Hi all, Since my last mail, I’ve added a wide variety of analysis tools, and OCR. Enabling the NumPy ecosystem in another process unlocks many other cool features. I made an experimental converter to translate standard spreadsheets to Python ones. It handles several hundred functions so far.
I’m still looking for feedback on the architecture and code. Microsoft added Python to Excel a few years ago and LibreOffice should do the same for competitive and relevance reasons. It’s a great language for both children and data scientists. This is the main document about the features: https://github.com/KeithCu/writeragent/blob/master/docs/enabling_numpy_in_libreoffice.md Are there any Calc developers left? Thanks, Keith On Tue, May 26, 2026 at 7:58 PM Keith Curtis <[email protected]> wrote: > Hello all, > > Since my last email, I've made more progress on the =Python() feature. It > now uses a custom pickle format optimized for numerics to make IPC faster, > supports multiple ranges like =PYTHON("np.mean(data)", A1:A10, C1:C10), and > integrates a rich text editor based on Monaco that handles Python (and TeX) > syntax highlighting, etc. Monaco needs some extra packages in the external > venv, the LO side remains pure Python. > > I discovered that Excel added Python support a couple of years ago, but > their design looks messy with xl() references instead of just passing in > ranges, and which must break their existing recalc logic. They also > apparently force Python in the cloud instead of locally. > > It feels like LibreOffice could have had a cleaner native =PYTHON() (with > numpy, etc.) years ago. The good news is that I've implemented many of > the essential features in a simple, robust, and fast enough way. I'd love > to get feedback on the code and architecture. Who are the Calc maintainers > these days? > > Thank you, > > -Keith > > https://github.com/KeithCu/writeragent > >
