Merge pull request #272 from egraphs-good/builtins

Easily create pure egglog output of jitted function

Author: saulshanabrook

python/egglog/egraph.py

@@ -1241,10 +1241,10 @@ def register(
 
     def _register_commands(self, cmds: list[Command]) -> None:
         self._add_decls(*cmds)
-        egg_cmds = list(map(self._command_to_egg, cmds))
+        egg_cmds = [egg_cmd for cmd in cmds if (egg_cmd := self._command_to_egg(cmd)) is not None]
         self._egraph.run_program(*egg_cmds)
 
-    def _command_to_egg(self, cmd: Command) -> bindings._Command:
+    def _command_to_egg(self, cmd: Command) -> bindings._Command | None:
         ruleset_name = ""
         cmd_decl: CommandDecl
         match cmd:

python/egglog/egraph_state.py

@@ -7,7 +7,7 @@
 import re
 from collections import defaultdict
 from dataclasses import dataclass, field
-from typing import TYPE_CHECKING, overload
+from typing import TYPE_CHECKING, Literal, overload
 
 from typing_extensions import assert_never
 
@@ -116,7 +116,8 @@ def ruleset_to_egg(self, name: str) -> None:
                     if rule in added_rules:
                         continue
                     cmd = self.command_to_egg(rule, name)
-                    self.egraph.run_program(cmd)
+                    if cmd is not None:
+                        self.egraph.run_program(cmd)
                     added_rules.add(rule)
             case CombinedRulesetDecl(rulesets):
                 if name in self.rulesets:
@@ -126,10 +127,13 @@ def ruleset_to_egg(self, name: str) -> None:
                     self.ruleset_to_egg(ruleset)
                 self.egraph.run_program(bindings.UnstableCombinedRuleset(name, list(rulesets)))
 
-    def command_to_egg(self, cmd: CommandDecl, ruleset: str) -> bindings._Command:
+    def command_to_egg(self, cmd: CommandDecl, ruleset: str) -> bindings._Command | None:
         match cmd:
             case ActionCommandDecl(action):
-                return bindings.ActionCommand(self.action_to_egg(action, expr_to_let=True))
+                action_egg = self.action_to_egg(action, expr_to_let=True)
+                if not action_egg:
+                    return None
+                return bindings.ActionCommand(action_egg)
             case RewriteDecl(tp, lhs, rhs, conditions) | BiRewriteDecl(tp, lhs, rhs, conditions):
                 self.type_ref_to_egg(tp)
                 rewrite = bindings.Rewrite(
@@ -166,7 +170,13 @@ def command_to_egg(self, cmd: CommandDecl, ruleset: str) -> bindings._Command:
             case _:
                 assert_never(cmd)
 
-    def action_to_egg(self, action: ActionDecl, expr_to_let: bool = False) -> bindings._Action:
+    @overload
+    def action_to_egg(self, action: ActionDecl) -> bindings._Action: ...
+
+    @overload
+    def action_to_egg(self, action: ActionDecl, expr_to_let: Literal[True] = ...) -> bindings._Action | None: ...
+
+    def action_to_egg(self, action: ActionDecl, expr_to_let: bool = False) -> bindings._Action | None:  # noqa: C901, PLR0911, PLR0912
         match action:
             case LetDecl(name, typed_expr):
                 var_decl = VarDecl(name, True)
@@ -179,7 +189,11 @@ def action_to_egg(self, action: ActionDecl, expr_to_let: bool = False) -> bindin
                 return bindings.Set(span(), call_.name, call_.args, self._expr_to_egg(rhs))
             case ExprActionDecl(typed_expr):
                 if expr_to_let:
-                    typed_expr = self._transform_let(typed_expr)
+                    maybe_typed_expr = self._transform_let(typed_expr)
+                    if maybe_typed_expr:
+                        typed_expr = maybe_typed_expr
+                    else:
+                        return None
                 return bindings.Expr_(span(), self.typed_expr_to_egg(typed_expr))
             case ChangeDecl(tp, call, change):
                 self.type_ref_to_egg(tp)
@@ -351,13 +365,13 @@ def typed_expr_to_egg(self, typed_expr_decl: TypedExprDecl, transform_let: bool
         self.type_ref_to_egg(typed_expr_decl.tp)
         return self._expr_to_egg(typed_expr_decl.expr)
 
