Stefania created CASSANDRA-12863:
------------------------------------
Summary: cqlsh COPY FROM cannot parse timestamp in partition key
if table contains a counter value
Key: CASSANDRA-12863
URL: https://issues.apache.org/jira/browse/CASSANDRA-12863
Project: Cassandra
Issue Type: Bug
Components: Tools
Reporter: Stefania
Assignee: Stefania
Fix For: 2.2.x, 3.0.x, 3.x
This sample table:
{code}
CREATE TABLE test (columnname text, day timestamp, israndom boolean,
columnvalue text, counter counter, PRIMARY KEY ((columnname, day, israndom),
columnvalue));
{code}
with this sample data:
{code}
origins|2016-10-01 00:00:00+0000|False|ACTUAL|6
origins|2016-10-01 00:00:00+0000|False|ADGMOB|4
origins|2016-10-01 00:00:00+0000|False|ANONPM|4
origins|2016-10-01 00:00:00+0000|False|CSRT2L|76
origins|2016-10-01 00:00:00+0000|False|DIAGOP|18
origins|2016-10-01 00:00:00+0000|False|E-SOFT|17
origins|2016-10-01 00:00:00+0000|False|E-TASK|10
{code}
when imported with
{code}
COPY ks.test FROM 'test.csv' WITH DELIMITER = '|';
{code}
will generate a parse error:
{code}
Failed to import 7 rows: ParseError - can't interpret u"'2016-10-01
00:00:00+0000'" as a date with this format: %Y-%m-%d %H:%M:%S%z, given up
without retries
{code}
The problem is that when a counter value is present, we don't use prepared
statements and hence add quotes to certain types, such as timestamps. However,
because the timestamp is part of the partition key, we must parse it in order
to determine the routing token. Here lies the problem, we do not remove the
quotes before parsing the partition key, therefore ending up with a parse error.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
