Python module for parsing Apple Numbers .numbers files
github.com
github.com
I've seen this elsewhere in the wild. E.g. HDFS namenode file system image. But it seems like such a strange choice for a spreadsheet. Why not throw it into sqlite?
So this single file setup is actually a conversion of how a multi file directory structure would have been.
The choice of a document package makes sense when you see how even a spreadsheet might embed media etc... But synchronization applications like Box fail to handle packages even today so moving to a single container file allows keeping a lot of the old parsing code intact while achieving better support for things that prefer single files.
CREATE TABLE cells (sheet TEXT NOT NULL, x TEXT NOT NULL, y INTEGER NOT NULL, content TEXT, PRIMARY KEY (sheet, x, y))
Content can be text, numbers, formulas, or a reference to a media collection containing blobs. You can also pull the sheet out to a reference, etc.You could also do some cheeky stuff like making an rtree index to manage the visible cells so you only need to evaluate those in view for a quicker first paint.
This is not the only thing wrong in the world, but it is one of the things wrong in the world.
Given Joel Spolsky’s stories of weird Excel corner cases the Numbers team had their hands full making redundant software somehow different and yet remain grounded in something akin to printed sheets of paper.
https://www.joelonsoftware.com/2006/06/16/my-first-billg-rev...
If you're loading an entire sheet, I guarantee sqlite will be faster than deserializing protobuf.