-    def _transform_let(self, typed_expr: TypedExprDecl) -> TypedExprDecl:
+    def _transform_let(self, typed_expr: TypedExprDecl) -> TypedExprDecl | None:
         """
         Rewrites this expression as a let binding if it's not already a let binding.
         """
         var_decl = VarDecl(f"__expr_{hash(typed_expr)}", True)
         if var_decl in self.expr_to_egg_cache:
-            return TypedExprDecl(typed_expr.tp, var_decl)
+            return None
         var_egg = self._expr_to_egg(var_decl)
         cmd = bindings.ActionCommand(bindings.Let(span(), var_egg.name, self.typed_expr_to_egg(typed_expr)))
         try:
@@ -367,7 +381,7 @@ def _transform_let(self, typed_expr: TypedExprDecl) -> TypedExprDecl:
             return typed_expr
         self.expr_to_egg_cache[typed_expr.expr] = var_egg
         self.expr_to_egg_cache[var_decl] = var_egg
-        return TypedExprDecl(typed_expr.tp, var_decl)
+        return None
 
     @overload
     def _expr_to_egg(self, expr_decl: CallDecl) -> bindings.Call: ...

python/egglog/exp/array_api.py

@@ -290,8 +290,8 @@ def check_index(length: IntLike, idx: IntLike) -> Int:
     """
     Returns the index if 0 <= idx < length, otherwise returns Int.NEVER
     """
-    length = cast(Int, length)
-    idx = cast(Int, idx)
+    length = cast("Int", length)
+    idx = cast("Int", idx)
     return Int.if_(((idx >= 0) & (idx < length)), idx, Int.NEVER)
 
 
@@ -336,7 +336,7 @@ def abs(self) -> Float: ...
 
     @method(cost=2)
     @classmethod
-    def rational(cls, r: Rational) -> Float: ...
+    def rational(cls, r: BigRat) -> Float: ...
 
     @classmethod
     def from_int(cls, i: IntLike) -> Float: ...
@@ -362,15 +362,15 @@ def __eq__(self, other: FloatLike) -> Boolean: ...  # type: ignore[override]
 
 
 @array_api_ruleset.register
-def _float(fl: Float, f: f64, f2: f64, i: i64, r: Rational, r1: Rational):
+def _float(fl: Float, f: f64, f2: f64, i: i64, r: BigRat, r1: BigRat):
     return [
         rule(eq(fl).to(Float(f))).then(set_(fl.to_f64).to(f)),
         rewrite(Float(f).abs()).to(Float(f), f >= 0.0),
         rewrite(Float(f).abs()).to(Float(-f), f < 0.0),
         # Convert from float to rationl, if its a whole number i.e. can be converted to int
-        rewrite(Float(f)).to(Float.rational(Rational(f.to_i64(), 1)), eq(f64.from_i64(f.to_i64())).to(f)),
+        rewrite(Float(f)).to(Float.rational(BigRat(f.to_i64(), 1)), eq(f64.from_i64(f.to_i64())).to(f)),
         # always convert from int to rational
-        rewrite(Float.from_int(Int(i))).to(Float.rational(Rational(i, 1))),
+        rewrite(Float.from_int(Int(i))).to(Float.rational(BigRat(i, 1))),
         rewrite(Float(f) + Float(f2)).to(Float(f + f2)),
         rewrite(Float(f) - Float(f2)).to(Float(f - f2)),
         rewrite(Float(f) * Float(f2)).to(Float(f * f2)),
@@ -417,7 +417,7 @@ def range(cls, stop: IntLike) -> TupleInt:
     def from_vec(cls, vec: VecLike[Int, IntLike]) -> TupleInt: ...
 
