The default context and module-level functions
Every operation in Symi runs against a context — the symbol table, the
assumption records, and the hash-consed expression store. You can manage
contexts explicitly (see Construction and parsing), but
for everyday use the bindings provide a default context: a context created
automatically per thread (per module instance in WASM), shared by a set of
module-level functions that mirror the Context class methods one-for-one.
import symi
x = symi.symbol("x", positive=True)
print(symi.factor(x**2 + 2*x + 1)) # (1 + x)^2
print(symi.integrate(symi.sin(x), x))
print(symi.pi * x**2) # constants are valuesRules of the model:
- One default context per thread. The Rust core is thread-safe and its
public types are
Send + Sync, but a default context is a convenience for a caller who did not make one, so each thread gets its own rather than sharing one implicitly. The Python classes are unsendable and WebAssembly is single-threaded today, so in those hosts objects never cross threads anyway. Objects made by module functions and bydefault_context()interoperate freely; objects from an explicitcontext()must not be mixed in (as ever, contexts do not mix). - Assumptions are per-name and session-wide.
symbol("x", positive=True)refines whatxmeans everywhere in the default context, retroactively. Refinement is monotonic: strengthening an unknown property is allowed, contradicting an earlier declaration raises an error.clear_assumptionsis the escape hatch; a freshcontext()gives full isolation. - Memory is bounded by live objects. The expression store holds weak
references: expressions you no longer hold are reclaimed automatically, so
there is no module-level
reset_contextand long sessions do not leak.interner_lengthreports the live count. - Structural constructors use the default context.
matrix,point_2d,segment_2d,polygon_2d,triangle_2d, andcircle_2dare both exported classes and default-context constructors: callMatrix([[x, y]])in Python orsymi.matrix([[x, y]])through the JavaScript facade. Their methods on an explicitContext()remain available for isolated construction.undefined_functionis a module-level factory in Python and on the raw WASM module (undefinedFunction); it has no class constructor.reset_contextis likewise context-only: resetting the shared default context would strand every live object, and the weak store removes the memory reason to do it. - Python
import *shadows builtins. The module exportsabs,min, andmax; useimport symi as scin code that also needs the Python builtins.
Functions#
default_context#
Default context
default_context() -> ContextReturn a handle to the thread's default context — the same context every module-level function and direct structural-class constructor uses. Pass it to code written against the explicit-context API. Each call returns a new handle to the same underlying context.
clear_assumptions#
Default context
clear_assumptions(variable: VariableLike) -> NoneCalling forms
Explicit context: Context.clear_assumptions
clear_assumptions(variable: VariableLike) -> NoneReset one symbol's assumptions to the default (complex domain, every property unknown). This is the only sanctioned widening operation: it is explicit and per-symbol, and the symbol's identity is untouched, so existing expressions remain structurally valid. Use it when a monotonic-refinement conflict blocks a re-declaration you actually intend.
assumptions_of#
Default context
assumptions_of(variable: Any) -> list[str]Return the symbol's current assumptions as a native Python mapping or
JavaScript record with domain, positive, negative, nonzero, and
finite fields. The variable may be a name or same-context symbol. The domain
is one of complex, real, rational, integer, natural, or
positive_integer; truth-valued fields are yes, no, or unknown.
interner_length#
Default context
interner_length() -> intNumber of live expressions currently interned in the context. Dead entries awaiting the amortized sweep are not counted. Useful for observing memory in long-running sessions and in tests that assert the store stays bounded.
Example#
Module-level calls share one context; an explicit context isolates a computation.
import symi
shared = symi.symbol("shared")
print(symi.default_context().integrate(shared, "shared"))
isolated = symi.Context()
isolated_symbol = isolated.symbol("shared")
print(isolated.integrate(isolated_symbol, "shared"))