problem "fit" with current relation

[lepelletieralexandre]

Hello,

There is a bug when I use the "fit" method with "LogisticRegression" model.
In fact, I have this error :
MissingRelation: Severity: ERROR, Message: Relation "SCHEMA.TVUL_STAT_titanic_CHAID_INT_SANS_NA_ SELECT ION_OK_DECOUP_MAIN_TRAIN_TEST_train_DUMMIES_MODELE_titanic_test" does not exist, Sqlstate: 42V01, Routine: throwRelationDoesNotExist,

There are 2 spaces between "SELECT" for "SELECTION" word which have been added.
I found where the problem comes from.
In "fit" method here : https://github.com/vertica/VerticaPy/blob/master/verticapy/machine_learning/vertica/base.py , line 2354, you use "self.input_relation = input_relation.current_relation()". But, by default, the parameter "reindent" in "current_relation" function is TRUE.
line 432 here -> https://github.com/vertica/VerticaPy/blob/master/verticapy/core/vdataframe/_sys.py#L432 , you call "indent_vpy_sql" function, which is define here line 1219 -> https://github.com/vertica/VerticaPy/blob/master/verticapy/_utils/_sql/_format.py , where spaces are adds between "SELECT" word.

I think, when we use "fit" fonction, the "reindent" parameter from "current_relation" need to be False. And, in general, all function which call "current_relation" function to capture "input_relation" in the code need to define False value to "reindent" parameter ;
for exemple, line 7837 for class Unsupervised(VerticaModel).

Can you fix this problem?

Best regards,
Alexandre

[mail4umar]

Can you please tell me which version are you using? Because I am unable to reproduce your error on the latest branch.

from verticapy import vDataFrame
from verticapy.machine_learning.vertica.linear_model import LogisticRegression
from verticapy._utils._sql._sys import _executeSQL

# Create sample data
df = vDataFrame({
    "survived": [0, 1, 1, 0, 1],
    "age": [22, 38, 26, 35, 28],
    "fare": [7.25, 71.28, 7.92, 53.1, 8.05]
})

# Push to Vertica DB with a long name that includes "SELECTION"
long_table_name = "TVUL_STAT_titanic_CHAID_INT_SANS_NA_SELECTION_OK_DEMO_2"
df.to_db(long_table_name)

# Define the Logistic Regression model
model = LogisticRegression("titanic_model_selection_bug_2")

# Fit the model (this would also trigger the same splitting internally)
try:
    model.fit(input_relation=df, X=["age", "fare"], y="survived")
except Exception as e:
    print("Expected error during fit():")
    print(e)

[lepelletieralexandre]

Hello,

It's normal that you don't have the problem, because you don't reload the table "TVUL_STAT_titanic_CHAID_INT_SANS_NA_SELECTION_OK_DEMO_2", and the current_relation don't have "SELECTION" in the name.

you need to write these lines :
df.to_db(long_table_name)
df = vDataFrame("TVUL_STAT_titanic_CHAID_INT_SANS_NA_SELECTION_OK_DEMO_2")

If I take your exemple and I had these lines, I have the problem,

My verticapy version is 1.0.5

best regards