Unusual REV value format not detected as Revision

[rfc2822]

Hello,

The following REV line (part of a VCard 4) seems not to be parsed as Revision by ez-vcard:

REV;VALUE=timestamp:2016-11-19T17:44:22.057Z

However, ISO.8601.2004 (which is referenced by the VCard4 REV/TIMESTAMP definition)

• allows the following format: YY-MM-DDThh:mm:ssZ,
• says that "If necessary for a particular application a decimal fraction of hour, minute or second may be included."

I don't know whether this really applies to REV values, but I think the above value is allowed and should be parsed by correctly.

Further reference: https://forums.bitfire.at/topic/1259/davdroid-1-3-4-1-ose-nextcloud-11-0-beta-carddav-sync-error/

[rfc2822]

It seems that there are many ways to express date and time, see https://forums.bitfire.at/post/7504

I guess you will have to import a whole library or something like that to understand all these crazy formats … but a good start would be to parse the format with the milliseconds, too.

[mangstadt]

Note: I can't seem to find a freely available copy of ISO.8601.2004 online. I found something that looks like it, but not 100% sure it's exactly the same.

Section 4.3.2, which the vCard TIMESTAMP data type adheres to, doesn't say anything about decimal fractions, so my opinion is that REV doesn't have to support it. But what vCard producers choose to do in the wild is another thing, so it might be worth supporting if a lot of producers decide to break with the spec.

As you alluded to in your forum post, it appears there are other ways of representing a timestamp, which ez-vcard does not account for (see below). So yeah, that's a thing.

The time elements of a date and time of day expression shall be written in the following sequence.
a) For calendar dates: year – month – day of the month – time designator – hour – minute – second – zone designator
b) For ordinal dates: year – day of the year – time designator – hour – minute – second – zone designator
c) For week dates: year – week designator – week – day of the week – time designator – hour – minute – second – zone designator

The Appendix B.2.3 you mention contains examples for reduced accuracy date formats, which are not applicable to the TIMESTAMP data type.

[rfc2822]

In my tests, there are other time formats which don't work. Currently, Nextcloud Contacts uses the REV:20161218T201900 format, which is not recognized by ez-vcard. I have used this test class which tests the four formats mentioned in RFC 6350 4.3.5 TIMESTAMP:

package at.bitfire.vcard4android;

import static org.junit.Assert.*;
import org.junit.Test;

import ezvcard.Ezvcard;
import ezvcard.VCard;

public class EzVCardTest {

    @Test
    public void testREV_UTC() {
        VCard vCard = Ezvcard.parse("BEGIN:VCARD\r\n" +
                "VERSION:4.0\r\n" +
                "REV:20161218T201900Z\r\n" +
                "END:VCARD").first();
        assertNotNull(vCard.getRevision());
    }

    @Test
    public void testREV_WithoutTZ() {
        VCard vCard = Ezvcard.parse("BEGIN:VCARD\r\n" +
                "VERSION:4.0\r\n" +
                "REV:20161218T201900\r\n" +
                "END:VCARD").first();
        assertNotNull(vCard.getRevision());
    }

    @Test
    public void testREV_TZHourOffset() {
        VCard vCard = Ezvcard.parse("BEGIN:VCARD\r\n" +
                "VERSION:4.0\r\n" +
                "REV:20161218T201900-05\r\n" +
                "END:VCARD").first();
        assertNotNull(vCard.getRevision());
    }

    @Test
    public void testREV_TZHourAndMinOffset() {
        VCard vCard = Ezvcard.parse("BEGIN:VCARD\r\n" +
                "VERSION:4.0\r\n" +
                "REV:20161218T201900-0530\r\n" +
                "END:VCARD").first();
        assertNotNull(vCard.getRevision());
    }

}

and tests fail for

• testREV_TZHourOffset and
• testREV_WithoutTZ

[mangstadt]

Added support for the above date formats that failed in 9a6977e. Timestamps without UTC offsets will be assumed to be in the local computer's timezone.

[rfc2822]

Thank you very much!

[zeehio]

Thanks!

If I read the code properly, in the end timestamps with milliseconds do not seem to be supported. Let's hope no client uses them (nextcloud/contacts currently is using them) nextcloud/contacts#65

[mangstadt]

Ok, I will add that. Can you remind me what that timestamp format looks like?

[zeehio]

20161227T155712.034Z

I would express it like this:

\d{4}\d{2}\d{2}T\d{2}\d{2}\d{2}(\.\d*)?Z

[mangstadt]

Will the milliseconds always be 3 digits long?

[zeehio]

No, 1-6 digits can be expected

[mangstadt]

Can you elaborate? Thanks. :)

[zeehio]

http://stackoverflow.com/questions/25842840/representing-fraction-of-second-with-iso-86012004

The ISO does not specify the number of digits, however I haven't seen more than six ever.

[mangstadt]

Added in 27f4404. If it's greater than 3 digits, then it is rounded to the nearest millisecond.

[zeehio]

Thanks!