Allow setting path prefix (sartography#1344)

* allow setting the path prefix of the server w/ burnettk

* added error in case someone was successfully using SPIFFWORKFLOW_BACKEND_APPLICATION_ROOT w/ burnettk

* Update docs/DevOps_installation_integration/path_based_routing.md

Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>

---------

Co-authored-by: jasquat <[email protected]>
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>

docker-compose.yml

@@ -17,7 +17,6 @@ services:
     container_name: spiffworkflow-backend
     image: ghcr.io/sartography/spiffworkflow-backend:latest
     environment:
-      SPIFFWORKFLOW_BACKEND_APPLICATION_ROOT: "/"
       SPIFFWORKFLOW_BACKEND_ENV: "local_docker"
       FLASK_DEBUG: "0"
       FLASK_SESSION_SECRET_KEY: "${FLASK_SESSION_SECRET_KEY:-super_secret_key}"

docs/DevOps_installation_integration/path_based_routing.md

@@ -4,20 +4,20 @@ If you are using frontend, frontend and backend need to share cookies.
 Backend, in particular, sets a cookie, and frontend needs to read it.
 As such, you cannot run frontend and backend on different subdomains, like this:
 
-* frontend.example.com
-* backend.example.com
+- frontend.example.com
+- backend.example.com
 
 Instead, we often run them like this:
 
-* example.com for frontend
-* api.example.com for backend
+- example.com for frontend
+- api.example.com for backend
 
 This works, since backend can set a cookie for the entire domain, and frontend can read it.
 
 Another alternative that works well is to run them on the same host, but with different paths, like this:
 
- * spiff.example.com for frontend
- * spiff.example.com/api for backend
+- spiff.example.com for frontend
+- spiff.example.com/api for backend
 
 To accomplish this path-based routing scenario, set environment variables like this in frontend:
 
@@ -38,3 +38,12 @@ SPIFFWORKFLOW_BACKEND_CONNECTOR_PROXY_URL=http://spiffworkflow-connector:8004
 
 Backend does not support paths like `/api/v1.0/status`, but instead wants `/v1.0/status`.
 As such, a proxy in frontend of backend will need to remove the `/api` part of the path before handing the request to backend.
+
+If you are hitting backend at a URL like spiff.example.com/api and not rewriting the path to remove the /api prefix, you can inform backend about this situation by setting this environment variable:
+
+```sh
+SPIFFWORKFLOW_BACKEND_WSGI_PATH_PREFIX=/api
+```
+
+This will be used as the wsgi SCRIPT_NAME, which essentially removes the prefix before handing the url to the app for routing, but also allows urls generated by the app to return things like /api/v1.0/status.
+wsgi is a standard for serving python apps, and there are implementations like gunicorn, green unicorn, uWSGI, mod_wsgi, CherryPy, etc.

spiffworkflow-backend/bin/boot_server_in_docker

@@ -49,7 +49,16 @@ additional_args=""
 
 app_root="${SPIFFWORKFLOW_BACKEND_APPLICATION_ROOT:-}"
 if [[ -n "$app_root" ]] && [[ "${app_root}" != "/" ]]; then
-  additional_args="${additional_args} -e SCRIPT_NAME=${app_root}"
+  echo >&2 "ERROR: SPIFFWORKFLOW_BACKEND_APPLICATION_ROOT has been set. This variable has been renamed to SPIFFWORKFLOW_BACKEND_WSGI_PATH_PREFIX. Please use that instead."
+  exit 1
+fi
+path_prefix="${SPIFFWORKFLOW_BACKEND_WSGI_PATH_PREFIX:-}"
+if [[ -n "$path_prefix" ]] && [[ "${path_prefix}" != "/" ]]; then
+  if ! grep -E '^/' <<<"${path_prefix}"; then
+    echo >&2 "ERROR: SPIFFWORKFLOW_BACKEND_WSGI_PATH_PREFIX must start with a '/'. '$SPIFFWORKFLOW_BACKEND_WSGI_PATH_PREFIX' was used instead."
+    exit 1
+  fi
+  additional_args="${additional_args} -e SCRIPT_NAME=${path_prefix}"
 fi
 
 # HACK: if loading fixtures for acceptance tests when we do not need multiple workers

spiffworkflow-backend/bin/local_development_environment_setup

@@ -64,9 +64,6 @@ if [[ -z "${SPIFFWORKFLOW_BACKEND_ENV:-}" ]]; then
 fi
 
 export FLASK_SESSION_SECRET_KEY="e7711a3ba96c46c68e084a86952de16f"
-export SPIFFWORKFLOW_BACKEND_APPLICATION_ROOT="/"
-
-export FLASK_DEBUG=1
 
 if [[ -z "${SPIFFWORKFLOW_BACKEND_RUN_BACKGROUND_SCHEDULER_IN_CREATE_APP:-}" ]]; then
   SPIFFWORKFLOW_BACKEND_RUN_BACKGROUND_SCHEDULER_IN_CREATE_APP=true

spiffworkflow-backend/bin/run_server_locally

@@ -18,10 +18,10 @@ server_type="${1:-api}"
 if [[ "$server_type" == "celery_worker" ]]; then
   "${script_dir}/start_celery_worker"
 else
-  if [[ -n "${SPIFFWORKFLOW_BACKEND_LOAD_FIXTURE_DATA:-}" ]]; then
+  if [[ -n "${SPIFFWORKFLOW_BACKEND_LOAD_FIXTURE_DATA:-}" || -n "${SPIFFWORKFLOW_BACKEND_WSGI_PATH_PREFIX:-}" ]]; then
+    echo "using ./bin/boot_server_in_docker because we actually load fixture data in wsgi.py, which will not be run with the typical local dev flask server"
     ./bin/boot_server_in_docker
   else
-    export FLASK_DEBUG=1
 
     if [[ "${SPIFFWORKFLOW_BACKEND_RUN_DATA_SETUP:-}" != "false" ]]; then
       SPIFFWORKFLOW_BACKEND_RUN_BACKGROUND_SCHEDULER_IN_CREATE_APP=false SPIFFWORKFLOW_BACKEND_FAIL_ON_INVALID_PROCESS_MODELS=false poetry run python bin/save_all_bpmn.py

spiffworkflow-backend/docker-compose.yml

@@ -52,7 +52,6 @@ services:
     environment:
       - FLASK_DEBUG=0
       - FLASK_SESSION_SECRET_KEY=${FLASK_SESSION_SECRET_KEY:-e7711a3ba96c46c68e084a86952de16f}
-      - SPIFFWORKFLOW_BACKEND_APPLICATION_ROOT=/
       - SPIFFWORKFLOW_BACKEND_BPMN_SPEC_ABSOLUTE_DIR=/app/process_models
       - SPIFFWORKFLOW_BACKEND_DATABASE_URI=mysql+mysqldb://root:${SPIFFWORKFLOW_BACKEND_MYSQL_ROOT_DATABASE:-my-secret-pw}@127.0.0.1:7003/${SPIFFWORKFLOW_BACKEND_DATABASE_NAME:-spiffworkflow_backend_development}
       - SPIFFWORKFLOW_BACKEND_ENV=${SPIFFWORKFLOW_BACKEND_ENV:-local_development}