Misleading error message when an invalid y column name is provided to px.line

[joshHug]

When attempting to plot multiple y values using px.line (or similar), erroneous column names lead to a misleading error message that states that the length of the arguments is incorrect, when in fact the problem is that there the column name is wrong.

Simple to reproduce:

import plotly.express as px
import pandas as pd
import numpy as np
x = np.linspace(0, 10, 100)
data1 = x**2
data2 = x**3
df = pd.DataFrame({"x": x, "data1": data1, "data2": data2})
px.line(df, x = "x", y = ["data1", "data2"]) # works fine
px.line(df, x = "x", y = ["data1", "Data2"]) 

# this last line gives a misleading error message: 
# ValueError: All arguments should have the same length. The length of argument `y` is 2, 
# whereas the length of  previously-processed arguments ['x'] is 100

[nicolaskruchten]

Yes, we should definitely find a way to give you a better hint there that there's a typo in the contents of y. It's a bit tricky to do and we're love some help to figure out how to do it! The relevant code is near here: https://github.com/plotly/plotly.py/blob/master/packages/python/plotly/plotly/express/_core.py#L1247

[nicolaskruchten]

The reason it's tricky is that we land deep in that if block based on some determination that happened way ahead of time here https://github.com/plotly/plotly.py/blob/master/packages/python/plotly/plotly/express/_core.py#L1315 ... basically up front PX decides if it's in "wide mode" based on whether or not the contents of y are all column names, and then much later if they're not, it checks the length of the data y points to against the length of the data x points to.

[nicolaskruchten]

See also #3359 and #3474 and #2586

[abarpan3]

Hi, was exploring this issue a bit. So the build_dataframe function has another function is_col_list to check if the supplied argument for x/y is a list or not (hence wide-form) https://github.com/plotly/plotly.py/blob/master/packages/python/plotly/plotly/express/_core.py#L1339

Now this is_col_list function has a check inside it to see if a column is present in the dataframe, and if not it returns wide_mode as false.
https://github.com/plotly/plotly.py/blob/master/packages/python/plotly/plotly/express/_core.py#L1036

Here, in the is_col_list functionwe can catch the erroneous column name and give the user a more helpful error message. So I tried to make 2 changes,

1. Line 1035 : Checking if y value is of pd.Series type and returning wide_mode as false, if the condition is fulfilled. This takes care of any direct array inputs in x/y arguments like below example.
   px.scatter(df1,x = df1.sales, y = df2.market)

2. **Line 1042 - 1048 ** : Checking additionally if col_name supplied in list is present in df. And returning a more helpful error message. Attached image shows both changes

Could anyone kindly provide any feedback if this seems like the right approach?