Getting started: Python
Alpha releases are published on PyPI as prebuilt wheels for CPython 3.9 or
newer. Install the symi-cas distribution, then import symi:
python -m pip install --pre symi-casTo build the extension from source instead, run in the activated project environment:
cd symi-python
maturin develop --releasePrefer a qualified import so Symi names do not shadow Python builtins:
from fractions import Fraction
import symi
x = symi.symbol("x", positive=True)
polynomial = x**2 + 2*x + 1
print(polynomial.factor()) # (x + 1)^2
print(symi.integrate(symi.sin(x), x))
print(polynomial.substitute(x, 3))
print(symi.rational(2, 3) + Fraction(1, 3))
assert str(symi.integer(1 << 4096)) == str(1 << 4096)(x + 1)^2
-cos(x)
16
1Python int, excluding bool, is accepted anywhere an expression-like value
is expected and retains arbitrary precision. fractions.Fraction is converted
exactly. A float is never silently promoted to an exact expression.
Constants are values, not calls:
circle_area = symi.pi * symi.symbol("radius")**2
assert symi.evaluate_truth(symi.rational(1, 3) < 1) is TrueRelations are symbolic expressions. Use evaluate_truth for an explicit
True, False, or None; use execute when a symbolic true/false expression
is wanted. Python truth testing of an undecidable symbolic value raises an
error instead of guessing.
Classes take Python's CapWords spelling — Expression, Matrix, Context,
Point2d — while methods, module functions, and keyword arguments keep the
Symi underscore spelling. For an isolated computation, create an explicit
context:
context = symi.Context()
x = context.symbol("x")
matrix = context.matrix(((1, x), (x, 1)))
print(matrix.determinant())Variable parameters accept either a name or a symbol from the same context.
Collections accept finite iterables according to their mathematical meaning:
matrices require ordered rows, while finite_set also accepts unordered host
collections. See Conversions and collections and
Context safety.
Errors carry stable structured attributes:
try:
symi.integer(1.5)
except TypeError as error:
print(error.category, error.operation, error.parameter_path)
print(error.received, error.expected, error.hint)Boundary validation raises TypeError for a wrong host type, ValueError for
an invalid exact conversion, and RuntimeError for a context or staleness
violation. The structured attributes above are present on boundary
diagnostics; a mathematical precondition reported from inside an operation
(for example a zero denominator in rational) raises a plain ValueError
carrying only its message.
The supplied symi.pyi describes expression-like inputs, constants,
structured objects, and context methods for editors and type checkers.