https://community.cloudera.com/t5/Support-Questions/Impala-table-definition/m-p/66666#M13689 <P>Hi&nbsp;<a href="https://community.cloudera.com/t5/user/viewprofilepage/user-id/24935">@toamitjain</a>,<BR /><BR />Firstly, I advice you to create the external -TEXTFILE- table too in <U>impala</U> (<EM>it's faster to create it in hive!</EM>).</P><PRE>CREATE EXTERNAL TABLE table1 (columns def) ROW FORMAT DELIMITED FIELDS TERMINATED BY ',' LOCATION 'hdfs_path';</PRE><P><BR />1- Because of your initial hdfs file probably is a text format file while the parquet tables use the parquet format, so you have do this recreation to enssure the <STRONG>convertision TEXT=&gt;PARQUET</STRONG> (but if your HDFS files are already in parquet format you can create an direct external parquet table on it..).<BR /><BR />2- E<SPAN class="short_text"><SPAN>xcept the <STRONG>slowness</STRONG> of your <STRONG>queries</STRONG> (because parquet table is very fast than text table) there is no issue, just the <STRONG>table</STRONG> and </SPAN></SPAN><SPAN class="short_text"><SPAN>the <STRONG>hdfs</STRONG> files will be <STRONG>separated</STRONG>..<BR /><BR />3- I think storing the date on&nbsp;<STRONG>timestamps</STRONG> is more profetionnal, so you can benifit from the timestamp <STRONG>functions</STRONG>..<BR /><BR />Good luck.<BR /></SPAN></SPAN></P><P><BR /><BR /></P> Tue, 24 Apr 2018 10:05:21 GMT AcharkiMed 2018-04-24T10:05:21Z