[NUTCH-2642] MoreIndexingFilter parses ISO 8601 UTC dates in local time zone

• Type: Bug
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 2.3.1, 1.14, 1.15
• Fix Version/s: 2.4, 1.16
• Component/s: indexer, plugin
• Assignee: Unassigned
• Reporter: John Lacey
• Created: 05/Sep/18 08:25
• Updated: 28/Jan/21 13:56
• Resolved: 07/Oct/18 17:10

Description

The ISO 8601 pattern in MoreIndexingFilter.getTime is "yyyy-MM-dd'T'HH:mm:ss'Z'". Note the literal Z.

https://github.com/apache/nutch/blob/b834b81/src/plugin/index-more/src/java/org/apache/nutch/indexer/more/MoreIndexingFilter.java#L142

Apache commons-lang's DateUtils uses the local time zone by default when parsing, and can't tell that a string matching this pattern is specifying an offset because the pattern doesn't have an offset, just a literal "Z":

https://github.com/apache/commons-lang/blob/b610707/src/main/java/org/apache/commons/lang3/time/DateUtils.java#L370

So, when parsing a date string such as "2018-09-04T12:34:56Z", the time is returned as a local time:

DateUtils.parseDate("2018-09-04T12:34:56Z", new String[] { "yyyy-MM-dd'T'HH:mm:ss'Z'" })
=> Tue Sep 04 12:34:56 PDT 2018 (1536089696000)

I think a reasonable fix would be to specify an offset pattern instead of a literal "Z": "yyyy-MM-dd'T'HH:mm:ssXXX". That would also allow arbitrary offsets, as well as "Z":

DateUtils.parseDate("2018-09-04T12:34:56Z", new String[] { "yyyy-MM-dd'T'HH:mm:ssXXX" })
=> Tue Sep 04 05:34:56 PDT 2018 (1536064496000)

Issue Links

• links to: GitHub Pull Request #385
• links to: GitHub Pull Request #386