     def __add__(self, other: TupleIntLike) -> TupleInt:
-        other = cast(TupleInt, other)
+        other = cast("TupleInt", other)
         return TupleInt(
             self.length() + other.length(), lambda i: Int.if_(i < self.length(), self[i], other[i - self.length()])
         )
@@ -475,14 +475,14 @@ def select(self, indices: TupleIntLike) -> TupleInt:
         """
         Return a new tuple with the elements at the given indices
         """
-        indices = cast(TupleInt, indices)
+        indices = cast("TupleInt", indices)
         return indices.map(lambda i: self[i])
 
     def deselect(self, indices: TupleIntLike) -> TupleInt:
         """
         Return a new tuple with the elements not at the given indices
         """
-        indices = cast(TupleInt, indices)
+        indices = cast("TupleInt", indices)
         return TupleInt.range(self.length()).filter(lambda i: ~indices.contains(i)).map(lambda i: self[i])
 
 
@@ -554,7 +554,7 @@ def __init__(self, length: IntLike, idx_fn: Callable[[Int], TupleInt]) -> None:
     @method(subsume=True)
     @classmethod
     def single(cls, i: TupleIntLike) -> TupleTupleInt:
-        i = cast(TupleInt, i)
+        i = cast("TupleInt", i)
         return TupleTupleInt(1, lambda _: i)
 
     @method(subsume=True)
@@ -564,7 +564,7 @@ def from_vec(cls, vec: Vec[TupleInt]) -> TupleTupleInt: ...
     def append(self, i: TupleIntLike) -> TupleTupleInt: ...
 
     def __add__(self, other: TupleTupleIntLike) -> TupleTupleInt:
-        other = cast(TupleTupleInt, other)
+        other = cast("TupleTupleInt", other)
         return TupleTupleInt(
             self.length() + other.length(),
             lambda i: TupleInt.if_(i < self.length(), self[i], other[i - self.length()]),
@@ -840,7 +840,7 @@ def _value(i: Int, f: Float, b: Boolean, v: Value, v1: Value, i1: Int, f1: Float
 
     yield rewrite(Value.float(f).sqrt()).to(Value.float(f ** (0.5)))
 
-    yield rewrite(Value.float(Float.rational(Rational(0, 1))) + v).to(v)
+    yield rewrite(Value.float(Float.rational(BigRat(0, 1))) + v).to(v)
 
     yield rewrite(Value.if_(TRUE, v, v1)).to(v)
     yield rewrite(Value.if_(FALSE, v, v1)).to(v1)
@@ -862,7 +862,7 @@ def append(self, i: ValueLike) -> TupleValue: ...
     def from_vec(cls, vec: Vec[Value]) -> TupleValue: ...
 
     def __add__(self, other: TupleValueLike) -> TupleValue:
-        other = cast(TupleValue, other)
+        other = cast("TupleValue", other)
         return TupleValue(
             self.length() + other.length(),
             lambda i: Value.if_(i < self.length(), self[i], other[i - self.length()]),
@@ -875,13 +875,13 @@ def __getitem__(self, i: Int) -> Value: ...
     def foldl_boolean(self, f: Callable[[Boolean, Value], Boolean], init: BooleanLike) -> Boolean: ...
 
     def contains(self, value: ValueLike) -> Boolean:
-        value = cast(Value, value)
+        value = cast("Value", value)
         return self.foldl_boolean(lambda acc, j: acc | (value == j), FALSE)
 
     @method(subsume=True)
     @classmethod
     def from_tuple_int(cls, ti: TupleIntLike) -> TupleValue:
-        ti = cast(TupleInt, ti)
+        ti = cast("TupleInt", ti)
         return TupleValue(ti.length(), lambda i: Value.int(ti[i]))
 
 
@@ -1259,7 +1259,7 @@ def append(self, i: NDArrayLike) -> TupleNDArray: ...
     def from_vec(cls, vec: Vec[NDArray]) -> TupleNDArray: ...
 
     def __add__(self, other: TupleNDArrayLike) -> TupleNDArray:
-        other = cast(TupleNDArray, other)
+        other = cast("TupleNDArray", other)
         return TupleNDArray(
             self.length() + other.length(),
             lambda i: NDArray.if_(i < self.length(), self[i], other[i - self.length()]),
@@ -1632,7 +1632,7 @@ def ndindex(shape: TupleIntLike) -> TupleTupleInt:
     """
     https://numpy.org/doc/stable/reference/generated/numpy.ndindex.html
     """
-    shape = cast(TupleInt, shape)
+    shape = cast("TupleInt", shape)
     return shape.map_tuple_int(TupleInt.range).product()
 
