[Python] azure-kusto-data insert duplicated and corrupted data when some pattern '2_' in it

Using azure-kusto-data@5.0.1 when running an ingest query using both URI or Ingest URI, it ingests a column with value null as soon as the previous field contains the characters 2_ in it.

I can run the query in the azure data explorer and it works fine, but using this method it ingests null in the “value” field.

azure data explorer table schema:

{
“Name”: “kpi_table”,
“OrderedColumns”:
[{
“Name”: “server_timestamp”,
“Type”: “System.DateTime”,
“CslType”: “datetime”
},
{
“Name”: “name”,
“Type”: “System.String”,
“CslType”: “string”
},
{
“Name”: “plant_id”,
“Type”: “System. String”,
“CslType”: “string”
},
{
“Name”: “device_key”,
“Type”: “System.String”,
“CslType”: “string”
},
{
“Name”: “value”,
“Type”: “System.Double”,
“CslType”: “real”
}]
}


Data to ingest:
2025-03-13T22:06:00Z, kpi_inverter_production, #ESSSAB2, #ESSSAB2_T010JPK01KKP03, 0.0201

Data ingested:
2025-03-13T22:06:00Z, kpi_inverter_production, #ESSSAB2, #ESSSAB2_T010JPK01KKP03, 0.0201
2025-03-13T22:06:00Z, kpi_inverter_production, #ESSSAB2, #ESSSAB2_T010JPK01KKP03, null

Performs a successful ingest and 1 second later according to ingestion_time() does another ingest with value to null.

If in the field corresponding to device_key, you replace the character 2 by 3, it ingests correctly, can it detect the pattern 2_ as special characters internally?

Code:

import pandas as pd
from azure.kusto.data import KustoClient

def __init__(self, url, database, client_id, client_secret, tenant_id):

kcsb = KustoConnectionStringBuilder.with_aad_application_key_authentication(
url, client_id, client_secret, tenant_id
)
self.conn = KustoClient(kcsb)

def execute_kql_query(self, query: str):
try:
logger.debug("Executing KQL query: %s", query)
response = self.conn.execute(self.database, query)
if response.get_exceptions():
raise Exception(response.get_exceptions())
return dataframe_from_result_table(response.primary_results[0])
except Exception as error:
logger.error("Error executing KQL query: %s", error)
return pd.DataFrame()


The query is a string with an .inline ingest, and although I know it is not recommended for production environments, the error is also reproduced with streaming ingest.

Continue reading...