Multiline pd.eval() seems to be broken

[resposit]

Code Sample:

In [93]: df
Out[93]:
   A  B
0  1  2
1  2  3
2  3  4
3  4  5

In [102]: pd.eval("""A = df.A - df.B
     ...: B = df.A + df.B
     ...: """,target=pd.DataFrame())

Problem description

....
/usr/local/lib64/python3.6/site-packages/pandas/core/computation/scope.py in resolve(self, key, is_local)
    201                 from pandas.core.computation.ops import UndefinedVariableError
    202
--> 203                 raise UndefinedVariableError(key, is_local)
    204
    205     def swapkey(self, old_key: str, new_key: str, new_value=None):

UndefinedVariableError: name 'df' is not defined

Expected Output

Out[106]:
   A  B
0 -1  1
1 -1  2
2 -1  3
3 -1  4

Output of pd.show_versions()

INSTALLED VERSIONS

commit : None
python : 3.6.8.final.0
python-bits : 64
OS : Linux
OS-release : 3.10.0-1062.9.1.el7.x86_64
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : None
LANG : en_US.UTF-8
LOCALE : en_US.UTF-8

pandas : 1.0.1
numpy : 1.18.1
pytz : 2019.3
dateutil : 2.8.0
pip : 9.0.3
setuptools : 45.2.0
Cython : None
pytest : None
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : 4.5.0
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 2.11.1
IPython : 7.12.0
pandas_datareader: None
bs4 : 4.8.2
bottleneck : None
fastparquet : None
gcsfs : None
lxml.etree : 4.5.0
matplotlib : 3.1.2
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pytables : None
pytest : None
pyxlsb : None
s3fs : None
scipy : 1.4.1
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : None
xlwt : None
xlsxwriter : None
numba : None

[MarcoGorelli]

What do you mean by "broken" - did the example you provided used to work in previous versions?

It looks like you're trying to pass multiple expressions to pandas.eval - only a single one is allowed, however (it doesn't seems like this is currently documented in https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.eval.html - are you interested in submitting a PR?)

Finally, even with a single line, I don't believe this example would work: [EDIT: with target=pd.DataFrame(), as in the OP's example, it does indeed work]

>>> pd.eval("A = df.A - df.B")
ValueError: cannot assign without a target object

Perhaps you meant

>>> df.eval("A = A-B")

? (even then, only a single expression is allowed) (EDIT: I wasn't quite right about this: only a single expression is allowed, but that expression may contain multiple lines, but only in DataFrame.eval and not in pandas.eval)

[resposit]

Hi Marco,
I'm very new to pandas and don't know if this used to work in previous version.
Anyway, as far as I understood, you can actually pass multiple expressions (only assignements) to .eval() in a multiline string, as discussed here:
https://stackoverflow.com/questions/53779986/dynamic-expression-evaluation-in-pandas-using-pd-eval

[MarcoGorelli]

Thanks @resposit - are you referring to the "Multiline Queries and Assignment" section?

If so, the command they give is a bit different (they're using DataFrame.eval rather than pd.eval, and they're prepending a @ symbol to the df variable) - the following still works:

>>> df.eval("""
... A = @df.A - @df.B
... B = @df.A + @df.B
... """
... )
   A  B
0 -1  3
1 -1  5
2 -1  7
3 -1  9

[resposit]

Hi @MarcoGorelli
yes I was referring to that section. I know they're using DataFrame.eval().
In fact, using DataFrame.eval() you can also simplify your code like this:

In [108]: df.eval('''
     ...: A = A-B
     ...: B = A+B
     ...: ''')
Out[108]:
   A  B
0 -1  1
1 -1  2
2 -1  3
3 -1  4

I'm instead interested in using pd.eval() which, in principal, should also support multiline assignements.

[MarcoGorelli]

I'm instead interested in using pd.eval() which, in principal, should also support multiline assignements.

OK, yes - if

df.eval(
    """
    A = A - B
    B = A + B
    """
)

works, there's no reason why

pd.eval(
    """
    A = df.A - df.B
    B = df.A + df.B
    """,
    target=pd.DataFrame(),
)

(which is what you originally wrote - apologies) shouldn't.

Looking at the source code, I don't know if the fact that it doesn't is intentional, pandas/core/computation/eval.py contains:

    if multi_line and target is None:
        raise ValueError(
            "multi-line expressions are only valid in the "
            "context of data, use DataFrame.eval"
        )

which suggests that, if target isn't None, then multi-line expressions are expected work. Will look into it