 

python/egglog/exp/array_api_jit.py

@@ -2,10 +2,14 @@
 from collections.abc import Callable
 from typing import TypeVar, cast
 
+import numpy as np
+
 from egglog import EGraph, try_evaling
 from egglog.exp.array_api import NDArray
 from egglog.exp.array_api_numba import array_api_numba_schedule
-from egglog.exp.array_api_program_gen import array_api_program_gen_schedule, ndarray_function_two
+from egglog.exp.array_api_program_gen import EvalProgram, array_api_program_gen_schedule, ndarray_function_two_program
+
+from .program_gen import Program
 
 X = TypeVar("X", bound=Callable)
 
@@ -14,15 +18,27 @@ def jit(fn: X) -> X:
     """
     Jit compiles a function
     """
-    sig = inspect.signature(fn)
-    arg1, arg2 = sig.parameters.keys()
-    egraph = EGraph()
+    egraph, res, res_optimized, program = function_to_program(fn, save_egglog_string=False)
+    fn_program = EvalProgram(program, {"np": np})
     with egraph.set_current():
-        res = fn(NDArray.var(arg1), NDArray.var(arg2))
-    res_optimized = egraph.simplify(res, array_api_numba_schedule)
-
-    fn_program = ndarray_function_two(res_optimized, NDArray.var(arg1), NDArray.var(arg2))
-    fn = try_evaling(array_api_program_gen_schedule, fn_program, fn_program.as_py_object)
+        fn = cast("X", try_evaling(array_api_program_gen_schedule, fn_program, fn_program.as_py_object))
     fn.initial_expr = res  # type: ignore[attr-defined]
     fn.expr = res_optimized  # type: ignore[attr-defined]
-    return cast(X, fn)
+    return fn
+
+
+def function_to_program(fn: Callable, save_egglog_string: bool) -> tuple[EGraph, NDArray, NDArray, Program]:
+    sig = inspect.signature(fn)
+    arg1, arg2 = sig.parameters.keys()
+    egraph = EGraph(save_egglog_string=save_egglog_string)
+    with egraph:
+        with egraph.set_current():
+            res = fn(NDArray.var(arg1), NDArray.var(arg2))
+        res_optimized = egraph.simplify(res, array_api_numba_schedule)
+
+    return (
+        egraph,
+        res,
+        res_optimized,
+        ndarray_function_two_program(res_optimized, NDArray.var(arg1), NDArray.var(arg2)),
+    )

python/egglog/exp/array_api_program_gen.py

@@ -14,13 +14,13 @@
 array_api_program_gen_ruleset = ruleset(name="array_api_program_gen_ruleset")
 array_api_program_gen_eval_ruleset = ruleset(name="array_api_program_gen_eval_ruleset")
 
-array_api_program_gen_schedule = (
+array_api_program_gen_combined_ruleset = (
     array_api_program_gen_ruleset
     | program_gen_ruleset
     | array_api_program_gen_eval_ruleset
-    | eval_program_rulseset
     | array_api_vec_to_cons_ruleset
-).saturate()
+)
+array_api_program_gen_schedule = (array_api_program_gen_combined_ruleset | eval_program_rulseset).saturate()
 
 
 @function
@@ -116,7 +116,7 @@ def float_program(x: Float) -> Program: ...
 
