Support for `pandas.DataFrame.eval()`?

[gipert]

I've just ran a quick test and it seems to me that pandas.DataFrame.eval() is not supported. Is that correct? Is there a way to evaluate string expressions on dataframes containing Awkward arrays?

[jpivarski]

Is it an error message saying that it's not supported?

If Pandas is just running Python eval with column names loaded into the namespace (as the documentation suggests), then I don't see why those strings couldn't operate directly on Awkward Arrays.

Maybe the AwkwardSeries objects need to be unwrapped before passing to eval and the result needs to be re-wrapped? (This is a question for @douglasdavis.)

[gipert]

I don't see why those strings couldn't operate directly on Awkward Arrays.

Indeed that was my thinking. This is what happens:

>>> import awkward_pandas as akpd
>>> import pandas as pd
>>> import awkward as ak
>>> df = pd.DataFrame(
    ...:   {
    ...:     "a": [1, 2, 3, 4],
    ...:     "b": akpd.from_awkward(ak.Array([[1, 2], [], [3], [4, 5, 6]]))
    ...:   }
    ...: )
>>> df.eval("b*2")
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[10], line 1
----> 1 df.eval("b*2")

File ~/.virtualenvs/legend/lib/python3.11/site-packages/pandas/core/frame.py:4566, in DataFrame.eval(self, expr, inplace, **kwargs)
   4563     kwargs["target"] = self
   4564 kwargs["resolvers"] = tuple(kwargs.get("resolvers", ())) + resolvers
-> 4566 return _eval(expr, inplace=inplace, **kwargs)

File ~/.virtualenvs/legend/lib/python3.11/site-packages/pandas/core/computation/eval.py:336, in eval(expr, parser, engine, local_dict, global_dict, resolvers, level, target, inplace)
    327 # get our (possibly passed-in) scope
    328 env = ensure_scope(
    329     level + 1,
    330     global_dict=global_dict,
   (...)
    333     target=target,
    334 )
--> 336 parsed_expr = Expr(expr, engine=engine, parser=parser, env=env)
    338 if engine == "numexpr" and (
    339     is_extension_array_dtype(parsed_expr.terms.return_type)
    340     or getattr(parsed_expr.terms, "operand_types", None) is not None
   (...)
    344     )
    345 ):
    346     warnings.warn(
    347         "Engine has switched to 'python' because numexpr does not support "
    348         "extension array dtypes. Please set your engine to python manually.",
    349         RuntimeWarning,
    350         stacklevel=find_stack_level(),
    351     )

File ~/.virtualenvs/legend/lib/python3.11/site-packages/pandas/core/computation/expr.py:809, in Expr.__init__(self, expr, engine, parser, env, level)
    807 self.parser = parser
    808 self._visitor = PARSERS[parser](self.env, self.engine, self.parser)
--> 809 self.terms = self.parse()

File ~/.virtualenvs/legend/lib/python3.11/site-packages/pandas/core/computation/expr.py:828, in Expr.parse(self)
    824 def parse(self):
    825     """
    826     Parse an expression.
    827     """
--> 828     return self._visitor.visit(self.expr)

File ~/.virtualenvs/legend/lib/python3.11/site-packages/pandas/core/computation/expr.py:415, in BaseExprVisitor.visit(self, node, **kwargs)
    413 method = f"visit_{type(node).__name__}"
    414 visitor = getattr(self, method)
--> 415 return visitor(node, **kwargs)

File ~/.virtualenvs/legend/lib/python3.11/site-packages/pandas/core/computation/expr.py:421, in BaseExprVisitor.visit_Module(self, node, **kwargs)
    419     raise SyntaxError("only a single expression is allowed")
    420 expr = node.body[0]
--> 421 return self.visit(expr, **kwargs)

File ~/.virtualenvs/legend/lib/python3.11/site-packages/pandas/core/computation/expr.py:415, in BaseExprVisitor.visit(self, node, **kwargs)
    413 method = f"visit_{type(node).__name__}"
    414 visitor = getattr(self, method)
--> 415 return visitor(node, **kwargs)

File ~/.virtualenvs/legend/lib/python3.11/site-packages/pandas/core/computation/expr.py:424, in BaseExprVisitor.visit_Expr(self, node, **kwargs)
    423 def visit_Expr(self, node, **kwargs):
--> 424     return self.visit(node.value, **kwargs)

File ~/.virtualenvs/legend/lib/python3.11/site-packages/pandas/core/computation/expr.py:415, in BaseExprVisitor.visit(self, node, **kwargs)
    413 method = f"visit_{type(node).__name__}"
    414 visitor = getattr(self, method)
--> 415 return visitor(node, **kwargs)

File ~/.virtualenvs/legend/lib/python3.11/site-packages/pandas/core/computation/expr.py:537, in BaseExprVisitor.visit_BinOp(self, node, **kwargs)
    535 op, op_class, left, right = self._maybe_transform_eq_ne(node)
    536 left, right = self._maybe_downcast_constants(left, right)
--> 537 return self._maybe_evaluate_binop(op, op_class, left, right)

File ~/.virtualenvs/legend/lib/python3.11/site-packages/pandas/core/computation/expr.py:507, in BaseExprVisitor._maybe_evaluate_binop(self, op, op_class, lhs, rhs, eval_in_python, maybe_eval_in_python)
    504 res = op(lhs, rhs)
    506 if res.has_invalid_return_type:
--> 507     raise TypeError(
    508         f"unsupported operand type(s) for {res.op}: "
    509         f"'{lhs.type}' and '{rhs.type}'"
    510     )
    512 if self.engine != "pytables" and (
    513     res.op in CMP_OPS_SYMS
    514     and getattr(lhs, "is_datetime", False)
   (...)
    517     # all date ops must be done in python bc numexpr doesn't work
    518     # well with NaT
    519     return self._maybe_eval(res, self.binary_ops)

TypeError: unsupported operand type(s) for *: 'awkward' and '<class 'int'>'

[gipert]

Just for the context: I'm writing some code to evaluate algebraic expressions from a config file on tables made by jagged and rectangular columns. I was hoping to be able to write almost no code by using pandas.eval()...