I don't understand this issue? Can you explain more elaborately?

It seems the issue is that Polars will "iterate" a dict, so you get just the .keys()

• When constructing expr, dicts are interpreted as lists of keys #9109

So the the examples effectively become:

pl.DataFrame({"a": 1}).with_columns(["a"])

pl.DataFrame({"a": 1}).with_columns(["not_exist"])
# ColumnNotFoundError: not_exist

The issue is the behavior of with_columns if you give a dictionary as input .
As @cmdlineluser explain, it only reads the keys of the dictionary and ignore the values. It is not really intuitive and error prone. I do not really see the purpose of the behavior. I would prefer that a dict is not a valid input of with_columns.

This is just a side-effect of a dictionary being an Iterable, I don't believe using a dictionary is intended.

>>> list({1: 2, 3: 4})
[1, 3]

I think we can just fix this by requiring a Sequence instead of an Iterable, no?

Edit: I don't think Sequence would work, we need generators to work here.

Just ran into this as well. It would be great for this to error rather than fail silently.

I no longer think that this should raise. This is just user error, and the fact that python yields the keys when a dictionary is iterated is simply something to watch out for.

Oh, I hadn't realised that with_columns could accept a list / iterable at all, with names being taken from the expressions (so .with_columns(["x"]) is valid). I had thought that with_columns required a name -> value mapping.

We should just raise an error when a dictionary input is given

It's possible to correctly use a dictionary for with_columns() either of these syntaxes would work -

>>> data = {"a": pl.col("a") * 2}
>>> pl.select(a=1).with_columns(**data)
shape: (1, 1)
┌─────┐
│ a   │
│ --- │
│ i32 │
╞═════╡
│ 2   │
└─────┘
>>> pl.select(a=1).with_columns(expr.alias(name) for name, expr in data.items())
shape: (1, 1)
┌─────┐
│ a   │
│ --- │
│ i32 │
╞═════╡
│ 2   │
└─────┘

Hi team, can I take up this issue and come up with a fix for it?

Hi team, can I take up this issue and come up with a fix for it?

Feel free to