 
 @array_api_program_gen_ruleset.register
-def _float_program(f: Float, g: Float, f64_: f64, i: Int, r: Rational):
+def _float_program(f: Float, g: Float, f64_: f64, i: Int, r: BigRat):
     yield rewrite(float_program(Float(f64_))).to(Program(f64_.to_string()))
     yield rewrite(float_program(f.abs())).to(Program("np.abs(") + float_program(f) + ")")
     yield rewrite(float_program(Float.from_int(i))).to(int_program(i))
@@ -126,10 +126,10 @@ def _float_program(f: Float, g: Float, f64_: f64, i: Int, r: Rational):
     yield rewrite(float_program(f / g)).to(Program("(") + float_program(f) + " / " + float_program(g) + ")")
     yield rewrite(float_program(Float.rational(r))).to(
         Program("float(") + Program(r.numer.to_string()) + " / " + Program(r.denom.to_string()) + ")",
-        ne(r.denom).to(i64(1)),
+        ne(r.denom).to(BigInt(1)),
     )
     yield rewrite(float_program(Float.rational(r))).to(
-        Program("float(") + Program(r.numer.to_string()) + ")", eq(r.denom).to(i64(1))
+        Program("float(") + Program(r.numer.to_string()) + ")", eq(r.denom).to(BigInt(1))
     )
 
 

python/tests/__snapshots__/test_array_api/test_jit[lda][expr].py

@@ -17,7 +17,7 @@
         )
     ),
     DType.float64,
-) / NDArray.scalar(Value.float(Float.rational(Rational(150, 1))))
+) / NDArray.scalar(Value.float(Float.rational(BigRat(BigInt.from_string("150"), BigInt.from_string("1")))))
 _NDArray_4 = zeros(TupleInt.from_vec(Vec[Int](Int(3), Int(4))), OptionalDType.some(DType.float64), OptionalDevice.some(_NDArray_1.device))
 _MultiAxisIndexKeyItem_1 = MultiAxisIndexKeyItem.slice(Slice())
 _IndexKey_1 = IndexKey.multi_axis(MultiAxisIndexKey.from_vec(Vec[MultiAxisIndexKeyItem](MultiAxisIndexKeyItem.int(Int(0)), _MultiAxisIndexKeyItem_1)))
@@ -36,15 +36,24 @@
 _NDArray_9 = square(_NDArray_8 - expand_dims(sum(_NDArray_8, _OptionalIntOrTuple_1) / NDArray.scalar(Value.int(_NDArray_8.shape[Int(0)]))))
 _NDArray_10 = sqrt(sum(_NDArray_9, _OptionalIntOrTuple_1) / NDArray.scalar(Value.int(_NDArray_9.shape[Int(0)])))
 _NDArray_11 = copy(_NDArray_10)
-_NDArray_11[IndexKey.ndarray(_NDArray_10 == NDArray.scalar(Value.int(Int(0))))] = NDArray.scalar(Value.float(Float.rational(Rational(1, 1))))
-_TupleNDArray_1 = svd(sqrt(asarray(NDArray.scalar(Value.float(Float.rational(Rational(1, 147)))), OptionalDType.some(DType.float64))) * (_NDArray_8 / _NDArray_11), Boolean(False))
+_NDArray_11[IndexKey.ndarray(_NDArray_10 == NDArray.scalar(Value.int(Int(0))))] = NDArray.scalar(
+    Value.float(Float.rational(BigRat(BigInt.from_string("1"), BigInt.from_string("1"))))
+)
+_TupleNDArray_1 = svd(
+    sqrt(asarray(NDArray.scalar(Value.float(Float.rational(BigRat(BigInt.from_string("1"), BigInt.from_string("147"))))), OptionalDType.some(DType.float64)))
+    * (_NDArray_8 / _NDArray_11),
+    Boolean(False),
+)
 _Slice_1 = Slice(OptionalInt.none, OptionalInt.some(sum(astype(_TupleNDArray_1[Int(1)] > NDArray.scalar(Value.float(Float(0.0001))), DType.int32)).to_value().to_int))
 _NDArray_12 = (
     _TupleNDArray_1[Int(2)][IndexKey.multi_axis(MultiAxisIndexKey.from_vec(Vec[MultiAxisIndexKeyItem](MultiAxisIndexKeyItem.slice(_Slice_1), _MultiAxisIndexKeyItem_1)))]
     / _NDArray_11
 ).T / _TupleNDArray_1[Int(1)][IndexKey.slice(_Slice_1)]
 _TupleNDArray_2 = svd(
-    (sqrt((NDArray.scalar(Value.int(Int(150))) * _NDArray_3) * NDArray.scalar(Value.float(Float.rational(Rational(1, 2))))) * (_NDArray_4 - (_NDArray_3 @ _NDArray_4)).T).T
+    (
+        sqrt((NDArray.scalar(Value.int(Int(150))) * _NDArray_3) * NDArray.scalar(Value.float(Float.rational(BigRat(BigInt.from_string("1"), BigInt.from_string("2"))))))
+        * (_NDArray_4 - (_NDArray_3 @ _NDArray_4)).T
+    ).T
     @ _NDArray_12,
     Boolean(False),
 )

