egglog 6.1.0__cp310-none-win_amd64.whl → 7.1.0__cp310-none-win_amd64.whl

egglog/pretty.py

@@ -0,0 +1,464 @@
+"""
+Pretty printing for declerations.
+"""
+
+from __future__ import annotations
+
+from collections import Counter, defaultdict
+from dataclasses import dataclass, field
+from typing import TYPE_CHECKING, TypeAlias
+
+import black
+from typing_extensions import assert_never
+
+from .declarations import *
+
+if TYPE_CHECKING:
+    from collections.abc import Mapping
+
+__all__ = [
+    "pretty_decl",
+    "pretty_callable_ref",
+    "BINARY_METHODS",
+    "UNARY_METHODS",
+]
+MAX_LINE_LENGTH = 110
+LINE_DIFFERENCE = 10
+BLACK_MODE = black.Mode(line_length=180)
+
+# Use this special character in place of the args, so that if the args are inlined
+# in the viz, they will replace it
+ARG_STR = "·"
+
+# Special methods which we might want to use as functions
+# Mapping to the operator they represent for pretty printing them
+# https://docs.python.org/3/reference/datamodel.html
+BINARY_METHODS = {
+    "__lt__": "<",
+    "__le__": "<=",
+    "__eq__": "==",
+    "__ne__": "!=",
+    "__gt__": ">",
+    "__ge__": ">=",
+    # Numeric
+    "__add__": "+",
+    "__sub__": "-",
+    "__mul__": "*",
+    "__matmul__": "@",
+    "__truediv__": "/",
+    "__floordiv__": "//",
+    "__mod__": "%",
+    # TODO: Support divmod, with tuple return value
+    # "__divmod__": "divmod",
+    # TODO: Three arg power
+    "__pow__": "**",
+    "__lshift__": "<<",
+    "__rshift__": ">>",
+    "__and__": "&",
+    "__xor__": "^",
+    "__or__": "|",
+}
+
+
+UNARY_METHODS = {
+    "__pos__": "+",
+    "__neg__": "-",
+    "__invert__": "~",
+}
+
+AllDecls: TypeAlias = RulesetDecl | CombinedRulesetDecl | CommandDecl | ActionDecl | FactDecl | ExprDecl | ScheduleDecl
+
+
+def pretty_decl(
+    decls: Declarations, decl: AllDecls, *, wrapping_fn: str | None = None, ruleset_name: str | None = None
+) -> str:
+    """
+    Pretty print a decleration.
+
+    This will use re-format the result and put the expression on the last line, preceeded by the statements.
+    """
+    traverse = TraverseContext()
+    traverse(decl, toplevel=True)
+    pretty = traverse.pretty(decls)
+    expr = pretty(decl, ruleset_name=ruleset_name)
+    if wrapping_fn:
+        expr = f"{wrapping_fn}({expr})"
+    program = "\n".join([*pretty.statements, expr])
+    try:
+        # TODO: Try replacing with ruff for speed
+        # https://github.com/amyreese/ruff-api
+        return black.format_str(program, mode=BLACK_MODE).strip()
+    except black.parsing.InvalidInput:
+        return program
+
+
+def pretty_callable_ref(
+    decls: Declarations,
+    ref: CallableRef,
+    first_arg: ExprDecl | None = None,
+    bound_tp_params: tuple[JustTypeRef, ...] | None = None,
+) -> str:
+    """
+    Pretty print a callable reference, using a dummy value for
+    the args if the function is not in the form `f(x, ...)`.
+
+    To be used in the visualization.
+    """
+    # Pass in three dummy args, which are the max used for any operation that
+    # is not a generic function call
+    args: list[ExprDecl] = [VarDecl(ARG_STR)] * 3
+    if first_arg:
+        args.insert(0, first_arg)
+    res = PrettyContext(decls, defaultdict(lambda: 0))._call_inner(
+        ref, args, bound_tp_params=bound_tp_params, parens=False
+    )
+    # Either returns a function or a function with args. If args are provided, they would just be called,
+    # on the function, so return them, because they are dummies
+    return res[0] if isinstance(res, tuple) else res
+
+
+# TODO: Add a different pretty callable ref that doesnt fill in wholes but instead returns the function
+# so that things like Math.__add__ will be represented properly
+
+
+@dataclass
+class TraverseContext:
+    """
+    State for traversing expressions (or declerations that contain expressions), so we can know how many parents each
+    expression has.
+    """
+
+    # All expressions we have seen (incremented the parent counts of all children)
+    _seen: set[AllDecls] = field(default_factory=set)
+    # The number of parents for each expressions
+    parents: Counter[AllDecls] = field(default_factory=Counter)
+
+    def pretty(self, decls: Declarations) -> PrettyContext:
+        """
+        Create a pretty context from the state of this traverse context.
