[
https://issues.apache.org/jira/browse/DRILL-8071?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17456752#comment-17456752
]
ASF GitHub Bot commented on DRILL-8071:
---------------------------------------
cgivre commented on pull request #2400:
URL: https://github.com/apache/drill/pull/2400#issuecomment-990319067
@pjfanning
What if we added a boolean config option `nullOnInvalidDate` or something
like that, and we could let the user decide what they want? Personally, I like
the idea of being able to read through the data w/o error even when it is
invalid.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
> format-excel should use POI DataFormatter
> -----------------------------------------
>
> Key: DRILL-8071
> URL: https://issues.apache.org/jira/browse/DRILL-8071
> Project: Apache Drill
> Issue Type: Improvement
> Components: Execution - Data Types
> Affects Versions: 1.19.0
> Reporter: PJ Fanning
> Priority: Major
>
> The existing ExcelBatchReader uses the raw data values from the cells. This
> raw data ignores formatting set on the cells. As an example, numbers and
> dates are stored as doubles. With the POI DataFormatter, you can get the cell
> style applied so that the data will appear as it does when you view the data
> in Excel itself.
> [https://poi.apache.org/apidocs/dev/org/apache/poi/ss/usermodel/DataFormatter.html#formatCellValue-org.apache.poi.ss.usermodel.Cell-]
>
> A big number like 123456789.987654 could be stored as double that is more
> like 123456789.9876539999999 when represented in decimal format (because this
> might be the closest match that double can represent). The cell format could
> say that cell has 6 decimal places after the decimal point so the formatter
> would round the number back to the value that it displayed in Excel as.
> Even if you choose not to use the DataFormatter, you have unprotected calls
> to `cell.getNumericCellValue()` and that could easily throw an exception (if
> the data is not stored a number). Even `cell.getStringCellValue()` can throw
> an exception - for similar reasons.
>
> There is also custom code for handling the conversion of the raw numbers
> representing dates/timestamps but this also seems like a bad idea. The Cell
> class has getLocalDateTimeCellValue and this has the right logic for
> converting 1904 and 1900 based dates - yes, Excel uses 2 different formats.
> Code that processes excel files is a real pain to get right because the
> Microsoft storage format is really bad.
>
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
