hakib / fast-excel-python Goto Github PK

You really don't need to look at Pandas and Tablib because they both depend on Openpyxl for the reading. You should start with that and the additional optimisation of ws.iter_rows(values_only=True) or just plain ws.values which avoids the creation of cell objects altogether. Getting integers is surprisingly costly in Python but the real bottleneck is the underlying parser. We use etree because it's the best way to handle reading and writing all the thousands of possible objects, but it's anything but fast. I've run some bottlenecks with other parsers and they're all much, much faster but the conversion to Python objects is slow. I might look at the underlying Rust XML parser; converting rows at once might be an acceptable compromise.
Iterating does, of course, allow streaming conversions and can be combined with multiprocessing for handling multiple sheets in the same workbook at once.