+        """
+        return PrettyContext(decls, self.parents)
+
+    def __call__(self, decl: AllDecls, toplevel: bool = False) -> None:  # noqa: C901
+        if not toplevel:
+            self.parents[decl] += 1
+        if decl in self._seen:
+            return
+        match decl:
+            case RewriteDecl(_, lhs, rhs, conditions) | BiRewriteDecl(_, lhs, rhs, conditions):
+                self(lhs)
+                self(rhs)
+                for cond in conditions:
+                    self(cond)
+            case RuleDecl(head, body, _):
+                for action in head:
+                    self(action)
+                for fact in body:
+                    self(fact)
+            case SetDecl(_, lhs, rhs) | UnionDecl(_, lhs, rhs):
+                self(lhs)
+                self(rhs)
+            case LetDecl(_, d) | ExprActionDecl(d) | ExprFactDecl(d):
+                self(d.expr)
+            case ChangeDecl(_, d, _) | SaturateDecl(d) | RepeatDecl(d, _) | ActionCommandDecl(d):
+                self(d)
+            case PanicDecl(_) | VarDecl(_) | LitDecl(_) | PyObjectDecl(_):
+                pass
+            case EqDecl(_, decls) | SequenceDecl(decls) | RulesetDecl(decls):
+                for de in decls:
+                    self(de)
+            case CallDecl(_, exprs, _):
+                for e in exprs:
+                    self(e.expr)
+            case RunDecl(_, until):
+                if until:
+                    for f in until:
+                        self(f)
+            case PartialCallDecl(c):
+                self(c)
+            case CombinedRulesetDecl(_):
+                pass
+            case _:
+                assert_never(decl)
+
+        self._seen.add(decl)
+
+
+@dataclass
+class PrettyContext:
+    """
+
+    We need to build up a list of all the expressions we are pretty printing, so that we can see who has parents and who is mutated
+    and create temp variables for them.
+
+    """
+
+    decls: Declarations
+    parents: Mapping[AllDecls, int]
+
+    # All the expressions we have saved as names
+    names: dict[AllDecls, str] = field(default_factory=dict)
+    # A list of statements assigning variables or calling destructive ops
+    statements: list[str] = field(default_factory=list)
+    # Mapping of type to the number of times we have generated a name for that type, used to generate unique names
+    _gen_name_types: dict[str, int] = field(default_factory=lambda: defaultdict(lambda: 0))
+
+    def __call__(
+        self, decl: AllDecls, *, unwrap_lit: bool = False, parens: bool = False, ruleset_name: str | None = None
+    ) -> str:
+        if decl in self.names:
+            return self.names[decl]
+        expr, tp_name = self.uncached(decl, unwrap_lit=unwrap_lit, parens=parens, ruleset_name=ruleset_name)
+        # We use a heuristic to decide whether to name this sub-expression as a variable
+        # The rough goal is to reduce the number of newlines, given our line length of ~180
+        # We determine it's worth making a new line for this expression if the total characters
+        # it would take up is > than some constant (~ line length).
+        line_diff: int = len(expr) - LINE_DIFFERENCE
+        n_parents = self.parents[decl]
+        if n_parents > 1 and n_parents * line_diff > MAX_LINE_LENGTH:
+            self.names[decl] = expr_name = self._name_expr(tp_name, expr, copy_identifier=False)
+            return expr_name
+        return expr
+
+    def uncached(self, decl: AllDecls, *, unwrap_lit: bool, parens: bool, ruleset_name: str | None) -> tuple[str, str]:  # noqa: PLR0911
+        match decl:
+            case LitDecl(value):
+                match value:
+                    case None:
+                        return "Unit()", "Unit"
+                    case bool(b):
+                        return str(b) if unwrap_lit else f"Bool({b})", "Bool"
+                    case int(i):
+                        return str(i) if unwrap_lit else f"i64({i})", "i64"
+                    case float(f):
+                        return str(f) if unwrap_lit else f"f64({f})", "f64"
+                    case str(s):
+                        return repr(s) if unwrap_lit else f"String({s!r})", "String"
+                assert_never(value)
+            case VarDecl(name):
+                return name, name
+            case CallDecl(_, _, _):
+                return self._call(decl, parens)
+            case PartialCallDecl(CallDecl(ref, typed_args, _)):
+                arg_strs = (_pretty_callable(ref), *(self(a.expr, parens=False, unwrap_lit=True) for a in typed_args))
+                return f"UnstableFn({', '.join(arg_strs)})", "fn"
+            case PyObjectDecl(value):
+                return repr(value) if unwrap_lit else f"PyObject({value!r})", "PyObject"
+            case ActionCommandDecl(action):
+                return self(action), "action"
+            case RewriteDecl(_, lhs, rhs, conditions) | BiRewriteDecl(_, lhs, rhs, conditions):
+                args = ", ".join(map(self, (rhs, *conditions)))
+                fn = "rewrite" if isinstance(decl, RewriteDecl) else "birewrite"
+                return f"{fn}({self(lhs)}).to({args})", "rewrite"
+            case RuleDecl(head, body, name):
+                l = ", ".join(map(self, body))
+                if name:
+                    l += f", name={name}"
