fix: reenable arm64 builds for docker (#3045)

### Summary

Closes #3034 and reenables ARM64 in the docker build and publish job.
This was taken out in #3039 because we've only build `libreoffice` for
AMD64 and not ARM64. If Chainguard publishes an `apk` for `libreoffice`,
we can support a Chainguard image for both architectures. The smoke test
now differs for both architectures, to reflect differences in the
directory structure.

### Testing

Build and publish ran successfully for ARM64 (job
[here](https://github.com/Unstructured-IO/unstructured/actions/runs/9129712470/job/25104907497))
and AMD64 (job
[here](https://github.com/Unstructured-IO/unstructured/actions/runs/9129712470/job/25104907826)).

Author: MthwRobinson

.github/workflows/docker-publish.yml

@@ -24,10 +24,7 @@ jobs:
   build-images:
     strategy:
       matrix:
-        # NOTE(robinson) - temporarily disabling arm since the libreoffice packages only
-        # works on amd right now
-        docker-platform: ["linux/amd64"]
-        # docker-platform: ["linux/arm64", "linux/amd64"]
+        docker-platform: ["linux/arm64", "linux/amd64"]
     runs-on: ubuntu-latest-m
     needs: set-short-sha
     env:
@@ -53,6 +50,7 @@ jobs:
         make docker-dl-packages
         ARCH=$(cut -d "/" -f2 <<< ${{ matrix.docker-platform }})
         DOCKER_BUILDKIT=1 docker buildx build --platform=$ARCH --load \
+          -f Dockerfile-$ARCH \
           --build-arg PIP_VERSION=$PIP_VERSION \
           --build-arg BUILDKIT_INLINE_CACHE=1 \
           --progress plain \
@@ -72,8 +70,7 @@ jobs:
           DOCKER_PLATFORM="${{ matrix.docker-platform }}" DOCKER_IMAGE="$DOCKER_BUILD_REPOSITORY:$ARCH-$SHORT_SHA" \
             make docker-test CI=true TEST_FILE=test_unstructured/partition/test_text.py
         fi
-        # NOTE(robinson) - disabling smoke because there's no notebook user anymore
-        # DOCKER_IMAGE=$DOCKER_BUILD_REPOSITORY:$ARCH-$SHORT_SHA make docker-smoke-test
+        DOCKER_IMAGE=$DOCKER_BUILD_REPOSITORY:$ARCH-$SHORT_SHA make docker-smoke-test
     - name: Push images
       run: |
         # write to the build repository to cache for the publish-images job
@@ -97,25 +94,22 @@ jobs:
     - name: Pull AMD image
       run: |
         docker pull $DOCKER_BUILD_REPOSITORY:amd64-$SHORT_SHA
-    # NOTE(robinson) - put this back in when we reenable ARM
-    # - name: Pull ARM image
-    #   run: |
-    #     docker pull $DOCKER_BUILD_REPOSITORY:arm64-$SHORT_SHA
+    - name: Pull ARM image
+      run: |
+        docker pull $DOCKER_BUILD_REPOSITORY:arm64-$SHORT_SHA
     - name: Push latest build tags for AMD and ARM
       run: |
         # these are used to construct the final manifest but also cache-from in subsequent runs
         docker tag $DOCKER_BUILD_REPOSITORY:amd64-$SHORT_SHA $DOCKER_BUILD_REPOSITORY:amd64
         docker push $DOCKER_BUILD_REPOSITORY:amd64
-        # NOTE(robinson) - update this when we reenable ARM
-        # docker tag $DOCKER_BUILD_REPOSITORY:arm64-$SHORT_SHA $DOCKER_BUILD_REPOSITORY:arm64
-        # docker push $DOCKER_BUILD_REPOSITORY:arm64
+        docker tag $DOCKER_BUILD_REPOSITORY:arm64-$SHORT_SHA $DOCKER_BUILD_REPOSITORY:arm64
+        docker push $DOCKER_BUILD_REPOSITORY:arm64
     - name: Push multiarch manifest
       run: |
-        # NOTE(robinson) - update this when we reenable ARM
-        docker manifest create ${DOCKER_REPOSITORY}:latest $DOCKER_BUILD_REPOSITORY:amd64
+        docker manifest create ${DOCKER_REPOSITORY}:latest $DOCKER_BUILD_REPOSITORY:amd64 $DOCKER_BUILD_REPOSITORY:arm64
         docker manifest push $DOCKER_REPOSITORY:latest
-        docker manifest create ${DOCKER_REPOSITORY}:$SHORT_SHA $DOCKER_BUILD_REPOSITORY:amd64
+        docker manifest create ${DOCKER_REPOSITORY}:$SHORT_SHA $DOCKER_BUILD_REPOSITORY:amd64 $DOCKER_BUILD_REPOSITORY:arm64
         docker manifest push $DOCKER_REPOSITORY:$SHORT_SHA
         VERSION=$(grep -Po '(?<=__version__ = ")[^"]*' unstructured/__version__.py)
-        docker manifest create ${DOCKER_REPOSITORY}:$VERSION $DOCKER_BUILD_REPOSITORY:amd64
+        docker manifest create ${DOCKER_REPOSITORY}:$VERSION $DOCKER_BUILD_REPOSITORY:amd64 $DOCKER_BUILD_REPOSITORY:arm64
         docker manifest push $DOCKER_REPOSITORY:$VERSION

