[
https://issues.apache.org/jira/browse/NIFI-14106?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17941715#comment-17941715
]
ASF subversion and git services commented on NIFI-14106:
--------------------------------------------------------
Commit c31691eecf2870dac6ed341d03ad2b782249d920 in nifi's branch
refs/heads/main from dan-s1
[ https://gitbox.apache.org/repos/asf?p=nifi.git;h=c31691eecf ]
NIFI-14106 Updated Cell Copy Policy in SplitExcel to Preserve Date and Time
Precision (#9853)
- Updated following changes implemented in Apache POI 5.4.1
Signed-off-by: David Handermann <[email protected]>
> Loss of dates and times precision when running Excel with Dates through
> SplitExcel
> ----------------------------------------------------------------------------------
>
> Key: NIFI-14106
> URL: https://issues.apache.org/jira/browse/NIFI-14106
> Project: Apache NiFi
> Issue Type: Bug
> Reporter: Daniel Stieglitz
> Assignee: Daniel Stieglitz
> Priority: Major
> Attachments: fileWithIssue.xlsx
>
> Time Spent: 3h 10m
> Remaining Estimate: 0h
>
> The SplitExcel processor seems to be dropping date format information needed
> in order to correctly determine dates and times. I set up two flows sending
> the attached file to both:
> +Flow without SplitExcel+
> GetFile -> ConvertRecord (configured with ExcelReader with Schema Access
> Strategy 'Use Starting Row' and CSVRecordSetWriter (configured with Schema
> Access Strategy 'Inherit Record Schema', Date Format MM/dd/yyyy, Time Format
> HH:mm:ss and Timestamp Format MM/dd/yyyy'T'hh:mm:ss)
> +Flow with SplitExcel+
> GetFile -> SplitExcel -> ConvertRecord (configured with ExcelReader with
> Schema Access Strategy 'Use Starting Row' and CSVRecordSetWriter (configured
> with Schema Access Strategy 'Inherit Record Schema', Date Format MM/dd/yyyy,
> Time Format HH:mm:ss and Timestamp Format MM/dd/yyyy'T'hh:mm:ss)
> The contents from the flow without SplitExcel correctly outputs the dates and
> times
> {code:java}
> transaction_id,transaction_date,transaction_time
> 75,01/01/2023,09:53:44
> 78,01/01/2023,09:55:16
> 80,01/01/2023,10:00:39
> 81,01/01/2023,10:03:55
> 82,01/01/2023,10:14:49{code}
> and its schema is
> {code:java}
> {"type":"record","name":"nifiRecord","namespace":"org.apache.nifi","fields":[{"name":"transaction_id","type":["long","null"]},{"name":"transaction_date","type":[{"type":"int","logicalType":"date"},"null"]},{"name":"transaction_time","type":[{"type":"int","logicalType":"time-millis"},"null"]}]}{code}
> while the contents from the flow with SplitExcel does not correctly output
> the dates and times
> {code:java}
> transaction_id,transaction_date,transaction_time
> 75,44927,-1
> 78,44927,-1
> 80,44927,-1
> 81,44927,-1
> 82,44927,-1{code}
> and its schema is
> {code:java}
> {"type":"record","name":"nifiRecord","namespace":"org.apache.nifi","fields":[{"name":"transaction_id","type":["long","null"]},{"name":"transaction_date","type":["long","null"]},{"name":"transaction_time","type":["long","null"]}]}{code}
> Even if the schema for the SplitExcel flow is specified as the one produced
> by flow without the SplitExcel processor and the header from the attached
> file is removed, the results are not the correct date and times
> {code:java}
> 75,01/01/1970,23:59:59
> 78,01/01/1970,23:59:59
> 80,01/01/1970,23:59:59
> 81,01/01/1970,23:59:59
> 82,01/01/1970,23:59:59{code}
> It would seem when copying the contents from the original tab into a new
> Excel spreadsheet (with one tab) that for all dates/times/timestamps the
> correct epoch must be copied over.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