python/tests/__snapshots__/test_array_api/test_jit[lda][initial_expr].py

@@ -41,8 +41,14 @@
     OptionalInt.some(Int(0)),
 )
 _NDArray_6 = std(_NDArray_5, _OptionalIntOrTuple_1)
-_NDArray_6[IndexKey.ndarray(std(_NDArray_5, _OptionalIntOrTuple_1) == NDArray.scalar(Value.int(Int(0))))] = NDArray.scalar(Value.float(Float.rational(Rational(1, 1))))
-_TupleNDArray_1 = svd(sqrt(asarray(NDArray.scalar(Value.float(Float.rational(Rational(1, 147)))), OptionalDType.some(DType.float64))) * (_NDArray_5 / _NDArray_6), Boolean(False))
+_NDArray_6[IndexKey.ndarray(std(_NDArray_5, _OptionalIntOrTuple_1) == NDArray.scalar(Value.int(Int(0))))] = NDArray.scalar(
+    Value.float(Float.rational(BigRat(BigInt.from_string("1"), BigInt.from_string("1"))))
+)
+_TupleNDArray_1 = svd(
+    sqrt(asarray(NDArray.scalar(Value.float(Float.rational(BigRat(BigInt.from_string("1"), BigInt.from_string("147"))))), OptionalDType.some(DType.float64)))
+    * (_NDArray_5 / _NDArray_6),
+    Boolean(False),
+)
 _Slice_1 = Slice(OptionalInt.none, OptionalInt.some(sum(astype(_TupleNDArray_1[Int(1)] > NDArray.scalar(Value.float(Float(0.0001))), DType.int32)).to_value().to_int))
 _NDArray_7 = asarray(reshape(asarray(_NDArray_2), TupleInt.from_vec(Vec[Int](Int(-1)))))
 _NDArray_8 = unique_values(concat(TupleNDArray.from_vec(Vec[NDArray](unique_values(asarray(_NDArray_7))))))
@@ -61,9 +67,12 @@
 )
 _NDArray_10 = copy(_NDArray_9)
 _NDArray_10[IndexKey.ndarray(_NDArray_9 == NDArray.scalar(Value.int(Int(0))))] = NDArray.scalar(Value.float(Float(1.0)))
-_NDArray_11 = astype(unique_counts(_NDArray_2)[Int(1)], DType.float64) / NDArray.scalar(Value.float(Float.rational(Rational(150, 1))))
+_NDArray_11 = astype(unique_counts(_NDArray_2)[Int(1)], DType.float64) / NDArray.scalar(Value.float(Float.rational(BigRat(BigInt.from_string("150"), BigInt.from_string("1")))))
 _TupleNDArray_2 = svd(
-    (sqrt((NDArray.scalar(Value.int(Int(150))) * _NDArray_11) * NDArray.scalar(Value.float(Float.rational(Rational(1, 2))))) * (_NDArray_4 - (_NDArray_11 @ _NDArray_4)).T).T
+    (
+        sqrt((NDArray.scalar(Value.int(Int(150))) * _NDArray_11) * NDArray.scalar(Value.float(Float.rational(BigRat(BigInt.from_string("1"), BigInt.from_string("2"))))))
+        * (_NDArray_4 - (_NDArray_11 @ _NDArray_4)).T
+    ).T
     @ (
         (
             _TupleNDArray_1[Int(2)][IndexKey.multi_axis(MultiAxisIndexKey.from_vec(Vec[MultiAxisIndexKeyItem](MultiAxisIndexKeyItem.slice(_Slice_1), _MultiAxisIndexKeyItem_1)))]