Issue 15711

Event date is interpreted as January 1st when month or day omitted in the verbatim record

15711
Reporter: fmendez
Assignee: cgendreau
Type: Bug
Summary: Event date is interpreted as January 1st when month or day omitted in the verbatim record
Priority: Major
Status: Open
Created: 2014-05-28 12:45:11.425
Updated: 2016-02-05 16:21:45.349
        
Description: An occurrence record like [gbifid: 500544056, v_eventdate:	1979, eventdate	: 01/01/79 00:00, v_year: 1979, year:1979, v_month:0, month:1, v_day:0, day:1] contains an incorrect interpretation of the eventdate field; the month and day have been interpreted as Jan 1st. The eventdate shouldn't be interpreted when there's an omission of year,month or day in the verbatim version of the record.

Note: an attachment contains sample records that show the issue.]]>
    
Attachment query_result_jan_1.xlsx


Author: omeyn@gbif.org
Comment: More analysis reveals that approx 8.1M records have an eventDate of Jan 1, but approx 7.9M of those are Jan 1 in the verbatim record (given either as day and month or as part of the verbatim dateString). There is definitely a bug in our handling of incomplete dateStrings (i.e. year only or year-month only) but is probably only responsible for approx 200k of the Jan 1 spike.
Created: 2014-05-29 17:01:10.935
Updated: 2014-05-29 17:01:10.935


Author: omeyn@gbif.org
Comment: I've added tests to the TemporalInterpreterTest of occ project that show this problem. Fix is TODO.
Created: 2014-06-16 14:02:49.268
Updated: 2014-06-16 14:02:49.268