> On Nov. 14, 2013, 1:44 a.m., Thejas Nair wrote: > > serde/src/java/org/apache/hadoop/hive/serde2/lazy/LazyTimestamp.java, line > > 89 > > <https://reviews.apache.org/r/15105/diff/2/?file=374513#file374513line89> > > > > Catching an exception and converting it will be extremely slow. > >
Changed to use Pattern/Matcher to detect the timestamp format. - Venki ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/15105/#review28853 ----------------------------------------------------------- On Feb. 6, 2014, 10:37 p.m., Venki Korukanti wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/15105/ > ----------------------------------------------------------- > > (Updated Feb. 6, 2014, 10:37 p.m.) > > > Review request for hive and Mark Grover. > > > Bugs: HIVE-3844 > https://issues.apache.org/jira/browse/HIVE-3844 > > > Repository: hive-git > > > Description > ------- > > Currently LazyTimestamp can only interpret timestamps in JDBC format. If > there are UNIX epoch styles timestamps it treats them as invalid and output > NULL. This patch is to support all three types of formats mentioned here > https://cwiki.apache.org/confluence/display/Hive/LanguageManual+Types#LanguageManualTypes-Timestamps. > > BigDecimal is used to avoid rounding off errors with atoi/atof type of > conversions. > > > Diffs > ----- > > data/files/timestamp_data.txt PRE-CREATION > ql/src/test/queries/clientpositive/timestamp_4.q PRE-CREATION > ql/src/test/queries/clientpositive/timestamp_null.q 36f3541 > ql/src/test/results/clientpositive/timestamp_4.q.out PRE-CREATION > ql/src/test/results/clientpositive/timestamp_null.q.out 57269d7 > serde/src/java/org/apache/hadoop/hive/serde2/io/TimestampWritable.java > 435d6c6 > serde/src/java/org/apache/hadoop/hive/serde2/lazy/LazyTimestamp.java > 27895c5 > > Diff: https://reviews.apache.org/r/15105/diff/ > > > Testing > ------- > > Fix includes a unittest which reads from a file that has most types of > timestamp formats. > > > Thanks, > > Venki Korukanti > >
