I'd expect ORDER BY trunc(ts, "DD") to work. E.g. on my system this works:   [localhost:21000] > select timestamp_col, tinyint_col from functional_hbase.alltypestiny order by trunc(timestamp_col, 'DD'), tinyint_col desc; +---------------------+-------------+ | timestamp_col | tinyint_col | +---------------------+-------------+ | 2009-01-01 00:01:00 | 1 | | 2009-01-01 00:00:00 | 0 | | 2009-02-01 00:01:00 | 1 | | 2009-02-01 00:00:00 | 0 | | 2009-03-01 00:01:00 | 1 | | 2009-03-01 00:00:00 | 0 | | 2009-04-01 00:01:00 | 1 | | 2009-04-01 00:00:00 | 0 | +---------------------+-------------+ ... View more

If you want to implement a C++ UDF though, I'd recommend starting with the docs here: https://www.cloudera.com/documentation/enterprise/latest/topics/impala_udf.html. There are some examples of string manipulation UDFs on that page. ... View more

  [localhost:21000] > select concat(substring(l_comment, 1, 3), regexp_replace(substring(l_comment, 4, length(l_comment) - 3), '[^ ]', '*'), substring(l_comment, length(l_comment) - 3)) from tpch.lineitem limit 5; Query: select concat(substring(l_comment, 1, 3), regexp_replace(substring(l_comment, 4, length(l_comment) - 3), '[^ ]', '*'), substring(l_comment, length(l_comment) - 3)) from tpch.lineitem limit 5 Query submitted at: 2017-09-25 07:38:09 (Coordinator: http://tarmstrong-box:25000) Query progress can be monitored at: http://tarmstrong-box:25000/query_plan?query_id=34f1a993e3cb99a:51d89bf800000000 +--------------------------------------------------------------------------------------------------------------------------------------------------------------+ | concat(substring(l_comment, 1, 3), regexp_replace(substring(l_comment, 4, length(l_comment) - 3), '[^ ]', '*'), substring(l_comment, length(l_comment) - 3)) | +--------------------------------------------------------------------------------------------------------------------------------------------------------------+ | egu*** ****** ***** *** the | | ly ***** ************* ***** **** old | | rio***** ******** ******* *** dep | | lit*** ******** **** **n de | | pe***** ****** ***** **y re | +--------------------------------------------------------------------------------------------------------------------------------------------------------------+   I'd recommend doing it with builtin functions since it will be easier to maintain. I included an example above of how you might do it using regexp_replace. I expect it will be quite fast - Impala's query compilation can inline functions like length() and substring() so those are essentially free in Impala (unlike many other SQL engines). The main cost is regexp_replace() but I'd expect that to be quite fast too.   ... View more

Impala doesn't have a date data type, it does have a timestamp type though: https://www.cloudera.com/documentation/enterprise/latest/topics/impala_timestamp.html. to_date() converts timestamp to a string.https://www.cloudera.com/documentation/enterprise/latest/topics/impala_datetime_functions.html   If you want to remove the time portion of the timestamp you can use trunc(ts, "DD") to get the timestamp of midnight on that day: https://www.cloudera.com/documentation/enterprise/latest/topics/impala_datetime_functions.html   I don't understand your question about "order by". Order by should work for all scalar data types in Impala (timestamp, double, etc). What do you want it to do and what is it doing now? ... View more

The user that Impala is running under needs to be able to remove and recreate the scratch directory at startup (i.e. /tmp/impala-scratch). This is done to ensure that the directory is free of old files and that Impala has ownership of the directory.   Based on the log message that user doesn't have the required permissions to do that.   I suspect if you just delete that directory and let Impala create it at startup, that will solve your problem. ... View more

The Impala daemon wasn't able to set up the scratch directories during startup. The reason will be logged in one of the impalad*.WARNING logs, probably one of the first messages in there. ... View more

I'm not sure there are risks specifically. The best practice is to use Cloudera manager to configure memory limits for different services, so this is the right way to configure things.   Cloudera Manager does have support to help set up memory limits for applications: https://www.cloudera.com/documentation/enterprise/latest/topics/cm_mc_autoconfig.html#concept_xjy_vb3_rn .    For a production system, it's important to put thought into how much memory your system needs and how it's allocated between different services.  E.g. as an earlier poster saw, 256MB is not enough memory to do much interesting with Impala.     ... View more