[lldb] Proof of concept data formatter compiler for Python

[kastiglione]

A compiler from Python to the assembly syntax of the lldb data formatter bytecode.

[llvmbot]

@llvm/pr-subscribers-lldb

Author: Dave Lee (kastiglione)

Changes

---

Full diff: https://github.com/llvm/llvm-project/pull/113734.diff

2 Files Affected:

• (added) lldb/examples/formatter-bytecode/optional_summary.py (+14)
• (added) lldb/examples/formatter-bytecode/python_to_assembly.py (+145)

diff --git a/lldb/examples/formatter-bytecode/optional_summary.py b/lldb/examples/formatter-bytecode/optional_summary.py
new file mode 100644
index 00000000000000..68e672d86613d1
--- /dev/null
+++ b/lldb/examples/formatter-bytecode/optional_summary.py
@@ -0,0 +1,14 @@
+def OptionalSummaryProvider(valobj, _):
+    failure = 2
+    storage = valobj.GetChildMemberWithName("Storage")
+    hasVal = storage.GetChildMemberWithName("hasVal").GetValueAsUnsigned(failure)
+    if hasVal == failure:
+        return "<could not read Optional>"
+
+    if hasVal == 0:
+        return "None"
+
+    underlying_type = storage.GetType().GetTemplateArgumentType(0)
+    value = storage.GetChildMemberWithName("value")
+    value = value.Cast(underlying_type)
+    return value.GetSummary()
diff --git a/lldb/examples/formatter-bytecode/python_to_assembly.py b/lldb/examples/formatter-bytecode/python_to_assembly.py
new file mode 100755
index 00000000000000..6e2adbe093fdac
--- /dev/null
+++ b/lldb/examples/formatter-bytecode/python_to_assembly.py
@@ -0,0 +1,145 @@
+#!/usr/bin/python3
+
+import ast
+import io
+import sys
+from typing import Any
+
+BUILTINS = {
+    "Cast": "@cast",
+    "GetChildMemberWithName": "@get_child_with_name",
+    "GetSummary": "@get_summary",
+    "GetTemplateArgumentType": "@get_template_argument_type",
+    "GetType": "@get_type",
+    "GetValueAsUnsigned": "@get_value_as_unsigned",
+}
+
+COMPS = {
+    ast.Eq: "=",
+    ast.NotEq: "!=",
+    ast.Lt: "<",
+    ast.LtE: "=<",
+    ast.Gt: ">",
+    ast.GtE: "=>",
+}
+
+class Compiler(ast.NodeVisitor):
+    # Track the stack index of locals variables.
+    #
+    # This is essentially an ordered dictionary, where the key is an index on
+    # the stack, and the value is the name of the variable whose value is at
+    # that index.
+    #
+    # Ex: `locals[0]` is the name of the first value pushed on the stack, etc.
+    locals: list[str]
+
+    buffer: io.StringIO
+    final_buffer: io.StringIO
+
+    def __init__(self) -> None:
+        self.locals = []
+        self.buffer = io.StringIO()
+        self.final_buffer = io.StringIO()
+
+    def visit_FunctionDef(self, node: ast.FunctionDef) -> None:
+        # Initialize `locals` with the (positional) arguments.
+        self.locals = [arg.arg for arg in node.args.args]
+        self.generic_visit(node)
+        self.locals.clear()
+
+    def visit_Compare(self, node: ast.Compare) -> None:
+        self.visit(node.left)
+        # XXX: Does not handle multiple comparisons, ex: `0 < x < 10`
+        self.visit(node.comparators[0])
+        self._output(COMPS[type(node.ops[0])])
+
+    def visit_If(self, node: ast.If) -> None:
+        self.visit(node.test)
+
+        # Does the body `return`?
+        has_return = any(isinstance(x, ast.Return) for x in node.body)
+
+        self._output("{")
+        self._visit_each(node.body)
+        if not node.orelse and not has_return:
+            # No else, and no early exit: a simple `if`
+            self._output("} if")
+            return
+
+        self._output("}")
+        if node.orelse:
+            # Handle else.
+            self._output("{")
+            self._visit_each(node.orelse)
+            self._output("} ifelse")
+        elif has_return:
+            # Convert early exit into an `ifelse`.
+            self._output("{")
+            self._output("} ifelse", final=True)
+
+    def visit_Constant(self, node: ast.Constant) -> None:
+        if isinstance(node.value, str):
+            self._output(f'"{node.value}"')
+        elif isinstance(node.value, bool):
+            self._output(int(node.value))
+        else:
+            self._output(node.value)
+
+    def visit_Call(self, node: ast.Call) -> None:
+        if isinstance(node.func, ast.Attribute):
+            # The receiver is the left hande side of the dot.
+            receiver = node.func.value
+            method = node.func.attr
+            if selector := BUILTINS.get(method):
+                # Visit the method's receiver to have its value on the stack.
+                self.visit(receiver)
+                # Visit the args to position them on the stack.
+                self._visit_each(node.args)
+                self._output(f"{selector} call")
+            else:
+                # TODO: fail
+                print(f"error: unsupported method {node.func.attr}", file=sys.stderr)
+
+    def visit_Assign(self, node: ast.Assign) -> None:
+        # Visit RHS first, putting values on the stack.
+        self.visit(node.value)
+        # Determine the name(s). Either a single Name, or a Tuple of Names.
+        target = node.targets[0]
+        if isinstance(target, ast.Name):
+            names = [target.id]
+        elif isinstance(target, ast.Tuple):
+            # These tuple elements are Name nodes.
+            names = [x.id for x in target.elts]
+
+        # Forget any previous bindings of these names.
+        # Their values are orphaned on the stack.
+        for local in self.locals:
+            if local in names:
+                old_idx = self.locals.index(local)
+                self.locals[old_idx] = ""
+
+        self.locals.extend(names)
+
+    def visit_Name(self, node: ast.Name) -> None:
+        idx = self.locals.index(node.id)
+        self._output(f"{idx} pick # {node.id}")
+
+    def _visit_each(self, nodes: list[ast.AST]) -> None:
+        for child in nodes:
+            self.visit(child)
+
+    def _output(self, x: Any, final: bool = False) -> None:
+        dest = self.final_buffer if final else self.buffer
+        print(x, file=dest)
+
+    @property
+    def output(self) -> str:
+        return compiler.buffer.getvalue() + compiler.final_buffer.getvalue()
+
+
+if __name__ == "__main__":
+    with open(sys.argv[1]) as f:
+        root = ast.parse(f.read())
+    compiler = Compiler()
+    compiler.visit(root)
+    print(compiler.output)

[github-actions]

⚠️ Python code formatter, darker found issues in your code. ⚠️

You can test this locally with the following command:

darker --check --diff -r eb9f4756bc3daaa4b19f4f46521dc05180814de4...57223942e91c47d0a61b148a65247cd9cbb16496 lldb/examples/formatter-bytecode/optional_summary.py lldb/examples/formatter-bytecode/python_to_assembly.py

View the diff from darker here.

--- python_to_assembly.py	2024-10-25 19:56:12.000000 +0000
+++ python_to_assembly.py	2024-10-25 19:59:11.920862 +0000
@@ -20,10 +20,11 @@
     ast.Lt: "<",
     ast.LtE: "=<",
     ast.Gt: ">",
     ast.GtE: "=>",
 }
+
 
 class Compiler(ast.NodeVisitor):
     # Track the stack index of locals variables.
     #
     # This is essentially an ordered dictionary, where the key is an index on