Refactor for Dialect.

Author: richardkiss

clvm/__init__.py

@@ -1,6 +1,7 @@
 from .SExp import SExp
+from .dialect import Dialect  # noqa
 from .operators import (  # noqa
-    QUOTE_ATOM,
+    QUOTE_ATOM,  # deprecated
     KEYWORD_TO_ATOM,
     KEYWORD_FROM_ATOM,
 )

clvm/chainable_multi_op_fn.py

@@ -0,0 +1,26 @@
+from dataclasses import dataclass
+from typing import Optional, Tuple
+
+from .types import CLVMObjectType, MultiOpFn, OperatorDict
+
+
+@dataclass
+class ChainableMultiOpFn:
+    """
+    This structure handles clvm operators. Given an atom, it looks it up in a `dict`, then
+    falls back to calling `unknown_op_handler`.
+    """
+    op_lookup: OperatorDict
+    unknown_op_handler: MultiOpFn
+
+    def __call__(
+        self, op: bytes, arguments: CLVMObjectType, max_cost: Optional[int] = None
+    ) -> Tuple[int, CLVMObjectType]:
+        f = self.op_lookup.get(op)
+        if f:
+            try:
+                return f(arguments)
+            except TypeError:
+                # some operators require `max_cost`
+                return f(arguments, max_cost)
+        return self.unknown_op_handler(op, arguments, max_cost)

clvm/chia_dialect.py

@@ -0,0 +1,37 @@
+from .casts import int_to_bytes
+from .dialect import ConversionFn, Dialect, new_dialect, opcode_table_for_backend
+
+KEYWORDS = (
+    # core opcodes 0x01-x08
+    ". q a i c f r l x "
+
+    # opcodes on atoms as strings 0x09-0x0f
+    "= >s sha256 substr strlen concat . "
+
+    # opcodes on atoms as ints 0x10-0x17
+    "+ - * / divmod > ash lsh "
+
+    # opcodes on atoms as vectors of bools 0x18-0x1c
+    "logand logior logxor lognot . "
+
+    # opcodes for bls 1381 0x1d-0x1f
+    "point_add pubkey_for_exp . "
+
+    # bool opcodes 0x20-0x23
+    "not any all . "
+
+    # misc 0x24
+    "softfork "
+).split()
+
+KEYWORD_FROM_ATOM = {int_to_bytes(k): v for k, v in enumerate(KEYWORDS)}
+KEYWORD_TO_ATOM = {v: k for k, v in KEYWORD_FROM_ATOM.items()}
+
+
+def chia_dialect(strict: bool, to_python: ConversionFn, backend=None) -> Dialect:
+    quote_kw = KEYWORD_TO_ATOM["q"]
+    apply_kw = KEYWORD_TO_ATOM["a"]
+    dialect = new_dialect(quote_kw, apply_kw, strict, to_python, backend=backend)
+    table = opcode_table_for_backend(KEYWORD_TO_ATOM, backend=backend)
+    dialect.update(table)
+    return dialect

