corruption loop for a data frame in PySpark - pyspark

Question

I'm making a loop for corrupt my dataset on PySpark, and i want to control the errors.
First: I made a list of type: 9 errors and 1 with no error.
#code_erreur= [ "replace","inverse","inserte","delete","espace","NA"]
code_erreur= ["replace"]
varibale =["VARIABLEA","VARIABLEB"]
I start with just code the error : replace.
Select a random letter, on a random "varibale ", to replace by antoher random letter .
My input:
VARIABLEA  | VARIABLEB
BLUE        | WHITE
PINK         | DARK
My expected output:
VARIABLEA  | VARIABLEB
BLTE        | WHITE
PINK         | DARM
And I made a loop:
def algo_corruption(lettre,code_erreur,nombre_erreur,varibale,data):
alp=(list(string.ascii_uppercase))
table_corruption=[]
for i in range(1,data.count()):
code_erreur_choisi =random.choice(code_erreur)
varibale_choisie =random.choice(varibale)
(table_corruption.append((code_erreur_choisi,
varibale_choisie)))
cols=["code_erreur_choisi","varibale_choisie"]
result = spark.createDataFrame(table_corruption, cols)
result= result.withColumn("id", monotonically_increasing_id())
data= data.withColumn("id", monotonically_increasing_id())
data_join_result= data.join(result, "id","inner").drop("id")
for j in range(1,data_join_result.count()):
if data_join_result.filter(col("code_erreur_choisi") == "replace"):
data_corrp = (data_join_result[varibale_choisie].replace(random.choice(data_join_result.collect()[j][varibale_choisie]),
random.choice(alp))
display(data_corrp)
else:
print("erreur pas encore codée")
But that doesn't work, I always have errors like:
ValueError: R
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<command-3614001298541202> in <module>()
----> 1 algo_corruption(code_erreur,varibale,extrac_base_train)
<command-1181487391956055> in algo_corruption(code_erreur, varibale, data)
29 if data_join_result.filter(col("code_erreur_choisi") == "replace"):
30 data_corrp = (data_join_result[varibale_choisie].replace(random.choice(data_join_result.collect()[j][varibale_choisie]),
---> 31 data_join_result.collect()[j][lettre_choisie]))
32 display(data_corrp)
33
/databricks/spark/python/pyspark/sql/types.py in __getitem__(self, item)
1517 raise KeyError(item)
1518 except ValueError:
-> 1519 raise ValueError(item)
1520
1521 def __getattr__(self, item):
ValueError: R