H
hour of the day (24-hour clock)
H
minimal-digit hour. Example: 0 for midnight.
HH
two-digit hour, zero-padded if needed. Example: 00 for midnight.
To my understanding H would match between 0 to 23, but it doesn't if there is not a separator between hours, minutes, seconds and milliseconds (but it does match if I have separatos like H:mm:ss.SSS). So I guess that the problem I'm having is related to the way that the H match the pattern when there is not a separator.
I think the problem is in the underlying Joda library that the date filter uses. When it sees H the builder looks for one or two digits -- it is not limited to one. That will consume '94', which is not a valid hour.
You could use mutate+gsub to change the timestamp to "20210128 9.4501065" and then parse that using "yyyyMMdd H.mmssSSS"
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.