+                r = ", ".join(map(self, head))
+                return f"rule({l}).then({r})", "rule"
+            case SetDecl(_, lhs, rhs):
+                return f"set_({self(lhs)}).to({self(rhs)})", "action"
+            case UnionDecl(_, lhs, rhs):
+                return f"union({self(lhs)}).with_({self(rhs)})", "action"
+            case LetDecl(name, expr):
+                return f"let({name!r}, {self(expr.expr)})", "action"
+            case ExprActionDecl(expr):
+                return self(expr.expr), "action"
+            case ExprFactDecl(expr):
+                return self(expr.expr), "fact"
+            case ChangeDecl(_, expr, change):
+                return f"{change}({self(expr)})", "action"
+            case PanicDecl(s):
+                return f"panic({s!r})", "action"
+            case EqDecl(_, exprs):
+                first, *rest = exprs
+                return f"eq({self(first)}).to({', '.join(map(self, rest))})", "fact"
+            case RulesetDecl(rules):
+                if ruleset_name:
+                    return f"ruleset(name={ruleset_name!r})", f"ruleset_{ruleset_name}"
+                args = ", ".join(map(self, rules))
+                return f"ruleset({args})", "ruleset"
+            case CombinedRulesetDecl(rulesets):
+                if ruleset_name:
+                    rulesets = (*rulesets, f"name={ruleset_name!r})")
+                return f"unstable_combine_rulesets({', '.join(rulesets)})", "combined_ruleset"
+            case SaturateDecl(schedule):
+                return f"{self(schedule, parens=True)}.saturate()", "schedule"
+            case RepeatDecl(schedule, times):
+                return f"{self(schedule, parens=True)} * {times}", "schedule"
+            case SequenceDecl(schedules):
+                if len(schedules) == 2:
+                    return f"{self(schedules[0], parens=True)} + {self(schedules[1], parens=True)}", "schedule"
+                args = ", ".join(map(self, schedules))
+                return f"seq({args})", "schedule"
+            case RunDecl(ruleset_name, until):
+                ruleset = self.decls._rulesets[ruleset_name]
+                ruleset_str = self(ruleset, ruleset_name=ruleset_name)
+                if not until:
+                    return ruleset_str, "schedule"
+                args = ", ".join(map(self, until))
+                return f"run({ruleset_str}, {args})", "schedule"
+        assert_never(decl)
+
+    def _call(
+        self,
+        decl: CallDecl,
+        parens: bool,
+    ) -> tuple[str, str]:
+        """
+        Pretty print the call. Also returns if it was saved as a name.
+
+        :param parens: If true, wrap the call in parens if it is a binary method call.
+        """
+        args = [a.expr for a in decl.args]
+        ref = decl.callable
+        # Special case !=
+        if decl.callable == FunctionRef("!="):
+            l, r = self(args[0]), self(args[1])
+            return f"ne({l}).to({r})", "Unit"
+        function_decl = self.decls.get_callable_decl(ref).to_function_decl()
+        signature = function_decl.signature
+
+        # Determine how many of the last arguments are defaults, by iterating from the end and comparing the arg with the default
+        n_defaults = 0
+        # Dont try counting defaults for function application
+        if isinstance(signature, FunctionSignature):
+            for arg, default in zip(
+                reversed(args), reversed(signature.arg_defaults), strict=not signature.var_arg_type
+            ):
+                if arg != default:
+                    break
+                n_defaults += 1
+        if n_defaults:
+            args = args[:-n_defaults]
+
+        # If this is a function application, the type is the first type arg of the function object
+        if signature == "fn-app":
+            tp_name = decl.args[0].tp.args[0].name
+        else:
+            assert isinstance(signature, FunctionSignature)
+            tp_name = signature.semantic_return_type.name
+        if isinstance(signature, FunctionSignature) and signature.mutates:
+            first_arg = args[0]
+            expr_str = self(first_arg)
+            # copy an identifier expression iff it has multiple parents (b/c then we can't mutate it directly)
+            has_multiple_parents = self.parents[first_arg] > 1
+            self.names[decl] = expr_name = self._name_expr(tp_name, expr_str, copy_identifier=has_multiple_parents)
+            # Set the first arg to be the name of the mutated arg and return the name
+            args[0] = VarDecl(expr_name)
+        else:
+            expr_name = None
+        res = self._call_inner(ref, args, decl.bound_tp_params, parens)
+        expr = (
+            f"{res[0]}({', '.join(self(a, parens=False, unwrap_lit=True) for a in res[1])})"
+            if isinstance(res, tuple)
+            else res
+        )
+        # If we have a name, then we mutated
+        if expr_name:
+            self.statements.append(expr)
+            return expr_name, tp_name
+        return expr, tp_name
+
+    def _call_inner(  # noqa: PLR0911
+        self, ref: CallableRef, args: list[ExprDecl], bound_tp_params: tuple[JustTypeRef, ...] | None, parens: bool
+    ) -> tuple[str, list[ExprDecl]] | str:
+        """
+        Pretty print the call, returning either the full function call or a tuple of the function and the args.
