pydata/numexpr

Health v2 · maintenance only

Adoption not part of the score

2535 stars · 226 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

NumExpr is a fast numerical expression evaluator for NumPy arrays that compiles expressions like '3*a+4*b' into op-codes executed by an integrated virtual machine. It avoids intermediate memory allocations and uses multi-threading (optionally Intel MKL/VML) to speed up large array computations.

Use cases

• evaluate arithmetic expressions on large numpy arrays faster
• reduce memory usage in pandas query computations
• speed up element-wise math on big arrays with multiple cores
• accelerate numerical expressions in PyTables queries
• compute complex array formulas without temporary allocations

When to choose

• you work with large NumPy/Pandas arrays and expression evaluation is a bottleneck
• you want multi-core speedups without rewriting vectorized code
• you need lower memory overhead than NumPy's intermediate arrays

When to avoid

• your arrays are small, where NumPy is equally fast or faster
• you need general-purpose symbolic math or arbitrary Python code evaluation
• your expressions are trivial like 'a + 1' with little expected speedup

Facets

library · maturity stable

math machine-learning data-science data-science performance python cross-platform numpy pandas numerical-computing expression-evaluator multithreading array-computing algorithms

Member repositories

For agents

markdown · JSON · MCP: product_card(name="pydata/numexpr")