Performance tips

The new version of Paperpile is generally much faster across all user-relevant metrics for the vast majority of users. We estimate that 99.3% of users won’t need to read this guide.

However, some users with large libraries (tens of thousands of references), a high number of folders or labels (thousands), slow computers (e.g., Chromebooks), or any combination of these factors may experience performance issues.

This guide provides some background information and tips on how to manage potential performance issues.

Background: What’s changed in the new Paperpile?

The new version of Paperpile uses a “local first” approach: it stores your data locally on your computer and syncs all changes with our server in the background.

This approach makes the app feel more like a desktop application, with instantaneous actions and no need to wait for server data. Paperpile is also now prepared to work offline and as a standalone desktop app in the future (stay tuned!).

A potential downside is that your data needs to be synced before it can be viewed and more complex calculations are performed on your local computer instead of our servers.

Tips

• Wait for the initial sync to complete. For libraries with a few thousand references, the sync only takes seconds. For larger libraries, it can take longer, depending on the library size and your network speed. Paperpile attempts to download the data in one batch through an optimized connection. If this takes longer than a minute, it switches to downloading in smaller batches, which can take even more time.
• Expect high CPU load and short freezes after the initial sync. Once the data is downloaded, Paperpile saves it to a local database and performs calculations to efficiently display data (e.g., for filtering, sorting, and counting). For large libraries on slow computers, this can stress the CPU, leading to stutters or freezes. These issues will subside once the calculations are complete.
• Close the tab instead of signing out. Some users log out of Paperpile after each session. For security reasons, signing out deletes all local data, requiring a full sync the next time you sign in. If no one else uses your browser, we recommend simply closing the Paperpile tab instead of logging out each time.
• Delete references from the Trash in smaller batches. To reduce library size, delete references you no longer need by moving them to the Trash and then deleting them permanently. Currently, deleting a large number of references from the Trash can be slow. Until this is addressed (see below), we recommend deleting references in batches of no more than 5,000 and waiting until the library has fully synced afterward (wait for the progress bar under “My Library” to disappear).
• Take care when working with many folders or labels. If you have an unusually large number of folders/labels or references with many assigned labels, adding or removing labels/folders can be slow. Each action triggers a recalculation of the data structures needed to display, filter, sort, and count references efficiently. This process can be CPU intensive, causing stutters or even complete freezing of the user interface. Until a fix is available in the coming weeks, we recommend making changes to labels or folders one at a time and waiting for the UI to become fully responsive before making another change.

How we will address these issues

Our goal is to eliminate these bottlenecks and any tradeoffs between desktop and web application functionality. We want Paperpile to deliver the best of both worlds without compromises for all users including power-users:

• We will release a fix for label/folder-related performance issues by the end of September 2024.
• We will make it easier to split large libraries into smaller ones and allow users to choose which libraries to sync on specific computers.
• We will improve initial sync times, with small libraries seeing the most significant improvements, but large libraries will also sync faster.
• We will make deleting references from the Trash much faster, so you can easily clean up your library if it has grown too large (autumn 2024).
• We will improve the overall speed of the application by adopting a new browser database technology (expected in 2025).