+        """
+        match ref:
+            case FunctionRef(name):
+                return name, args
+            case ClassMethodRef(class_name, method_name):
+                fn_str = str(JustTypeRef(class_name, bound_tp_params or ()))
+                if method_name != "__init__":
+                    fn_str += f".{method_name}"
+                return fn_str, args
+            case MethodRef(_class_name, method_name):
+                slf, *args = args
+                slf = self(slf, parens=True)
+                match method_name:
+                    case _ if method_name in UNARY_METHODS:
+                        expr = f"{UNARY_METHODS[method_name]}{slf}"
+                        return f"({expr})" if parens else expr
+                    case _ if method_name in BINARY_METHODS:
+                        expr = f"{slf} {BINARY_METHODS[method_name]} {self(args[0], parens=True, unwrap_lit=True)}"
+                        return f"({expr})" if parens else expr
+                    case "__getitem__":
+                        return f"{slf}[{self(args[0], unwrap_lit=True)}]"
+                    case "__call__":
+                        return slf, args
+                    case "__delitem__":
+                        return f"del {slf}[{self(args[0], unwrap_lit=True)}]"
+                    case "__setitem__":
+                        return f"{slf}[{self(args[0], unwrap_lit=True)}] = {self(args[1], unwrap_lit=True)}"
+                    case _:
+                        return f"{slf}.{method_name}", args
+            case ConstantRef(name):
+                return name
+            case ClassVariableRef(class_name, variable_name):
+                return f"{class_name}.{variable_name}"
+            case PropertyRef(_class_name, property_name):
+                return f"{self(args[0], parens=True)}.{property_name}"
+        assert_never(ref)
+
+    def _generate_name(self, typ: str) -> str:
+        self._gen_name_types[typ] += 1
+        return f"_{typ}_{self._gen_name_types[typ]}"
+
+    def _name_expr(self, tp_name: str, expr_str: str, copy_identifier: bool) -> str:
+        # tp_name =
+        # If the thing we are naming is already a variable, we don't need to name it
+        if expr_str.isidentifier():
+            if copy_identifier:
+                name = self._generate_name(tp_name)
+                self.statements.append(f"{name} = copy({expr_str})")
+            else:
+                name = expr_str
+        else:
+            name = self._generate_name(tp_name)
+            self.statements.append(f"{name} = {expr_str}")
+        return name
+
+
+def _pretty_callable(ref: CallableRef) -> str:
+    """
+    Returns a function call as a string.
+    """
+    match ref:
+        case FunctionRef(name):
+            return name
+        case (
+            ClassMethodRef(class_name, method_name)
+            | MethodRef(class_name, method_name)
+            | PropertyRef(class_name, method_name)
+        ):
+            return f"{class_name}.{method_name}"
+        case ConstantRef(_):
+            msg = "Constants should not be callable"
+            raise NotImplementedError(msg)
+        case ClassVariableRef(_, _):
+            msg = "Class variables should not be callable"
+            raise NotADirectoryError(msg)
+    assert_never(ref)
+
+
+def _plot_line_length(expr: object):  # pragma: no cover
+    """
+    Plots the number of line lengths based on different max lengths
+    """
+    global MAX_LINE_LENGTH, LINE_DIFFERENCE
+    import altair as alt
+    import pandas as pd
+
+    sizes = []
+    for line_length in range(40, 180, 10):
+        MAX_LINE_LENGTH = line_length
+        for diff in range(0, 40, 5):
+            LINE_DIFFERENCE = diff
+            new_l = len(str(expr).split())
+            sizes.append((line_length, diff, new_l))
+
+    df = pd.DataFrame(sizes, columns=["MAX_LINE_LENGTH", "LENGTH_DIFFERENCE", "n"])  # noqa: PD901
+
+    return alt.Chart(df).mark_rect().encode(x="MAX_LINE_LENGTH:O", y="LENGTH_DIFFERENCE:O", color="n:Q")