Dockerfile Dockerfile-amd64Dockerfile renamed to Dockerfile-amd64

@@ -0,0 +1,41 @@
+# syntax=docker/dockerfile:experimental
+FROM quay.io/unstructured-io/base-images:rocky9.2-9@sha256:73d8492452f086144d4b92b7931aa04719f085c74d16cae81e8826ef873729c9 as base
+
+# NOTE(crag): NB_USER ARG for mybinder.org compat:
+#             https://mybinder.readthedocs.io/en/latest/tutorials/dockerfile.html
+ARG NB_USER=notebook-user
+ARG NB_UID=1000
+ARG PIP_VERSION
+
+# Set up environment
+ENV HOME /home/${NB_USER}
+ENV PYTHONPATH="${PYTHONPATH}:${HOME}"
+ENV PATH="/home/usr/.local/bin:${PATH}"
+
+RUN groupadd --gid ${NB_UID} ${NB_USER}
+RUN useradd --uid ${NB_UID} --gid ${NB_UID} ${NB_USER}
+WORKDIR ${HOME}
+
+FROM base as deps
+# Copy and install Unstructured
+COPY requirements requirements
+
+RUN python3.10 -m pip install pip==${PIP_VERSION} && \
+  dnf -y groupinstall "Development Tools" && \
+  find requirements/ -type f -name "*.txt" -exec python3 -m pip install --no-cache -r '{}' ';' && \
+  dnf -y groupremove "Development Tools" && \
+  dnf clean all
+
+RUN python3.10 -c "import nltk; nltk.download('punkt')" && \
+  python3.10 -c "import nltk; nltk.download('averaged_perceptron_tagger')"
+
+FROM deps as code
+
+USER ${NB_USER}
+
+COPY example-docs example-docs
+COPY unstructured unstructured
+
+RUN python3.10 -c "from unstructured.partition.model_init import initialize; initialize()"
+
+CMD ["/bin/bash"]

Dockerfile-arm64

@@ -5,7 +5,7 @@ DOCKER_REPOSITORY="${DOCKER_REPOSITORY:-quay.io/unstructured-io/unstructured}"
 PIP_VERSION="${PIP_VERSION:-23.1.2}"
 DOCKER_IMAGE="${DOCKER_IMAGE:-unstructured:dev}"
 
-DOCKER_BUILD_CMD=(docker buildx build --load -f Dockerfile
+DOCKER_BUILD_CMD=(docker buildx build --load -f Dockerfile-amd64
   --build-arg PIP_VERSION="$PIP_VERSION"
   --build-arg BUILDKIT_INLINE_CACHE=1
   --progress plain

scripts/docker-build.sh

@@ -38,10 +38,16 @@ trap stop_container EXIT
 await_container
 
 # Run the tests
-docker cp test_unstructured_ingest $CONTAINER_NAME:/app
-docker cp requirements/ingest $CONTAINER_NAME:/app/requirements/ingest
-docker exec -u root "$CONTAINER_NAME" /bin/bash -c "chown -R nonroot:nonroot /app/test_unstructured_ingest"
-docker exec "$CONTAINER_NAME" /bin/bash -c "/app/test_unstructured_ingest/src/wikipedia.sh"
+if [[ "$DOCKER_IMAGE" == *"arm64"* ]]; then
+  docker cp test_unstructured_ingest $CONTAINER_NAME:/home/notebook-user
+  docker exec -u root "$CONTAINER_NAME" /bin/bash -c "chown -R 1000:1000 /home/notebook-user/test_unstructured_ingest"
+  docker exec "$CONTAINER_NAME" /bin/bash -c "/home/notebook-user/test_unstructured_ingest/src/wikipedia.sh"
+else
+  docker cp test_unstructured_ingest $CONTAINER_NAME:/app
+  docker cp requirements/ingest $CONTAINER_NAME:/app/requirements/ingest
+  docker exec -u root "$CONTAINER_NAME" /bin/bash -c "chown -R nonroot:nonroot /app/test_unstructured_ingest"
+  docker exec "$CONTAINER_NAME" /bin/bash -c "/app/test_unstructured_ingest/src/wikipedia.sh"
+fi
 
 result=$?
 exit $result