https://github.com/cilladev/xlstream
Fastest streaming Excel formula evaluation engine - no dependency graph, no full-workbook buffering. Rust core with Python bindings
https://github.com/cilladev/xlstream
excel python rust
Last synced: about 2 months ago
JSON representation
Fastest streaming Excel formula evaluation engine - no dependency graph, no full-workbook buffering. Rust core with Python bindings
- Host: GitHub
- URL: https://github.com/cilladev/xlstream
- Owner: cilladev
- License: other
- Created: 2026-04-17T15:22:54.000Z (2 months ago)
- Default Branch: main
- Last Pushed: 2026-04-23T02:48:54.000Z (about 2 months ago)
- Last Synced: 2026-04-23T03:02:32.057Z (about 2 months ago)
- Topics: excel, python, rust
- Language: Rust
- Homepage: https://xlstream.dev
- Size: 774 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- Changelog: CHANGELOG.md
- Contributing: CONTRIBUTING.md
- License: LICENSE-APACHE
- Security: SECURITY.md
- Roadmap: docs/roadmap/README.md
Awesome Lists containing this project
README
# xlstream
`xlstream` is a **streaming Excel formula evaluation engine** written in Rust, with Python bindings. It reads `.xlsx` files row-by-row, evaluates formulas in a single (or two-pass) streaming traversal (no dependency graph, no full-workbook buffering), and writes the results to a new `.xlsx` — all in bounded memory regardless of file size.
We are not building a general-purpose spreadsheet engine. We are building the *fastest, leanest* Excel formula evaluator — supporting ~465 of Excel's ~500 functions, everything that fits a streaming architecture.
## Performance
**425x faster** on the same workbook. [Full benchmarks](benchmarks/reports)
| | xlstream | formualizer |
| ------------------- | ------------------ | --------------------- |
| 700k rows x 20 cols | **48s** | 5h 40m |
| Peak memory | **734 MB** | 3.3 GB |
| Architecture | Streaming (2-pass) | Full dependency graph |
## Supported functions
113 functions + 13 operators across 9 categories. [Full list with cross-reference](docs/functions.md).
| Category | Count | Examples |
| ---------- | ----- | ------------------------------------------------- |
| Operators | 13 | `+`, `-`, `*`, `/`, `^`, `&`, `%`, comparisons |
| Logical | 11 | IF, IFS, SWITCH, IFERROR, AND, OR, NOT, XOR |
| Aggregates | 15 | SUM, SUMIF, SUMIFS, AVERAGE, COUNTIF, MEDIAN |
| Lookups | 7 | VLOOKUP, XLOOKUP, INDEX/MATCH, HLOOKUP, CHOOSE |
| Text | 19 | LEFT, UPPER, TRIM, CONCAT, TEXT, FIND, SUBSTITUTE |
| Date | 12 | TODAY, YEAR, EDATE, EOMONTH, DATEDIF, NETWORKDAYS |
| Math | 23 | ROUND, MOD, ABS, SQRT, LOG, SIN, PI, FLOOR |
| Info | 10 | ISNUMBER, ISTEXT, ISERROR, ISBLANK, ISREF, TYPE |
| Financial | 6 | PMT, PV, FV, NPV, IRR, RATE |
Excel has ~500 functions. ~465 are streaming-compatible. We're adding more each release — see the [roadmap](docs/roadmap/README.md).
## Install
```bash
pip install xlstream
```
## Usage
### Python
```python
import xlstream
result = xlstream.evaluate("input.xlsx", "output.xlsx")
# {'rows_processed': 700001, 'formulas_evaluated': 7000000, 'duration_ms': 48000}
# Parallel (row-sharded across cores)
result = xlstream.evaluate("input.xlsx", "output.xlsx", workers=8)
```
### CLI
```bash
xlstream evaluate input.xlsx -o output.xlsx -w 8 --verbose
```
### Rust
```rust
let summary = xlstream_eval::evaluate(&input, &output, Some(8))?;
```
## Error handling
```python
try:
xlstream.evaluate("input.xlsx", "output.xlsx")
except xlstream.UnsupportedFormula as e:
print(e) # names the cell, formula, and reason
except xlstream.FormulaParseError as e:
print(e)
except OSError as e:
print(e) # file not found, corrupt xlsx
```
## License
Dual-licensed under Apache-2.0 or MIT, at your option.