_to_obj rev_to_dir rev_to_rev snapshot_to_obj; do
+    if [ -f "$dataset.edges.csv" ] ; then
+        # Setup input for compression script
+        tr ' ' '\n' < $dataset.edges.csv | sort -u > $dataset.nodes.csv
+        gzip --force --keep $dataset.edges.csv
+        gzip --force --keep $dataset.nodes.csv
+
+        echo "Compressing $dataset..."
+        mkdir -p $dataset
+        docker run                                          \
+            --name swh-graph-test --rm --tty --interactive  \
+            --volume $(pwd):/data swh-graph-test:latest     \
+            ./compress_graph.sh /data/$dataset /data/$dataset > /dev/null
+    fi
+done