clvm/dialect.py

@@ -0,0 +1,159 @@
+from typing import Callable, Optional, Tuple
+
+try:
+    import clvm_rs
+except ImportError:
+    clvm_rs = None
+
+from . import core_ops, more_ops
+from .chainable_multi_op_fn import ChainableMultiOpFn
+from .handle_unknown_op import (
+    handle_unknown_op_softfork_ready,
+    handle_unknown_op_strict,
+)
+from .run_program import _run_program
+from .types import CLVMObjectType, ConversionFn, MultiOpFn, OperatorDict
+
+
+OP_REWRITE = {
+    "+": "add",
+    "-": "subtract",
+    "*": "multiply",
+    "/": "div",
+    "i": "if",
+    "c": "cons",
+    "f": "first",
+    "r": "rest",
+    "l": "listp",
+    "x": "raise",
+    "=": "eq",
+    ">": "gr",
+    ">s": "gr_bytes",
+}
+
+
+def op_table_for_module(mod):
+
+    # python-implemented operators don't take `max_cost` and rust-implemented operators do
+    # So we make the `max_cost` operator optional with this trick
+    # TODO: have python-implemented ops also take `max_cost` and unify the API.
+
+    def elide_max_cost(f):
+        def inner_op(sexp, max_cost=None):
+            try:
+                return f(sexp, max_cost)
+            except TypeError:
+                return f(sexp)
+        return inner_op
+
+    return {k: elide_max_cost(v) for k, v in mod.__dict__.items() if k.startswith("op_")}
+
+
+def op_imp_table_for_backend(backend):
+    if backend is None and clvm_rs:
+        backend = "native"
+
+    if backend == "native":
+        if clvm_rs is None:
+            raise RuntimeError("native backend not installed")
+        return clvm_rs.native_opcodes_dict()
+
+    table = {}
+    table.update(op_table_for_module(core_ops))
+    table.update(op_table_for_module(more_ops))
+    return table
+
+
+def op_atom_to_imp_table(op_imp_table, keyword_to_atom, op_rewrite=OP_REWRITE):
+    op_atom_to_imp_table = {}
+    for op, bytecode in keyword_to_atom.items():
+        op_name = "op_%s" % op_rewrite.get(op, op)
+        op_f = op_imp_table.get(op_name)
+        if op_f:
+            op_atom_to_imp_table[bytecode] = op_f
+    return op_atom_to_imp_table
+
+
+def opcode_table_for_backend(keyword_to_atom, backend):
+    op_imp_table = op_imp_table_for_backend(backend)
+    return op_atom_to_imp_table(op_imp_table, keyword_to_atom)
+
+
+class Dialect:
+    def __init__(
+        self,
+        quote_kw: bytes,
+        apply_kw: bytes,
+        multi_op_fn: MultiOpFn,
+        to_python: ConversionFn,
+    ):
+        self.quote_kw = quote_kw
+        self.apply_kw = apply_kw
+        self.opcode_lookup = dict()
+        self.multi_op_fn = ChainableMultiOpFn(self.opcode_lookup, multi_op_fn)
+        self.to_python = to_python
+
+    def update(self, d: OperatorDict) -> None:
+        self.opcode_lookup.update(d)
+
+    def clear(self) -> None:
+        self.opcode_lookup.clear()
+
+    def run_program(
+        self,
+        program: CLVMObjectType,
+        env: CLVMObjectType,
+        max_cost: int,
+        pre_eval_f: Optional[
+            Callable[[CLVMObjectType, CLVMObjectType], Tuple[int, CLVMObjectType]]
+        ] = None,
+    ) -> Tuple[int, CLVMObjectType]:
+        cost, r = _run_program(
+            program,
+            env,
+            self.multi_op_fn,
+            self.quote_kw,
+            self.apply_kw,
+            max_cost,
+            pre_eval_f,
+        )
+        return cost, self.to_python(r)
+
+
+def native_new_dialect(
+    quote_kw: bytes, apply_kw: bytes, strict: bool, to_python: ConversionFn
+) -> Dialect:
+    unknown_op_callback = (
+        clvm_rs.NATIVE_OP_UNKNOWN_STRICT
+        if strict
+        else clvm_rs.NATIVE_OP_UNKNOWN_NON_STRICT
+    )
+    dialect = clvm_rs.Dialect(
+        quote_kw,
+        apply_kw,
+        unknown_op_callback,
+        to_python=to_python,
+    )
+    return dialect
+
+
+def python_new_dialect(
+    quote_kw: bytes, apply_kw: bytes, strict: bool, to_python: ConversionFn
+) -> Dialect:
+    unknown_op_callback = (
+        handle_unknown_op_strict if strict else handle_unknown_op_softfork_ready
+    )
+    dialect = Dialect(
+        quote_kw,
+        apply_kw,
+        unknown_op_callback,
+        to_python=to_python,
+    )
+    return dialect
+
+
+def new_dialect(quote_kw: bytes, apply_kw: bytes, strict: bool, to_python: ConversionFn, backend=None):
+    if backend is None:
+        backend = "python" if clvm_rs is None else "native"
+    backend_f = native_new_dialect if backend == "native" else python_new_dialect
+    return backend_f(quote_kw, apply_kw, strict, to_python)

