A Penpot file that opens after reimport has passed the first part of a backup check. For a document I need to keep editing, I also want a list of the objects that came back and a way to compare their positions.
Keep the source and test the copy
For this check we exported a synthetic Penpot document through the GUI. The archive was 93,430 bytes. Its ZIP records passed a CRC check, and we computed a SHA-256 digest so we could identify that exact file later. The digest identifies the archive; it does not verify the contents of a restored project.
We imported the archive into a separate file and checked that its file ID differed from the source. The source remained available. Only then did we compare the archived document with the imported copy.
Compare by object ID
The fixture had seven pages and 56 objects, including page roots. We compared page names and IDs, object ID sets, object names, types, and the x, y, width and height fields. Match the sets of IDs before comparing geometry. Otherwise a difference in traversal order can make two unchanged objects look like a mismatch.
Some archive and API type names differ. In this fixture we mapped frame to board, rect to rectangle, and circle to ellipse. For one path we compared its stored bounding rectangle. We reloaded the browser and read the imported copy again to confirm that the checked fields persisted.
The GUI round trip worked for this fixture; the plugin API backup export still returned an error. We did not verify full rich text, fonts, fills, libraries, interaction behavior, version history or cross-version recovery. The Penpot backup test and its limits include the exact test scope. I maintain FOSS Desk. AI drafted this article from the recorded test evidence.
