从 Pandas Dataframe 到 pyodbc (Azure SQL DB) 使用 ZF3D386AADEEB6063D016E4DFEE 转换字符串时从字符串转换失败：

Question

I'm trying to load Salesforce data to Azure SQL Database incrementally by launching a Python script on Azure Databricks.

Since I'm not able to install Devart ODBC in Azure Databricks, I'm trying to use simple_salesforce to get data from salesforce:

import pandas as pd
import pyodbc
from simple_salesforce import Salesforce, SalesforceLogin, SFType
from sqlalchemy.types import Integer, Text, String, DateTime
from sqlalchemy import create_engine
import urllib

sf = Salesforce(password = password, username=username, security_token=jeton)
rep_qr = "SELECT SOMETHING FROM Account WHERE CONDITION"
soql = prep_qr.format(','.join(field_names))
results = sf.query_all(soql)['records']

I get the following result (an example):

[OrderedDict([('attributes', OrderedDict([('type', 'Account'), ('url', '/services/data/v42.0/sobjects/Account/0014K000009aoU3QAI')])), ('Id', XY1), (Name, Y), (Date, 2020-11-24T09:16:17.000+0000)])]

Then I converted the output to a pandas Dataframe:

results = pd.DataFrame(sf.query_all(soql)['records'])
results.drop(columns=['attributes'], inplace=True) #to keep only the columns

I got something like this (just an example):

Id	Name	Date
XY1	Y	2020-11-24T09:16:17.000+0000

In order to ingest this data into Azure SQL Database I have used "sqlalchemy" to convert the Dataframe into sql, then pyodbc will take in charge the insertion part into the destination (Azure SQL Database), as shown bellow:

df = pd.DataFrame(results)
df.reset_index(drop=True, inplace=True) #just to remove the index from dataframe

#Creating the engine from and pyodbc which is connected to Azure SQL Database:
params = urllib.parse.quote_plus \
                (r'DRIVER={ODBC Driver 17 for SQL Server};SERVER=' + server + ';DATABASE=' + database + ';UID=' + username + ';PWD=' + password)
conn_str = 'mssql+pyodbc:///?odbc_connect={}'.format(params)
engine_azure = create_engine(conn_str, echo=True)
df.to_sql('account',engine_azure,if_exists='append', index=False)

But I get the following error:

sqlalchemy.exc.DataError: (pyodbc.DataError) ('22007', '[22007] [Microsoft][ODBC Driver 17 for SQL Server][SQL Server]Conversion failed when converting date and/or time from character string. (241) (SQLExecDirectW)')

I think the problem is the library simple_salesforce brigns the date/time in this format:

2020-11-24T09:16:17.000+0000

But in Azure SQL Database it should be something like this:

2020-11-24T09:16:17.000

The problem here is I'm loading the tables dynamically (I don't even know the tables nor the columns that I'm loading) the reason why I can't cast these data type, I need a way to pass datatype to pyodbc automatically.

What can you recommend please?

Thanks,

Answer 1

If the date/time values are consistently returned as strings of the form 2020-11-24T11:22:33.000+0000 then you can use pandas' .apply() method to convert the strings to the 2020-11-24 11:22:33.000 format that SQL Server will accept:

df = pd.DataFrame(
    [
        (1, "2020-11-24T11:22:33.000+0000"),
        (2, None),
        (3, "2020-11-24T12:13:14.000+0000"),
    ],
    columns=["id", "dtm"],
)
print(df)
"""console output:
   id                           dtm
0   1  2020-11-24T11:22:33.000+0000
1   2                          None
2   3  2020-11-24T12:13:14.000+0000
"""

df["dtm"] = df["dtm"].apply(lambda x: x[:23].replace("T", " ") if x else None)
print(df)
"""console output:
   id                      dtm
0   1  2020-11-24 11:22:33.000
1   2                     None
2   3  2020-11-24 12:13:14.000
"""

df.to_sql(
    table_name,
    engine,
    index=False,
    if_exists="append",
)

with engine.begin() as conn:
    pprint(conn.execute(sa.text(f"SELECT * FROM {table_name}")).fetchall())
"""console output:
[(1, datetime.datetime(2020, 11, 24, 11, 22, 33)),
 (2, None),
 (3, datetime.datetime(2020, 11, 24, 12, 13, 14))]
"""