clvm/handle_unknown_op.py

@@ -0,0 +1,124 @@
+from typing import Tuple
+
+from .CLVMObject import CLVMObject
+from .EvalError import EvalError
+
+from .costs import (
+    ARITH_BASE_COST,
+    ARITH_COST_PER_BYTE,
+    ARITH_COST_PER_ARG,
+    MUL_BASE_COST,
+    MUL_COST_PER_OP,
+    MUL_LINEAR_COST_PER_BYTE,
+    MUL_SQUARE_COST_PER_BYTE_DIVIDER,
+    CONCAT_BASE_COST,
+    CONCAT_COST_PER_ARG,
+    CONCAT_COST_PER_BYTE,
+)
+
+
+def handle_unknown_op_strict(op, arguments, _max_cost=None):
+    raise EvalError("unimplemented operator", arguments.to(op))
+
+
+def args_len(op_name, args):
+    for arg in args.as_iter():
+        if arg.pair:
+            raise EvalError("%s requires int args" % op_name, arg)
+        yield len(arg.as_atom())
+
+
+# unknown ops are reserved if they start with 0xffff
+# otherwise, unknown ops are no-ops, but they have costs. The cost is computed
+# like this:
+
+# byte index (reverse):
+# | 4 | 3 | 2 | 1 | 0          |
+# +---+---+---+---+------------+
+# | multiplier    |XX | XXXXXX |
+# +---+---+---+---+---+--------+
+#  ^               ^    ^
+#  |               |    + 6 bits ignored when computing cost
+# cost_multiplier  |
+#                  + 2 bits
+#                    cost_function
+
+# 1 is always added to the multiplier before using it to multiply the cost, this
+# is since cost may not be 0.
+
+# cost_function is 2 bits and defines how cost is computed based on arguments:
+# 0: constant, cost is 1 * (multiplier + 1)
+# 1: computed like operator add, multiplied by (multiplier + 1)
+# 2: computed like operator mul, multiplied by (multiplier + 1)
+# 3: computed like operator concat, multiplied by (multiplier + 1)
+
+# this means that unknown ops where cost_function is 1, 2, or 3, may still be
+# fatal errors if the arguments passed are not atoms.
+
+
+def handle_unknown_op_softfork_ready(
+    op: bytes, args: CLVMObject, max_cost: int
+) -> Tuple[int, CLVMObject]:
+    # any opcode starting with ffff is reserved (i.e. fatal error)
+    # opcodes are not allowed to be empty
+    if len(op) == 0 or op[:2] == b"\xff\xff":
+        raise EvalError("reserved operator", args.to(op))
+
+    # all other unknown opcodes are no-ops
+    # the cost of the no-ops is determined by the opcode number, except the
+    # 6 least significant bits.
+
+    cost_function = (op[-1] & 0b11000000) >> 6
+    # the multiplier cannot be 0. it starts at 1
+
+    if len(op) > 5:
+        raise EvalError("invalid operator", args.to(op))
+
+    cost_multiplier = int.from_bytes(op[:-1], "big", signed=False) + 1
+
+    # 0 = constant
+    # 1 = like op_add/op_sub
+    # 2 = like op_multiply
+    # 3 = like op_concat
+    if cost_function == 0:
+        cost = 1
+    elif cost_function == 1:
+        # like op_add
+        cost = ARITH_BASE_COST
+        arg_size = 0
+        for length in args_len("unknown op", args):
+            arg_size += length
+            cost += ARITH_COST_PER_ARG
+        cost += arg_size * ARITH_COST_PER_BYTE
+    elif cost_function == 2:
+        # like op_multiply
+        cost = MUL_BASE_COST
+        operands = args_len("unknown op", args)
+        try:
+            vs = next(operands)
+            for rs in operands:
+                cost += MUL_COST_PER_OP
+                cost += (rs + vs) * MUL_LINEAR_COST_PER_BYTE
+                cost += (rs * vs) // MUL_SQUARE_COST_PER_BYTE_DIVIDER
+                # this is an estimate, since we don't want to actually multiply the
+                # values
+                vs += rs
+        except StopIteration:
+            pass
+
+    elif cost_function == 3:
+        # like concat
+        cost = CONCAT_BASE_COST
+        length = 0
+        for arg in args.as_iter():
+            if arg.pair:
+                raise EvalError("unknown op on list", arg)
+            cost += CONCAT_COST_PER_ARG
+            length += len(arg.atom)
+        cost += length * CONCAT_COST_PER_BYTE
+
+    cost *= cost_multiplier
+    if cost >= 2**32:
+        raise EvalError("invalid operator", args.to(op))
+
+    return (cost, args.to(b""))