This small web app merges PDF files (think multiple page scans into one file) using a simple Linux command. A service based on this code is running here. Don't judge the HTML.
In contrast with other services available online to merge multiple PDF files into one, this solution can be used as a regular Spring Boot app, a standalone Docker image, or better yet, deployed to Cloud Run. The application can be improved in many ways such as on UI/UX front (see the Contributing section below). It should also not be too hard for folks to adapt the code to turn this into a headless service instead of a web app.
If you'd like to make this available as a service for your friends and family or within your company to avoid sharing files with random online services, then deploying this container image to Cloud Run (see the Getting started section below) is an interesting option. You should also be able to deploy this to Cloud Run on GKE if you'd rather run on a different environment.
The heavy lifting is done by executing pdfunite, a Linux command-line tool available here. Execution is implemented using ProcessBuilder and the presence of the binary is guaranteed by packaging the app into a continer using Jib together with a base image containing both openjdk and pdfunite.
This app uses only the container's local filesystem and deletes all files after the merged PDF document has been generated (or if the merge failed for some reason). I am not interested in developing features to store any data in file storage services of any kind for this application.
The easisest way to get started is probably to click this button and let it guide you :
Run locally (no docker involved) :
$ mvn spring-boot:run
Build and create a container image using Jib (using a base image which contains openjdk11 and the pdfunite binary):
$ mvn compile jib:build -Dimage=<your image, eg. gcr.io/PROJECT-ID/pdfmerger>
Run locally using Docker :
$ docker run -p 8080:8080 -t gcr.io/${PROJECT-ID}/pdfmerger
Deploy to Cloud Run using this command (or use the console) :
$ gcloud run deploy --image gcr.io/${PROJECT-ID}/pdfmerger --platform managed --allow-unauthenticated --memory=1024Mi
You can also deploy this pre-built public image: gcr.io/alexismp-pdfmerger/pdfmerger-unite
Concurrency can be hard. The implementation in this repository accounts for Cloud Run's built-in concurrency to make sure multiple users don't end up mixing their files (ugh!) since they share the same filesystem. Files for every given user are grouped with a common prefix and their order is preserved in a List.
You could use a somewhat simpler implementation and set the concurrency of Cloud Run to 1, essentially mimicking what FaaS products usually do, but this would likely increase the number of cold starts and thus worsen the user experience.
Please follow the instructions in CONTRIBUTING.md. Forking is great too!
Yes, the UI is crap and the UX is just as bad. Some improvement ideas include :
- manipulate an arbitrary number of files (selecting multiple files in a single input doesn't help with order). The backend is not written for a hard-coded number of files.
- re-order files before uploading them (in which case a single input for mutiple files could work)
- provide an upload tracker for large files
- preview files before they're uploaded
- client-side testing that the files actually contain PDF (server-side testing is implemented)
This list is not an official Google product. Links on this list also are not necessarily to official Google products.